Amplitude, Inc. (AMPL) 2026 Q2 法說會逐字稿

Good morning, and welcome to Amplitude's second quarter 2026 earnings conference call. I'm John Streppa, Head of Investor Relations. And joining me today are Spenser Skates, CEO and Co-Founder of Amplitude; and Andrew Casey, Chief Financial Officer. During today's call, management will make forward-looking statements, including statements regarding our financial outlook for the third quarter and full-year 2026, the expected performance of oYitchuin Wongur products, our expected quarterly and long-term growth, investments and our overall future prospects.

早安，歡迎參加 Amplitude 2026 年第二季財報電話會議。我是投資人關係主管 John Streppa。今天與我一同出席的有 Amplitude 執行長暨共同創辦人 Spenser Skates，以及財務長 Andrew Casey。在今天的電話會議中，管理層將發表前瞻性陳述，包括關於我們對 2026 年第三季與全年之財務展望、oYitchuin Wongur 產品的預期表現、我們預期的季度與長期成長、投資，以及我們整體未來前景的相關陳述。

These forward-looking statements are based on current information, assumptions and expectations and are subject to risks and uncertainties, some of which are beyond our control that could cause actual results to differ materially from those described in these statements.

這些前瞻性陳述係基於目前資訊、假設與預期，並受風險與不確定性影響，其中部分超出我們的控制範圍，可能導致實際結果與這些陳述所描述者出現重大差異。

Further information on the risks that could cause actual results to differ is included in our filings with the Securities and Exchange Commission. You are cautioned not to place undue reliance on these forward-looking statements, and we assume no obligation to update these statements after today's call, except as required by law. Certain financial measures used on today's call are expressed on a non-GAAP basis.

有關可能導致實際結果出現差異之風險的更多資訊，已載於我們向美國證券交易委員會提交的文件中。敬請注意勿過度依賴這些前瞻性陳述；除法律要求外，我們不承擔在今日電話會議後更新這些陳述之義務。今日電話會議中使用的若干財務衡量指標係以非 GAAP 基礎表達。

We use these non-GAAP financial measures internally to facilitate analysis of our financial and business trends and for internal planning and forecasting purposes. These non-GAAP financial measures have limitations and should not be used in isolation from or as a substitute for financial information prepared in accordance with GAAP.

我們在內部使用這些非 GAAP 財務衡量指標，以利分析我們的財務與業務趨勢，並用於內部規劃與預測目的。這些非 GAAP 財務衡量指標具有其限制，不應單獨使用，亦不應作為依 GAAP 編製之財務資訊的替代。

Additional information regarding these non-GAAP financial measures and a reconciliation between these GAAP and non-GAAP financial measures are included in our earnings press release and the supplemental financial information, which can be found on our Investor Relations website at investors.amplitude.com.

關於這些非 GAAP 財務衡量指標的更多資訊，以及 GAAP 與非 GAAP 財務衡量指標之間的調節表，均載於我們的財報新聞稿與補充財務資訊中，可於我們的投資人關係網站 investors.amplitude.com 查閱。

And with that, I'll hand the call over to Spenser.

接下來，我把電話交給 Spenser。

Thanks, John, and good afternoon, everyone. Welcome to Amplitude's second quarter 2026 earnings call. Today, I'll cover three things. First, our Q2 results. Second, how we transformed Amplitude into an AI company and why every company I talk to now wants to learn how they can do the same.

謝謝你，John，各位下午好。歡迎參加 Amplitude 2026 年第二季財報電話會議。今天我將涵蓋三件事。第一，我們第二季的業績結果。第二，我們如何在過去兩年將 Amplitude 轉型為一家 AI 公司，以及為何我現在交流的每一家企業都想了解他們如何也能做到同樣的事。

Third, a look at our product and a spotlight on our customers. Let me start with the numbers. Q2 revenue was $101 million, up 21% year-over-year. Total annual recurring revenue was $410 million, up 22% year-over-year and up $36 million from last quarter. That was made up of 2 parts; inorganic ARR from Statsig of $17 million and organic ARR growth of $19 million.

第三，回顧我們的產品並聚焦介紹我們的客戶。我先從數字開始。第二季營收為 1.01 億美元，年增 21%。年度經常性收入（ARR）總額為 4.10 億美元，年增 22%，較上季增加 3,600 萬美元。其中包含兩部分：來自 Statsig 的非有機 ARR 為 1,700 萬美元，以及有機 ARR 成長 1,900 萬美元。

Andrew will walk through the details. Non-GAAP operating loss was $1.5 million. Customers with more than $100,000 in ARR grew to 824, an increase of 30% year-over-year. Both AI natives and large enterprises are driving this growth. Let me step back and tell you about our transformation and then how we're helping customers along their AI journeys.

Andrew 將說明細節。非 GAAP 營業虧損為 150 萬美元。ARR 超過 10 萬美元的客戶數成長至 824 家，年增 30%。AI 原生公司與大型企業共同推動了這項成長。我先退一步，談談我們的轉型，以及我們如何協助客戶踏上他們的 AI 旅程。

We help companies build better products. Every company wants to transform to deliver software products in an AI-native way. We've made that transformation at Amplitude over the last two years, and now our customers are looking to learn from us. Becoming an AI company starts with the organization. Two years ago, we first transformed our engineering team by bringing in AI engineers who built with it for years.

我們協助企業打造更好的產品。每家公司都希望以 AI 原生的方式交付軟體產品，完成轉型。過去兩年我們已在 Amplitude 完成這項轉型，而現在客戶也希望向我們學習。成為一家 AI 公司要從組織開始。兩年前，我們先改造工程團隊，延攬已使用 AI 建構產品多年的 AI 工程師。

Then we moved into adjacent functions like product management, design and the more technical parts of go-to-market. We also brought in AI expertise through acquisition. Founders and other members of the team from these companies have taken leadership roles across Amplitude. I have focused on bringing in leaders who are former founders and who have a technical background. Gab, our Chief Product Officer, started multiple companies, including Loom Systems, which sold to ServiceNow in 2020.

接著我們擴展到相鄰職能，例如產品管理、設計，以及 go-to-market 中較技術性的部分。我們也透過併購引入 AI 專業能力。這些公司的創辦人與其他團隊成員，已在 Amplitude 各部門擔任領導職務。我專注於延攬具技術背景、且曾為創辦人的領導者。我們的產品長 Gab 曾創辦多家公司，包括 Loom Systems，該公司於 2020 年出售給 ServiceNow。

In addition, Nate, our Chief Commercial Officer, has a degree in math and physics and started his career as an engineer programming in C++ and Java and building databases. Most recently, we added Angela Ferrante as SVP of Marketing. Angela founded Laudable, which went through Y Combinator Summer 2021, sold it in 2025 and is a technical marketing leader who builds apps with AI in her spare time. In addition to all of this, we're continually reeducating everyone at Amplitude through initiatives like AI Week, unlimited token spend and a living token leaderboard. This has all resulted in 3 times the number of pull requests in 6 months.

此外，我們的商務長 Nate 擁有數學與物理學位，職涯起步於工程師，使用 C++ 與 Java 進行程式開發並建置資料庫。最近，我們延攬 Angela Ferrante 擔任行銷資深副總裁。Angela 創辦 Laudable，曾入選 Y Combinator 2021 年夏季梯次，並於 2025 年出售；她是具技術背景的行銷領導者，閒暇時也會用 AI 開發應用程式。除此之外，我們也透過 AI Week、無上限 token 支出，以及動態 token 排行榜等計畫，持續讓 Amplitude 全員再教育。這些努力使我們在 6 個月內的 pull request 數量提升至 3 倍。

We've reduced our pull request cycle from 5 hours to 44 minutes. Bug reports are down 55%. 5% of our pull requests are submitted from designers and product managers with no engineering involvement. We've leveraged AI to shorten our closing process by a day. We built customer health dashboards that enable our sellers and leaders to track customer usage, bringing our own Amplitude data alongside Salesforce data and data from other sources.

我們將 pull request 的週期從 5 小時縮短到 44 分鐘。錯誤回報下降 55%。其中有 5% 的 pull request 是由設計師與產品經理在沒有工程師參與的情況下提交。我們運用 AI 將結帳（closing）流程縮短了 1 天。我們建立了客戶健康度儀表板，讓業務與主管能追蹤客戶使用情況，將我們自家的 Amplitude 資料與 Salesforce 資料及其他來源資料一併整合。

When I talk with our customers, they are all focused on how they can transform their business to be AI-native, like we have done at Amplitude. The AI landscape is changing rapidly, and they want to learn how to adapt. Our customers are on a spectrum of AI adoption. Our job is to meet them where they are and then educate them on how to take the next step. We work with leading AI companies to learn what the bleeding edge in product development looks like.

當我與客戶交流時，他們都聚焦於如何像 Amplitude 一樣，將其業務轉型為 AI 原生。AI 版圖變化迅速，他們希望學習如何因應調整。我們的客戶在 AI 採用程度上呈現光譜分布。我們的工作是從他們目前的位置出發，並教育他們如何邁向下一步。我們與領先的 AI 公司合作，了解產品開發最前沿的樣貌。

We use that knowledge to educate the rest of the market, including the largest enterprises deploying at scale. More than 40 AI-native companies now pay us over $100,000 a year. Those customers include Harvey, Midjourney, Character AI and one of the leading foundational AI model companies. On the enterprise side, enterprises are now more than 68% of our ARR. This quarter included agreements with Paramount, Jaguar Land Rover and Domino's Pizza.

我們運用這些知識來教育市場其他參與者，包括以規模化方式部署的最大型企業。目前已有超過 40 家 AI 原生公司每年支付我們超過 10 萬美元。這些客戶包括 Harvey、Midjourney、Character AI，以及一家領先的基礎 AI 模型公司。在企業端，企業客戶目前占我們 ARR 的 68% 以上。本季也包含與 Paramount、Jaguar Land Rover 以及 Domino's Pizza 的合作協議。

We've improved our pricing and packaging. We reduced down to a single meter to make it simpler for enterprises to add additional products. We increased the amount of data on our free plan, so we're the best for those just getting started. Amplitude has the best pricing, whether you're a start-up or a large enterprises. One of the biggest changes with building an AI-native company we're seeing at Amplitude and with our peers in private markets is in the cost structure.

我們已改善定價與方案包裝。我們將計量方式精簡為單一計量表（meter），讓企業更容易加購其他產品。我們提高了免費方案可用的資料量，因此對於剛起步的客戶而言，我們是最佳選擇。無論你是新創或大型企業，Amplitude 都提供最具競爭力的定價。在 Amplitude 以及我們在私募市場的同業身上，我們看到打造 AI 原生公司的最大變化之一在於成本結構。

A lot of inference spend is required in order to deliver AI-native products, which increases the amount spent on cost of goods sold. On the other hand, you do not need to add as much operating expense to continue to grow a business at scale. We are embracing this change in cost structure as part of our transition to an AI-native company. For now, we expect gross margins to stay in the low-70s. We will offset that with a commensurate reduction in operating expenses.

為了交付 AI 原生產品，需要投入大量推論（inference）支出，這會提高銷貨成本的支出。另一方面，若要在規模化下持續成長，你不需要等比例增加那麼多營運費用。我們正將這項成本結構的變化視為轉型為 AI 原生公司的一部分並加以擁抱。目前，我們預期毛利率將維持在 70% 出頭的水準。我們將以相應幅度降低營運費用來抵銷其影響。

That allows us to continue to show the same leverage in operating income as we have planned. I am continuing to drive Amplitude to a 20%-plus operating margin business over the long term. We offer three products to meet customers wherever they are on their AI journey. Amplitude gives you the deepest understanding of how people use your product. Our agents increasingly do that discovery for you.

這讓我們能夠如同規劃般，持續在營業利益上展現相同的槓桿效益。我將持續推動 Amplitude 長期成為一家營業利益率 20% 以上的企業。我們提供三項產品，能在客戶 AI 旅程的任何階段滿足其需求。Amplitude 讓你對人們如何使用你的產品有最深入的理解。我們的代理人也愈來愈多地替你完成這類探索。

Statsig gives you feature flagging and experimentation built on the world's most advanced SaaS engine with an engineering-first view. It's also integrated natively with data warehouses. Wave is the future of product development, self-improving products where we automatically recommend what to build next based on signals from users. While we're early here, I'm actually excited to show you a demo today. Together, these three products close the product development loop, understand what's happening, measure what ships and ship what matters.

Statsig 提供功能旗標與實驗能力，建立在全球最先進的 SaaS 引擎之上，並以工程優先的視角設計。它也可與資料倉儲原生整合。Wave 是產品開發的未來：自我改進的產品，會根據使用者訊號自動推薦下一步該打造什麼。雖然我們仍在早期階段，但我其實很期待今天向各位展示一個示範。這三項產品合在一起，能閉合產品開發迴圈：理解正在發生什麼、衡量已上線的內容，並交付真正重要的功能。

That loop is how AI-native business is built. Let me go deeper on Amplitude. Global chat is becoming the primary way our customers interact with their product data. You ask it a question in plain language and it does the analysis, no dashboard building required. It's become the de facto way many companies do product analytics.

這個迴圈就是打造 AI 原生企業的方式。讓我更深入談談 Amplitude。Global chat 正在成為我們客戶與其產品資料互動的主要方式。你用自然語言提問，它就會完成分析，不需要建立儀表板。它已成為許多公司進行產品分析的事實標準方式。

Global agent finds the root cause behind 75% of customer questions and hands you the answer. There are 1.3 million global agent interactions every week and root cause discovery rates are improving by 1 percentage point every month. As of today, over 40% of all insights come from AI agents as opposed to humans, and we expect this to continue to grow. Today, for our demo, I want to show you custom agents, Statsig and Wave. Let's start with custom agents.

Global agent 能找出 75% 客戶問題背後的根因，並把答案交給你。每週有 130 萬次 global agent 互動，且根因發現率每月提升 1 個百分點。截至今天，超過 40% 的所有洞察來自 AI 代理人而非人類，我們預期這個比例會持續成長。今天的示範，我想向各位展示自訂代理人、Statsig 與 Wave。我們先從自訂代理人開始。

Custom agents are teammates that automate recurring workflows on your product data and push that work to other tools and systems. This is our chat interface. An increasing number of users are interacting with Amplitude mostly through chat and agents. I'll ask a question. Which group of users are most likely to purchase next week?

自訂代理人是隊友，能自動化你在產品資料上的重複性工作流程，並把成果推送到其他工具與系統。這是我們的聊天介面。愈來愈多使用者主要透過聊天與代理人來使用 Amplitude。我來問一個問題。哪一群使用者最有可能在下週購買？

Chat can now write its own code to perform this analysis. This unlocks the ability to run deeper analysis and create powerful new graphs and artifacts, including diagrams like you see here, out-of-time decile lift, an ROC curve, segment propensity. You can dig in by seeing the actual code used and step-by-step analysis. This type of deep analysis has never been available before in analytics tooling. We are no longer bound by the constraints of a UI.

Chat 現在可以自行撰寫程式碼來執行這項分析。這解鎖了更深入的分析能力，並能建立強大的新圖表與產出物，包括你在這裡看到的圖示、跨期十分位提升（out-of-time decile lift）、ROC 曲線、區隔傾向（segment propensity）。你可以透過查看實際使用的程式碼與逐步分析來深入探究。這種深度分析在分析工具中前所未見。我們不再受限於 UI 的限制。

We can also create automatic and recurring agents that run in the background. I give it these instructions. I want this analysis run every Monday morning, cross reference with marketing activity and Confluence, DM me the results in Slack. Amplitude then creates the agent that you see here. This is the entire prompt, including connectors to Atlassian and Slack.

我們也能建立自動且定期執行的代理人，在背景運作。我給它這些指示。我希望每週一早上執行這個分析，並與行銷活動及 Confluence 交叉比對，然後在 Slack 私訊我結果。Amplitude 接著會建立你在這裡看到的代理人。這就是完整的提示詞（prompt），包含連接 Atlassian 與 Slack 的連接器。

It will run regularly every Monday and push the results to me. We are building the best analytics agent across all data sources. Statsig is the leading product for experimentation and feature management. Statsig runs experiments natively on your cloud data warehouse, whether that is Snowflake, BigQuery, Databricks or Redshift. Let me show you what this looks like.

它會在每週一定期執行，並把結果推送給我。我們正在打造跨所有資料來源的最佳分析代理人。Statsig 是實驗與功能管理的領先產品。Statsig 可在你的雲端資料倉儲上原生執行實驗，不論是 Snowflake、BigQuery、Databricks 或 Redshift。讓我示範它看起來是什麼樣子。

Here is the results page for one of hundreds of experiments that an e-commerce customer is running. This experiment is testing a larger product image versus the default size. There's a lot of statistical machinery behind a good experiment, but the UI makes it simple for an engineer to run. Up top, they can monitor exposure, which is saying if the experiment is healthy or not. We expect to see a 50:50 split, so we're doing good.

這是一位電商客戶正在執行的數百個實驗之一的結果頁面。這個實驗在測試較大的產品圖片與預設尺寸的差異。一個好的實驗背後有大量統計機制，但 UI 讓工程師能輕鬆執行。在上方，他們可以監控曝光（exposure），也就是判斷實驗是否健康。我們預期看到 50:50 的分流，所以目前表現良好。

And as you can see over here, we're getting a healthy check. We move to the scorecard that has the results. This has a confidence interval of 95%. Statsig uses advanced techniques like CUPED and sequential testing that allows engineers to speed up time to decision. We have those turned on.

如你在這裡看到的，我們得到健康檢查通過。接著我們移到顯示結果的計分卡（scorecard）。這裡的信賴區間為 95%。Statsig 使用 CUPED、序列檢定等進階技術，讓工程師能加快做出決策的時間。我們已將這些功能開啟。

In monitoring, we see specific events we're tracking for this experiment. We're seeing positive results. The checkout event is up by 27.4%, plus or minus 2.3%. Cart conversion is up. Total purchase dollars is up, while cards per session is down.

在監控中，我們看到此實驗正在追蹤的特定事件。我們看到正向結果。結帳事件提升 27.4%，正負 2.3%。購物車轉換率上升。總購買金額上升，而每次工作階段的卡片數（cards per session）下降。

For the rollout of this feature, we have a progressive rollout, starting with employees moving to early access users, then early release and a scheduled rollout for everyone else. Statsig has a variety of advanced experimentation capabilities for rollout like feature gating, dynamic configs and automatic rollbacks. Together, these are the mechanisms that a team uses to ship a change gradually, tune it while live and pull back automatically it goes wrong. Last, I want to show you Wave, the future of product development. Wave allows for self-improving products that automatically recommend what to build next based on signals from your users.

在此功能的上線推廣上，我們採用漸進式推廣：先從員工開始，再到搶先體驗使用者，接著早期釋出，最後排程推廣給其他所有人。Statsig 具備多種用於推廣的進階實驗能力，例如功能閘控（feature gating）、動態設定（dynamic configs）與自動回滾（automatic rollbacks）。這些機制讓團隊能逐步交付變更、在上線期間調整，並在出問題時自動撤回。最後，我想向各位展示 Wave——產品開發的未來。Wave 讓產品能自我改進，根據使用者訊號自動推薦下一步該打造什麼。

Wave is magical. Wave looks across all the different data sources you have; analytics, experimentation, session replay, Guides and Surveys, feedback and many others. It then synthesizes that data into a set of product recommendations, plans those recommendations and then helps you create those changes in your product. I'm going to walk you through a real example Wave suggested and built for Amplitude's documentation site. On our documentation site, Wave found a spike in failed searches through looking at Session Replay and analytics data.

Wave 很神奇。Wave 會檢視你所有不同的資料來源：分析、實驗、Session Replay、Guides 與 Surveys、回饋以及其他許多來源。接著它會將這些資料綜合成一組產品建議，規劃這些建議，並協助你在產品中建立這些變更。我將帶你看一個 Wave 為 Amplitude 文件網站所建議並建置的真實案例。在我們的文件網站上，Wave 透過檢視 Session Replay 與分析資料，發現失敗搜尋出現尖峰。

The core problem was that search on our docs page fired on every key stroke, typing a single letter to start a search returned an empty no result state before the person finished typing their search leading to a bad experience for users. Wave explains the reach of this issue. Every user who uses search, it has an expected impact of decreasing total search failures by 80%. Then Wave has automatically created a visual example of the problem below, so it's easy to understand. It also has a full explanation of the evidence.

核心問題在於：我們文件頁面的搜尋會在每次按鍵時觸發；當使用者只輸入一個字母開始搜尋時，在他完成輸入前就會回傳空的「無結果」狀態，導致使用者體驗不佳。Wave 說明了此問題的影響範圍。對每位使用搜尋的使用者而言，預期可將總搜尋失敗降低 80%。接著 Wave 會在下方自動建立問題的視覺化示例，讓人容易理解。它也提供完整的證據說明。

For the plan, Wave sketches a wire frame of the recommended update, setting a three-character minimum and a 200-millisecond debounce to trigger the search. Wave can also drive execution. It automatically created the poll request and cursor wrote the code. Mark, our Technical Writer, was able to merge this poll request and ship this. No engineers, no designers and no product manager.

在規劃方面，Wave 繪製了建議更新的線框圖：設定至少 3 個字元才觸發搜尋，並加入 200 毫秒的去彈跳（debounce）來觸發搜尋。Wave 也能推動執行。它自動建立了拉取請求（pull request），並由 Cursor 撰寫程式碼。我們的技術寫作者 Mark 能夠合併這個拉取請求並完成上線。不需要工程師、不需要設計師，也不需要產品經理。

Finally, Wave measures the results of the change. There is a massive decrease in total search failures. Simply amazing, simply amazing. Now, let's talk about some of our customers. We had a great quarter with both new lands and expansions.

最後，Wave 會衡量變更的結果。整體搜尋失敗次數大幅下降。簡直令人驚嘆，真的令人驚嘆。現在，讓我們談談一些客戶案例。本季我們在新客戶導入與既有客戶擴張方面都表現亮眼。

We added or expanded our relationship with customers, including Paramount Global, Jaguar Land Rover, Teladoc Health, Chime, Disney ad platforms, F5 Networks, Coursera, Grammarly, Kraken and Crunch Fitness, among others. I want to tell you three stories about how these customers are leveraging our platform. First is Coca-Cola FEMSA, which sells to hundreds of thousands of small shops across Latin America. Every shop is different, but for years, they have to run the same broad campaign to everyone because there is no way to tailor a message to that many retailers by hand. AI changed that.

我們新增或擴大了與多家客戶的合作關係，包括 Paramount Global、Jaguar Land Rover、Teladoc Health、Chime、Disney 廣告平台、F5 Networks、Coursera、Grammarly、Kraken 與 Crunch Fitness 等。我想分享三個故事，說明這些客戶如何運用我們的平台。第一個是 Coca-Cola FEMSA，他們在拉丁美洲向數十萬家小店銷售。每家店都不同，但多年來他們不得不對所有人投放同一個廣泛的行銷活動，因為不可能靠人工為如此多零售商逐一客製訊息。AI 改變了這一切。

They began sending each retailer its own recommendation every week written by AI. Their own teams were actually skeptical. A different message for every shop every week felt risky and no one knew if it was going to work. They used Amplitude to find out. Their AI campaigns actually had an 11% click-through rate, 4x higher than their previous approach.

他們開始每週向每位零售商發送一則由 AI 撰寫、專屬於該零售商的推薦內容。他們自己的團隊其實一開始持懷疑態度。每家店每週都收到不同訊息感覺風險很高，而且沒有人知道是否會奏效。他們使用 Amplitude 來驗證。他們的 AI 行銷活動點擊率實際達到 11%，是先前方法的 4 倍。

Our cohort analysis also showed that this lift lasted. Once a retailer engaged, its revenue stayed higher in the weeks that followed. That evidence turned skeptics at FEMSA into believers, and they scaled from a 2,500 store pilot to 690,000 retailers. The second is Replit. Replit is an AI app builder that allows non-technical builders to turn an idea into an app using AI.

我們的同群分析也顯示，這樣的提升具有持續性。一旦零售商開始互動，接下來幾週的營收都維持在較高水準。這些證據讓 FEMSA 的懷疑者轉為相信者，並將規模從 2,500 家門市的試點擴大到 690,000 家零售商。第二個是 Replit。Replit 是一個 AI 應用程式建置平台，讓非技術背景的創作者也能用 AI 把想法變成 App。

Replit has a large global user base of passionate builders that provide feedback. Replit is using Amplitude AI Feedback to understand how customers are engaging with their agents. They've connected AI Feedback to Zendesk, App Store Reviews, Twitter and Reddit and surfaced and prioritize what problems should be solved to increase their retention and engagement. It changed weeks of manual work on their end into a simple click with Amplitude. This is the next generation of product development at work.

Replit 擁有龐大的全球使用者基礎，且是一群熱情的創作者，會提供回饋。Replit 正在使用 Amplitude AI Feedback 來了解客戶如何與他們的代理（agents）互動。他們把 AI Feedback 連接到 Zendesk、App Store 評論、Twitter 與 Reddit，並將應優先解決、可提升留存與互動的問題浮現並排序。這把他們原本需要數週的人工工作，透過 Amplitude 變成一次簡單點擊。這就是下一代產品開發的實際運作方式。

Third is The Economist. The Economist is a print magazine that's in the midst of a transition to digital delivery and subscription. Their research arm built an AI assistant called Lens that answers questions for analysts and strategists using The Economist's content. Their normal analytics could show what users did, but not whether the AI's answers were any good. The team was reading sessions by hand, but they couldn't keep up.

第三個是《經濟學人》（The Economist）。《經濟學人》是一份紙本雜誌，正處於轉型為數位發行與訂閱的過程中。他們的研究部門打造了一個名為 Lens 的 AI 助理，使用《經濟學人》的內容為分析師與策略師回答問題。他們原本的分析工具能顯示使用者做了什麼，但無法判斷 AI 的回答品質好不好。團隊曾以人工逐一閱讀工作階段，但根本跟不上。

Amplitude Agent Analytics now scores every answer Lens gives automatically. They went from reading a handful of sample sessions to being able to see across all of them. Today, Lens holds a 96.9% task success rate and weekly failures are down 84%. That is the loop working, build with AI, measure whether it is good and fix what is not. To wrap up, the companies on the bleeding edge are choosing Amplitude.

Amplitude Agent Analytics 現在會自動為 Lens 的每個回答打分。他們從只能閱讀少量樣本工作階段，提升到能夠檢視全部工作階段的整體表現。目前，Lens 的任務成功率達 96.9%，每週失敗次數下降 84%。這就是迴圈在運作：用 AI 建置、衡量是否夠好，並修正不好的部分。總結來說，走在最前沿的公司正在選擇 Amplitude。

We've transformed Amplitude to be AI-native, and we're building the future on what can be done in analytics. Self-improving products are closer than ever with Wave. Our pace of innovation continues to accelerate, and we're building in a way that can scale with leverage. I'm extraordinarily excited about what's ahead.

我們已將 Amplitude 轉型為 AI 原生（AI-native），並在分析能力可達成的未來上持續建構。有了 Wave，自我改進的產品比以往任何時候都更接近實現。我們的創新速度持續加快，且我們以可透過槓桿擴張的方式在打造產品。我對未來的發展感到無比興奮。

With that, I'll hand it over to Andrew to walk you through the financials.

接下來我把時間交給 Andrew，請他帶大家看財務表現。

Thank you, Spenser. This was a strong quarter and a clear step forward in our execution, bringing our vision of how products will increasingly be developed and improved. We crossed $100 million in quarterly revenue. ARR reached $410 million, growing over 22% with the addition of the ARR assumed from the Statsig business and free cash flow was a record quarterly high of $23.7 million. We also returned $69 million in capital during the quarter as part of our share repurchase program.

謝謝你，Spenser。這是一個強勁的季度，也是我們執行力明顯向前邁進的一步，讓我們對產品將如何日益被開發與改進的願景更進一步落地。我們的季度營收突破 1 億美元。ARR 達到 4.10 億美元，在納入自 Statsig 業務假設承接的 ARR 後，成長超過 22%；自由現金流創下單季新高，達 2,370 萬美元。我們也在本季透過股票回購計畫回饋股東 6,900 萬美元資本。

We accomplished these milestones while integrating the Statsig technology and customers, managing through our own AI-native evolution and implementing our new pricing and packaging strategy. AI is changing how customers use Amplitude. The more our customers build with AI, the more they need to measure. Customers that adopt our AI into their workflows run nearly 10 times the number of analyses compared to those that are running things manually. This increases the value that customers receive from the data ingested into our platform and makes it more likely that they'll both ingest larger amounts of data and expand into additional products, which is the basis of our growth.

我們在整合 Statsig 的技術與客戶、推進自身 AI 原生轉型，以及導入新的定價與方案策略的同時，仍達成了這些里程碑。AI 正在改變客戶使用 Amplitude 的方式。客戶用 AI 建置得越多，就越需要衡量。將我們的 AI 納入工作流程的客戶，其分析次數幾乎是手動操作客戶的 10 倍。這提升了客戶從匯入我們平台的資料中獲得的價值，也使他們更可能匯入更大量的資料並擴展到更多產品，這正是我們成長的基礎。

Our new pricing and packaging is working. It supports our market consolidation strategy by providing customers with a lower overall cost if they consolidate applications onto our platform. It provides customers greater cost predictability and simplifies the quoting process for our sellers. In the second quarter, 70% of the ARR we closed was on the new model, up from 25% in the first quarter. Now, 28% of our total ARR is on the new pricing and packaging.

我們新的定價與方案組合正在發揮效果。它透過讓客戶把多個應用整合到我們平台上時享有更低的整體成本，來支持我們的市場整合策略。它也為客戶帶來更高的成本可預測性，並簡化我們銷售團隊的報價流程。在第二季，我們成交的 ARR 中有 70% 採用新模式，高於第一季的 25%。目前，我們總 ARR 的 28% 已採用新的定價與方案組合。

This is leading to average ARR increasing, higher multi-product attach and longer contract duration, which all contribute to greater durability of our revenue. Our margins reflect the choice. These are investments we are making to drive future growth with increasing profitability. Our gross margin was down over 1 point versus Q1 due to the integration of Statsig. We are working to optimize the new hosting environment and cloud structure, but it will take some time to improve from the low-50s gross margin closer to our expectation of 70-plus for the Statsig business.

這正帶動平均 ARR 提升、更高的多產品附加率，以及更長的合約期間，這些都讓我們的營收更具持久性。我們的毛利率也反映了這些選擇。這些是我們為了在提升獲利能力的同時推動未來成長所做的投資。由於整合 Statsig，我們的毛利率較第一季下降超過 1 個百分點。我們正在優化新的託管環境與雲端架構，但要把 Statsig 業務的毛利率從 50% 出頭提升到我們預期的 70% 以上，仍需要一些時間。

We're also experiencing higher customer adoption of AI capabilities and greater data ingestion into our platform, which combined has increased our costs and reduced our gross margins by an additional 2 points versus Q1. We have long maintained that we will grow with leverage. This investment in the cost of revenue places greater emphasis on the management of our operating expenses to a lower level in order to achieve the leverage. In Q2, we've managed down our sales and marketing to below 40% of revenue and G&A to the low teens, which is contributing to an increase in operating margins. We will continue to manage both areas lower as a percentage of revenue over time, and we will continue to invest in R&D to drive innovation.

我們也看到客戶對 AI 功能的採用提高、以及匯入平台的資料量增加，兩者合計使我們的成本上升，並使毛利率相較第一季再下降 2 個百分點。我們長期以來一直強調，我們將以槓桿方式成長。這項在營收成本上的投資，使我們更需要把營運費用管理到更低水準，以實現槓桿效益。在第二季，我們已將銷售與行銷費用降至營收的 40% 以下，並將一般與行政費用（G&A）控制在 10% 出頭，這正推動營業利益率提升。我們將持續把這兩項費用占營收比重隨時間進一步降低，同時也會持續投資研發以推動創新。

We are instrumenting our business to accelerate growth, capture market share and show leverage. One key metric we monitor is the usage of data compared to the entitlement for our customers as this is a primary monetization metric. Today, that metric is at an all-time high. This is the output from better pricing packaging and more usage driven by our AI features. We have increased the durability of our business through our RPO growth and reinvented our internal processes to capture scalability that AI offers.

我們正在為業務建立更完善的量測機制，以加速成長、取得市占並展現槓桿效益。我們監控的一項關鍵指標，是客戶的資料使用量相對於其權益（entitlement）的比例，因為這是主要的變現指標。目前，該指標已達到歷史新高。這是更好的定價與方案組合，以及由 AI 功能帶動更高使用量所產生的成果。我們也透過 RPO 成長提升了業務的持久性，並重塑內部流程，以掌握 AI 所帶來的可擴展性。

We are running the AI opportunity and taking share as we go. Turning to our second quarter results. As a reminder, all financial results that I'll be discussing, with the exception of revenue, are non-GAAP. Our GAAP financial results, along with a reconciliation between GAAP and non-GAAP results can be found in our earnings press release and supplemental financials on the Investor Relations page of our website. Second quarter revenue was $100.9 million, up 21% year-over-year and 8% quarter-over-quarter. Total ARR increased to $410 million, exiting the second quarter, an increase of 22% year-over-year and $36 million sequentially.

我們正在把握 AI 的機會，並在推進過程中持續擴大市占。接著談我們第二季的業績。提醒一下，除了營收之外，我將討論的所有財務結果皆為非 GAAP。我們的 GAAP 財務結果，以及 GAAP 與非 GAAP 結果之間的調節表，可在我們的財報新聞稿與網站投資人關係頁面的補充財務資料中查閱。第二季營收為 1.009 億美元，年增 21%，季增 8%。第二季結束時，總 ARR 增至 4.10 億美元，年增 22%，較前一季增加 3,600 萬美元。

This includes $17 million of incremental ARR from the Statsig business compared to the $16 million we expected to add when we shared our first quarter earnings. Total remaining performance obligations grew 35% year-over-year to $483 million. Current RPO was up 30% year-over-year and long-term RPO was up 47% year-over-year. Here are more details on the key elements of the quarter. We had a strong quarter for both new and expansion deals in the enterprise and platform sales were again particularly strong.

其中包含來自 Statsig 業務新增的 1,700 萬美元 ARR，較我們在分享第一季財報時預期新增的 1,600 萬美元更高。總剩餘履約義務（RPO）年增 35% 至 4.83 億美元。當期 RPO 年增 30%，長期 RPO 年增 47%。以下是本季關鍵要素的更多細節。本季在企業端的新單與擴張交易皆表現強勁，而平台銷售再次特別強勁。

48% of our customers now have multiple products, with 80% of our ARR coming from that cohort. We have over 26% of our ARR from customers with 5 or more products, up 2 times since the second quarter last year. In period, net dollar retention was 105% on a pro forma basis, led by cross-sell expansions across our customer base. This pro forma basis includes Statsig and Amplitude customers. Gross margin was 71% for the second quarter, down approximately 4 points from the second quarter of last year and down 4 points sequentially.

目前有 48% 的客戶使用多項產品，而我們 80% 的 ARR 來自這一客群。我們有超過 26% 的 ARR 來自使用 5 項或以上產品的客戶，較去年第二季成長至 2 倍。本期按備考（pro forma）基礎計算的淨美元留存率為 105%，主要由全客戶群的交叉銷售擴張所帶動。此備考基礎包含 Statsig 與 Amplitude 客戶。第二季毛利率為 71%，較去年第二季下降約 4 個百分點，較前一季亦下降 4 個百分點。

This was driven by continued growth in inference costs as customer adoption of our AI tools accelerated, along with the integration of the Statsig business in its hosting environment. Sales and marketing expenses were 39% of revenue, down from 44% in the second quarter of last year. G&A was 13% of revenue, down 1 point from the second quarter of last year. R&D was 21% of revenue, up approximately 3 points from the second quarter last year, reflecting investment to scale the Statsig opportunity and support for those customers. Total operating expenses were $73 million, or 72% of revenue.

這主要是因客戶對我們 AI 工具的採用加速，推動推理成本持續成長，以及 Statsig 業務在其託管環境中的整合所致。銷售與行銷費用占營收 39%，低於去年第二季的 44%。一般與行政（G&A）費用占營收 13%，較去年第二季下降 1 個百分點。研發（R&D）費用占營收 21%，較去年第二季上升約 3 個百分點，反映我們為擴大 Statsig 機會並支援其客戶所做的投資。總營業費用為 7,300 萬美元，占營收 72%。

Operating loss was $1.5 million or 1.4% of revenue. Net loss per share was negative $0.01 based on 129.4 million basic shares compared to $0.01 a year ago. Free cash flow in the quarter was $23.7 million, or 24% of revenue compared to $18.2 million, or 22% of revenue during the same period last year. We ended the quarter with approximately $162 million in cash and investments. We have conviction in the long-term value of our platform and have used and will use our cash to minimize the impacts of dilution.

營業虧損為 150 萬美元，占營收 1.4%。每股淨虧損為 -0.01 美元，係以 1.294 億股基本股數計算；去年同期為每股 0.01 美元。本季自由現金流為 2,370 萬美元，占營收 24%；去年同期為 1,820 萬美元，占營收 22%。本季結束時，我們持有約 1.62 億美元的現金與投資。我們對平台的長期價值深具信心，並已且將持續運用現金以將稀釋影響降至最低。

Our balance sheet position remains strong and allows us the opportunity to be more aggressive in our M&A strategy to accelerate our R&D road map when appropriate. Now turning to our outlook. As a reminder, the philosophy of how we set guidance is through the lens of execution. We are pleased with our overall progress on consolidating point solutions to our core platform and the adoption of our different AI technologies. We've instrumented our business and selling process to make it easier to use more of our platform.

我們的資產負債表仍然穩健，使我們在適當時機能更積極採取併購（M&A）策略，以加速我們的研發路線圖。接著談我們的展望。提醒一下，我們設定財測的理念是以執行為核心視角。我們對將點狀解決方案整合至核心平台的整體進展，以及不同 AI 技術的採用情況感到滿意。我們已將業務與銷售流程工具化，使客戶更容易使用我們平台的更多功能。

We believe that we are well positioned to continue to accelerate our growth in a profitable way. For the third quarter of 2026, we expect revenue to be between $105.6 million and $108 million, representing an annual growth rate of 21% at the midpoint. We expect non-GAAP operating income to be between $2.5 million and $4.5 million. And we expect non-GAAP net income per share to be between $0.02 and $0.03, assuming a weighted average shares outstanding of approximately 133 million as measured on a fully diluted basis. For the full-year 2026, we are raising our expectation for full-year revenue based on the performances in the second quarter to be between $407.2 million and $411.2 million, an annual growth rate of 19% at the midpoint.

我們相信，我們已具備良好定位，能以獲利的方式持續加速成長。針對 2026 年第三季，我們預期營收介於 1.056 億至 1.08 億美元之間，以中位數計算年成長率為 21%。我們預期非 GAAP 營業利益介於 250 萬至 450 萬美元之間。我們預期非 GAAP 每股淨利介於 0.02 至 0.03 美元之間，假設以完全稀釋基礎衡量的加權平均流通股數約為 1.33 億股。針對 2026 全年，我們基於第二季表現上調全年營收預期至 4.072 億至 4.112 億美元之間，以中位數計算年成長率為 19%。

We are also raising our expectation for the full year non-GAAP operating income due to performances in the second quarter and actions taken in the first half to be between $6.3 million and $9.3 million. We expect non-GAAP net income per share to be between $0.06 and $0.08, assuming weighted average shares outstanding of approximately 137.1 million as measured on a fully diluted basis. In closing, we are accelerating our pace of innovation, and we're growing the value that we can deliver to our customers. We have confidence in our ability to scale a durable and growing business while also bringing agentic analytics to the world.

我們也因第二季表現以及上半年採取的行動，上調全年非 GAAP 營業利益預期至 630 萬至 930 萬美元之間。我們預期非 GAAP 每股淨利介於 0.06 至 0.08 美元之間，假設以完全稀釋基礎衡量的加權平均流通股數約為 1.371 億股。最後，我們正在加快創新步伐，並提升我們能為客戶交付的價值。我們有信心在擴大一個具韌性且持續成長的業務規模的同時，也把代理式分析（agentic analytics）帶給全世界。

With that, we'll open up for Q&A. Over to you, John.

接下來，我們開放 Q&A。交給你了，John。

Thank you, Andrew. Going to Q&A. (Operator Instructions) Our first question today will come from the line of Mark Cash from Raymond James, followed by Jackson Ader at KeyBanc.

謝謝你，Andrew。我們進入 Q&A。(接線員指示) 今天第一個問題來自 Raymond James 的 Mark Cash，接著是 KeyBanc 的 Jackson Ader。

Thanks, John. Yeah. If I could start with Spenser, I really wanted to ask around Wave. I appreciate it's still limited beta, but I think you've been using internally for several months now. I guess, do you see Wave, if it could cause maybe a shift -- a company shifting away from using bespoke agents for specific use cases towards a broader AI-native product development platform from what you're seeing? And if so, how could that change your buyer, maybe the budgets you see in the addressable market over time?

謝謝，John。是的。如果我可以先從 Spenser 開始，我很想問一下關於 Wave。我知道目前仍是有限度的 beta，但我想你們內部已經使用了好幾個月。我想問的是，從你們看到的情況來看，Wave 是否可能帶來一個轉變——公司從針對特定使用情境採用客製化代理（bespoke agents），轉向更廣泛的 AI 原生產品開發平台？如果是，這會如何改變你們的買方輪廓，或是你們在可服務市場中長期看到的預算配置？

When you say bespoke, like, say, more on that, like-

你說的客製化是指什麼？可以再多說一點嗎——像是……

Yeah, Instead of using particular agents to do a specific task underlying because you have like a swarm of agents doing things underneath for Wave. So-

對。與其使用特定代理去做某個特定任務，因為在 Wave 底層其實有一群代理在做事情。所以——

I see what you're saying. I see. Okay. So yes, let me separate out a few different things. What we have on the Amplitude side, and I showed with custom agents is you have these agents that can look across the data and find insights for you and get to the root cause of questions and do that on a regular basis and kind of send it out.

我懂你的意思了。我明白。好。所以是的，讓我把幾件不同的事情拆開來說。在 Amplitude 這邊，我展示的自訂代理，是這些代理可以跨資料去找出洞察，幫你追溯問題的根因，並且能定期執行，然後把結果發送出去。

What Wave is doing, in particular, to your point, is it's kind of -- it's looking at all your data all the time and then saying, hey, here are points of friction. Here's something that's not working out should be. Here's a feature that I think you should emphasize more. Here's something that I think is a best practice that you're not doing. And so it's operating at a kind of higher level.

而 Wave 在做的事情，特別是如你所說，它有點像——它會一直看著你所有的資料，然後說：嘿，這裡有摩擦點。這裡有某件事沒有如預期運作。這裡有一個功能我認為你應該更強調。這裡有一個我認為是最佳實務、但你還沒做到的事情。所以它是在一個更高層級上運作。

In terms of the persona, I think what we're seeing is a convergence between engineers, product managers and designers into this AI builder persona. It's not really like you have engineers who are thinking about what to build and you have product managers who are also just chipping code. And so the best -- if you look at where the AI-native teams that everyone is aspiring to be, these roles are melding. So it's still the same problem we're solving, which is how do we help you build a better product, but we're just automating more of it because we're saying, hey, we're going to look at all the data all the time and then suggest recommendations. That -- like I've been -- we've been talking about self-improving products here at Amplitude for about nine years.

就角色樣貌而言，我認為我們看到的是工程師、產品經理與設計師正匯聚成一種「AI 建造者」的人設。不再像是工程師在思考要做什麼，而產品經理也只是順手寫點程式碼那樣。因此，最優秀的——如果你看大家都在追求成為的 AI 原生團隊——這些角色正在融合。所以我們解決的仍是同一個問題：如何幫你打造更好的產品；只是我們把更多流程自動化，因為我們在說：嘿，我們會一直看所有資料，然後提出建議。這——我一直——我們在 Amplitude 談「自我改進的產品」大概已經九年了。

And so I'm actually been blown away by what is possible with the technology today where it's just -- it's the perfect problem for AI in a lot of ways. The data sets are massive and complex. So, you can't get any human to look at them. And then the synthesis of, okay, here's what I think could be better and best practices is actually like extraordinarily impressive. And so what that means is that just by the fact that someone is using your software, like it's getting better because it's just translating recommendations.

所以我其實對今天技術所能做到的事情感到非常震撼——在很多方面，這正是 AI 的完美問題場景。資料集龐大且複雜。因此，你不可能讓任何人類去逐一檢視。而把它們綜合起來，像是「好，這裡我認為可以更好」以及最佳實務的提煉，實際上非常、非常令人印象深刻。這代表的是：只要有人在使用你的軟體，它就會變得更好，因為它會把建議轉化並輸出。

You no longer need someone to go into an Amplitude or to any data system and say, oh, here's what my interpretation of these results. So, I do think -- in terms of budget and persona, I do think, again, that means instead of having these distinct roles, you have engineering, product management and design merge. You're still doing digital product development, and that still rolls up to some leader, the same executive before. But yes, the way you do it looks different. Did I hit it on what you are looking for?

你不再需要有人進到 Amplitude 或任何資料系統裡說：喔，這是我對這些結果的解讀。所以我確實認為——就預算與角色而言——這再次意味著：與其有這些彼此分離的職能，你會看到工程、產品管理與設計合併。你仍然在做數位產品開發，而這仍然會向上匯報給某位領導者，也就是跟以前相同的高階主管。但沒錯，你做事的方式看起來會不同。我有回答到你想問的重點嗎？

Yeah, Absolutely. And if I could follow up with Andrew real quick. If my math is correct, the guidance for the year was raised by more than 2 times the beat for revenue and operating income. So, I was wondering if you could just go to the key drivers of lifting growth expectations Why you saw some pressure on pro forma expansion sequentially there in the quarter?

對，完全有。如果我可以很快追問 Andrew 一下。如果我算得沒錯，全年指引的上調幅度超過本季營收與營業利益超預期（beat）的 2 倍。所以我想請你談談上調成長預期的關鍵驅動因素，以及為什麼本季 pro forma 擴張（expansion）在季比上看到了些壓力？

And then what you're considered regarding margin leverage and levers while you're facing COGS pressure and ramping token spend internally? Thank you.

另外，在面臨 COGS 壓力、且內部 token 支出正在爬坡的同時，你們對於毛利槓桿與可用槓桿手段的考量是什麼？謝謝。

Yeah, sure. So, a couple of things. One, that when we look at our ability to actually generate revenue in the out quarters, one, we start with the strong balances we're booking that are showing up in our RPO. When you've got commitments from customers for a longer-term duration, you start to have better and better predictability about your future revenue.

好的，當然。有幾點。第一，當我們看我們在未來幾季實際創造營收的能力時，首先是從我們正在簽下的強勁合約餘額（balances）開始看，這些會反映在我們的 RPO 裡。當客戶做出更長期間的承諾時，你對未來營收的可預測性就會越來越高。

So, that's the first thing. It's one of the reasons why we emphasize that metric so much. The second thing is we look at how much our customers are actually responding to some of the initiatives we're putting out, and that comes in the form of our new product capabilities, our new pricing and packaging, areas where our sales team is running new promotions and activities, all those are bolstering our ability to see a stronger and stronger pipeline, and that pipeline progresses faster through its stages, which gives us greater and greater confidence that we'll add more and more in net new ARR.

這是第一點。這也是我們為什麼如此強調這個指標的原因之一。第二點是，我們會看客戶對我們推出的一些措施實際反應如何，這體現在我們新的產品能力、我們新的定價與包裝（pricing and packaging）、以及我們銷售團隊在推行的新促銷與活動等；這些都在強化我們看到更強管線（pipeline）的能力，而且該管線在各階段推進得更快，讓我們更有信心能新增更多的淨新增 ARR。

Now from a revenue perspective, as you know, the predominance of our business is all coming from our subscription revenue. So, those key factors on understanding what's the baseline, what can you see in your pipeline, what you expect to convert is what I refer to is our ability to go execute against the plans that are in front of us.

就營收結構而言，如你所知，我們業務的主要來源都是訂閱營收。因此，理解基準盤（baseline）是什麼、你在管線中能看到什麼、以及你預期能轉換多少，這些關鍵因素就是我所說的：我們執行眼前計畫的能力。

And sales teams have been doing a really good job of driving consolidation in the market, and that alone with our products is driving great conversions. So, that's the first thing. On some of the margin areas, I would tell you, look, -- we just -- in the case of the Google environment that we got or the Statsig, we're going to be focused on driving optimizations in that environment over a period of time. It's definitely lower. We said in the low-50s from a gross margin perspective.

而銷售團隊在推動市場整併（consolidation）方面做得非常好，光是這一點再加上我們的產品，就帶來很好的轉換。所以這是第一部分。至於一些毛利相關的面向，我會告訴你：你看——就我們取得的 Google 環境或 Statsig 而言，我們會在一段時間內專注於在該環境中推動最佳化。毛利率確實比較低。我們也說過，從毛利率角度來看是在 50% 出頭（low-50s）。

That comes from us taking on a whole new environment. Most of Amplitude, all of it, in fact, is on AWS. So, we took on a whole new cloud and hosting environment, and you have to go through the paces of really optimizing how you run those environments for customers. Our first objective was integrating, making sure there were no disruption in service. Now, we're moving quickly into how we can optimize those environments.

這是因為我們接手了一個全新的環境。Amplitude 的大部分——事實上全部——都在 AWS 上。因此我們接手了一個全新的雲端與託管環境，而你必須走完一整套流程，真正把你為客戶運行這些環境的方式最佳化。我們的第一個目標是完成整合，確保服務不中斷。現在，我們正快速轉向思考如何最佳化這些環境。

So, that's one big lever on the gross margin side. And we're constantly looking at how we can make investments to go drive greater efficiencies across all of our operating expense areas.

所以，這是毛利端的一個重要槓桿。同時，我們也持續在看，如何在所有營業費用領域進行投資，以推動更高的效率。

Great. Our next question will come from the line of Jackson Ader from KeyBanc, followed by Scott Berg.

很好。下一題將由 KeyBanc 的 Jackson Ader 提問，接著是 Scott Berg。

Hey, thanks, guys. Good to see you. I was curious on -- I guess, Andrew, kind of sticking with you and talking about rather than on the cost side, just on the operating expense side. We've seen really nice acceleration in organic ARR from the business.

嗨，謝謝各位。很高興見到你們。我想問——我想，Andrew，延續你剛剛的話題，與其談成本端，我想聚焦在營業費用端。我們看到公司有機 ARR 出現非常不錯的加速。

But if I take kind of a longer-term view, even on a non-GAAP basis, we're still around breakeven, right? And so I'm curious, as you're thinking about like driving more leverage and more incremental margin that you've talked about before on the income statement, what kind of impact should we expect that to have on the organic growth number, if at all?

但如果我拉長時間來看，即使在 non-GAAP 基礎上，我們仍大概在損益兩平附近，對吧？所以我想問的是，當你在思考推動更多槓桿、以及你之前在損益表上談到的更多增量利潤率（incremental margin）時，我們應該預期這會對有機成長數字造成什麼影響（如果有的話）？

Well, I'd tell you that, one, we still expect from an organic perspective, we've got a great set of products. Spenser just walked through a number of them that are brand new to the market. We think they have an enormous total addressable market that we can go after. So, revenue growth will be the predominance where we'll see increasing operating income. As far as leverage as a percentage of what that would be a percentage of operating income, I do expect over time that we'll be able to drive better and better gross -- cost to start revenue and increase gross margins over time.

我會告訴你，第一，就有機成長而言，我們仍然預期——我們有一套很棒的產品組合。Spenser 剛剛也走過其中幾項，這些都是全新推向市場的產品能力。我們認為它們對應的可服務總市場（TAM）非常龐大，我們可以去攻佔。因此，營收成長將會是我們看到營業利益提升的主要來源。至於槓桿、也就是營業利益占比會到什麼程度，我確實預期隨著時間推進，我們能把成本相對於營收（cost to start revenue）做得越來越好，並逐步提高毛利率。

It just takes time to go do those things, especially when you're seeing such a demand inflection from customers and increasing data lines. As I mentioned, we're at an all-time high for the amount of data ingested in the platform versus entitlements.

只是要做到這些需要時間，特別是在你看到客戶需求出現明顯拐點、以及資料量持續增加的情況下。如我提到的，我們平台相對於權益（entitlements）的資料攝取量已達到歷史新高。

When I first joined, that was in the low-60s. We're well into the 80s now as far as percentage of what customers have ingested versus what their entitlements are, and that portends increasing expansions on upsell, which is usually where we've had a lot of problems in the past of overselling and having to right-size contracts. For the first time, we're past those things and we're starting to see really good upsell, not just cross-sell driving growth.

我剛加入時，那個比例大概在 60% 出頭。現在就客戶實際攝取量相對於其權益的百分比而言，我們已經遠遠進入 80% 區間；這預示著透過加購（upsell）帶來的擴張會增加，而過去我們常遇到的問題是賣得過頭、之後不得不把合約調整到合適規模。第一次，我們已經走過那些狀況，開始看到非常不錯的加購表現，而不只是靠交叉銷售（cross-sell）來帶動成長。

So, revenue growth is the predominance of the first aspect of driving improving profitability. As far as the leverage goes, I think gross margins will improve over time. It's just going to take a while. And we still have a long way to go on sales and marketing is reducing that as a percentage of revenue. I think G&A has room.

所以，營收成長是推動獲利能力改善的第一個面向中最主要的因素。至於營運槓桿，我認為毛利率會隨時間推移而改善。只是需要一些時間。而且我們在銷售與行銷費用占營收比重的降低方面，仍然還有很長的路要走。我認為一般與行政費用（G&A）也還有改善空間。

And I do think that over time, we'll see greater and greater efficiencies with the R&D organization as they adopt more and more capabilities to build products at a faster rate.

而且我確實認為，隨著時間推移，研發組織在採用越來越多能力、以更快速度打造產品的同時，我們會看到越來越高的效率。

Okay. And then just a quick follow-up. Can you remind us, should there be any -- now that we're on a different kind of pricing and packaging model, a little bit more variable, I guess, if you will. But should there be any difference in terms of the seasonality of your revenue ramp or recognition as we move forward with the new packaging?

好的。接著快速追問一下。可以提醒我們一下嗎——既然我們現在採用不同的定價與包裝模式，我想可以說是更具變動性一些。但在我們推進新包裝之後，營收爬坡或認列的季節性，是否會有任何差異？

So on revenue, I'd say, you get a fairly predictable pattern under which revenue is recognized because as I said, most of our revenue in the future periods is designated by our RPO, the committed contracts. But ARR will follow a very typical seasonal pattern. My expectation are bit more on in the enterprise selling basis. Q1 will always be our weakest as far as net new ARR adds as we're adding new territories, adding new reps, implementing new strategic initiatives. This year, in particular, we're educating the sales teams on not only the new pricing and packaging, but a lot of the new products we have.

就營收而言，我會說營收認列會呈現相當可預測的模式，因為如我所說，我們未來期間的大部分營收是由我們的 RPO（剩餘履約義務），也就是已承諾合約所界定。但 ARR（年度經常性收入）會遵循非常典型的季節性模式。我的預期更多是基於企業銷售的節奏。就淨新增 ARR 而言，第一季一向會是我們最弱的季度，因為我們會新增銷售區域、增加新業務代表、並推動新的策略性計畫。尤其是今年，我們正在教育銷售團隊，不僅是新的定價與包裝，還包括我們推出的許多新產品。

So every year, you're going to have that. And so it will be a slow start and then pick up. This year, too, just to remind everybody, we also had some big changes in our sales and marketing leadership, which is predominance of what you see now flowing through and the cost benefit from a lower sales and marketing as a percentage of revenue, and that's from efficiencies we're driving.

所以每一年都會有這種情況。因此一開始會比較慢，之後再加速。另外也提醒大家，今年我們的銷售與行銷領導層也有一些重大變動，這正是你們現在看到逐步反映出來的主要原因；而銷售與行銷費用占營收比重下降所帶來的成本效益，來自於我們正在推動的效率提升。

Our next question will come from the line of Scott Berg from Needham, followed by Billy Fitzsimmons.

下一個問題來自 Needham 的 Scott Berg，接著是 Billy Fitzsimmons。

Hi, Spenser, Andrew, nice quarter. I wanted to follow up on sales enablement that Andrew was chatting about there. We did a couple of different customer checks in the quarter. And the one thing that we came back is, I don't think your existing customers are quite aware of all of the different modules and innovation that you've rolled out this year.

嗨，Spenser、Andrew，本季表現不錯。我想追問一下 Andrew 剛才提到的銷售賦能（sales enablement）。我們在本季做了幾次不同的客戶訪查。我們得到的一個回饋是，我不認為你們現有客戶充分了解你們今年推出的各種模組與創新。

Yes. Totally.

是的。完全同意。

I see Spenser smiling. I know that's a function of timing, obviously. And one customer didn't even know that you had acquired Statsig. So, I guess where are you kind of in that journey? When is the sales force properly ramped in that?

我看到 Spenser 在笑。我知道這顯然與時間點有關。而且有一位客戶甚至不知道你們已經收購了 Statsig。所以我想問，你們目前在這段旅程的哪個階段？銷售團隊什麼時候才能在這方面真正完成爬坡、到位？

I mean, the quarter's sales results were good as is, but obviously, better awareness there can be even more helpful.

我的意思是，本季的銷售結果本來就不錯，但顯然，如果能有更高的認知度，會更有幫助。

Yeah, to your point, I think a lot of people still bucket us in the analytics company, and it drives me absolutely crazy. Honestly, just sharing, hey, we have Statsig now, and this is bleeding edge feature experimentation. And you can use it too, and this is the same infrastructure. OpenAI runs internally, like awesome.

對，你說得沒錯，我覺得很多人仍然把我們歸類為一家分析公司，這真的讓我非常抓狂。老實說，光是分享「嘿，我們現在有 Statsig 了，這是最前沿的功能實驗（feature experimentation）。你也可以用，而且它是同一套基礎架構。OpenAI 內部就在用，超棒。」

A lot of customers don't even know that. And then same with Wave, I think they're just starting to understand Wave and then same with our other products. I think if you remember from the prepared remarks, like we do see ramping. So, we're moving customers from one to two to three to four to five to more products, but it's much slower and that drives me crazy. I think there is no substitute for the work of like, hey, we built something amazing.

很多客戶甚至不知道這件事。Wave 也是，我覺得他們才剛開始理解 Wave，其他產品也是一樣。我想如果你記得我們事先準備的講稿，我們確實看到在爬坡。所以，我們正在把客戶從一個產品帶到兩個、三個、四個、五個甚至更多產品，但速度慢很多，這讓我很抓狂。我認為沒有任何事情能取代那種扎實的工作：像是「嘿，我們做出了一個很了不起的東西」。

We have to educate the hundreds of people we have in our field. And then they have to educate the thousands of customers in market. Like that's just work and that's just the whole thing. Something I'm spending a lot of time with Nate, our Chief Commercial Officer as well as the rest of the executive team on in terms of how do we get that and do that more efficiently. We just had a kickoff a few weeks ago where we showed off a lot of what you saw today with Statsig and Wave and custom agents.

我們必須教育我們在前線的數百位同仁。然後他們必須教育市場上的數千位客戶。這就是工作，這就是整件事的本質。我也花很多時間和我們的商務長（Chief Commercial Officer）Nate 以及其他高階主管團隊一起思考：我們要如何更有效率地做到這件事。幾週前我們剛辦了一場 kickoff，展示了很多你們今天看到的內容，包括 Statsig、Wave 和自訂代理（custom agents）。

But that's not even to say the -- all the other products we have like Session Replay and Guides and Surveys and AI Feedback that can displace point solutions. So anyway, that is -- I think last year, we said it was the year of the platform. I think we still have a ways to go on educating people on it. I will say that the good news on it is the main thing customers are looking for is, hey, prove to me you guys are at the bleeding edge of where this field is going. And so my view is that analytics and the whole data -- behavioral data ecosystem is going to go through the same shift that coding has in the last two years, like that is still going to happen.

但這還沒算上我們其他所有產品，例如 Session Replay、Guides、Surveys，以及能取代單點解決方案的 AI Feedback。總之——我想去年我們說那是平台之年。我認為在教育大家理解平台這件事上，我們仍然還有一段路要走。不過我會說好消息是，客戶最在意的是：「嘿，證明你們站在這個領域發展方向的最前沿。」因此我的看法是，分析以及整個——行為資料（behavioral data）生態系，將會經歷和過去兩年寫程式（coding）同樣的轉變；這件事仍然會發生。

And so they want to -- we see it in like a lot of the stuff we've been demoing and our customers see it, too. And so they want to know, hey, am I working with the company that's bleeding edge on this? And so even if they're not necessarily ready to adopt a Wave or even a Statsig, I know that, okay, you at least help me take the first step to using some of the basics on these capabilities, and then I can add more even if it's maybe too overwhelming for me right at the start or I'm not ready as a company. So anyway, that's all to say we still have a bunch of work to do to make sure our field is equipped. There's definitely areas that do it extremely well, but then there's areas we need to do a better job on this.

所以他們想要——我們在很多我們一直在 demo 的內容中看得到，客戶也看得到。因此他們想知道：「嘿，我是不是在和一家在這方面最前沿的公司合作？」所以即使他們未必已準備好採用 Wave，甚至 Statsig，我也知道：好吧，至少你能幫我踏出第一步，先使用這些能力中的一些基礎功能；之後我可以再加更多，即使一開始可能對我來說太過龐雜，或是我們公司還沒準備好。總之，這些都在說明：我們仍然有很多工作要做，確保我們的前線團隊具備足夠的能力。確實有些區域做得非常好，但也有些區域我們需要做得更好。

So, I appreciate you calling that out.

所以，謝謝你把這點提出來。

Thanks for that, Spencer. And then from a -- my follow-up question is on the integration traction with Statsig. You all had a pretty aggressive goal, obviously, to move that asset into your organizations. Kind of where are you with it? Because the other customers that we spoke with were super excited about that. A couple of them were Statsig customers, et cetera.

謝謝你，Spencer。接著我追問的是——關於與 Statsig 的整合進展。你們顯然訂了相當積極的目標，要把那項資產納入你們的組織運作之中。你們目前進度到哪裡了？因為我們訪談的其他客戶對此非常興奮。其中有幾位本來就是 Statsig 的客戶等等。

So, just kind of understand, have you hit all your goals around that? And are you kind of at that point where now you can just deliver on product and sales versus just having to integrate the organization?

所以想了解一下，你們是否達成了相關的所有目標？以及你們是否已經到了這個階段：現在可以把重心放在產品與銷售交付上，而不是還需要花很多力氣整合組織？

Yeah. So as you imagine, Statsig as been around for five years, and there's a lot of work with getting it from a whole group of people who have never seen the code base or sold it or whatever else. I think we've kind of gotten through -- there's always stuff, but we've gotten through all of the urgent fires in running and delivering Statsig. So, that's great. Customers are very excited about how it's landing.

是的。如你所想像，Statsig 已經存在五年了，要讓一整群從未看過程式碼基底、或從未銷售過它等等的人上手，確實有很多工作要做。我想我們大致已經度過——當然永遠還會有事情——但我們已經處理完在營運與交付 Statsig 上所有緊急的火線問題。所以這很棒。客戶對它的落地情況非常興奮。

We want to make sure to give -- the fact that it's our main focus as opposed to at OpenAI, it was a little more of a side thing for them. It's all been received positively. So, that's good. Now we're starting to think about, okay, what's coming next for Statsig? So if you look at like statsig.com/updates, we're shipping stuff.

我們想確保能夠傳達——這是我們的主要重點；相較之下，在 OpenAI 那邊，對他們而言更像是一個比較偏支線的事情。整體回饋都很正面。所以，這很好。現在我們開始思考，好，Statsig 接下來會是什麼？所以如果你看像 statsig.com/updates，我們正在持續發布新功能。

We've been shipping stuff for the last few months. We're continuing to build on the road map. We're continuing to integrate with Amplitude much more tightly so that if you're on both, which a lot of our customers are, you get the benefits of being able to use data from one and the other. And I think a lot of -- the other thing we're seeing with Statsig is that there's a lot of demand from AI-natives in particular. So, one of the reasons we're really excited to join forces with Statsig is that they -- like a lot of the way the future product development is being run, like people are choosing Statsig for that.

過去幾個月我們一直在發布新功能。我們也持續依照產品路線圖推進。我們也持續更緊密地與 Amplitude 整合，讓同時使用兩者（我們很多客戶都是如此）的客戶，能享受到可以互相使用彼此資料的好處。而且我認為，我們在 Statsig 看到的另一件事是，特別是 AI 原生（AI-native）公司有很大的需求。所以，我們之所以非常興奮能與 Statsig 攜手合作，其中一個原因是——未來產品開發的運作方式在很大程度上就是如此，很多人因此選擇 Statsig。

So it's engineering-first teams that tend to be much more technical. They're building out whole software development harnesses. They want to manage how stuff is deployed in that harness and Statsig is set up really, really well to scale. As I mentioned, OpenAI runs a version of that infrastructure internally for themselves. And so they've tested that in tons of different ways over there, and we're doing the same thing except with everyone outside of OpenAI.

因此，通常是工程優先（engineering-first）的團隊，技術含量更高。他們在打造完整的軟體開發「工具鏈／框架」（harness）。他們希望在這個框架中管理部署方式，而 Statsig 的設計非常、非常適合擴展。如我先前提到，OpenAI 在內部為自己運行了一個版本的那套基礎設施。所以他們在那邊用各種方式大量測試過，而我們做的是同樣的事，只是服務對象是 OpenAI 以外的所有人。

And so there's a lot for us to do in terms of how do you set Statsig up to be a core part of the software development harness for all these bleeding-edge AI customers, and it's where kind of everyone wants to go over time. So, that's what we're focused on.

因此，我們有很多工作要做，包含如何把 Statsig 設定成這些最前沿 AI 客戶的軟體開發框架中的核心組成，而這也是隨著時間推移大家都想走向的方向。所以，這就是我們目前專注的重點。

Awesome. Thanks for taking my questions.

太棒了。謝謝你回答我的問題。

Of course, Scott.

當然可以，Scott。

Our next question will come from Billy Fitzsimmons from Piper Sandler, followed by Clark Wright from D.A. Davidson.

下一個問題來自 Piper Sandler 的 Billy Fitzsimmons，接著是 D.A. Davidson 的 Clark Wright。

Hey, guys. Good to see the results and guidance. I think one of the exciting things about Statsig is potentially the cross-sell opportunity. I know there are some things to do first. But last time I looked or last I checked, I think there were 80 of the 400 Statsig customers are on Amplitude already. So, there's a lot who aren't.

嗨，各位。很高興看到這次的業績與指引。我認為 Statsig 令人興奮的一點是潛在的交叉銷售機會。我知道在那之前還有一些事情要先做。但我上次看或上次確認時，我記得 400 個 Statsig 客戶中大約有 80 個已經在用 Amplitude。所以，還有很多客戶尚未使用。

Can you just help contextualize for us how we should think about the potential cross-sell opportunity of Amplitude into Statsig or potentially vice versa and how we should think about that flowing through the model long term?

你們能否協助我們理解：我們應該如何看待把 Amplitude 交叉銷售給 Statsig 客戶、或反過來的潛在機會？以及長期來看，這會如何反映在你們的模型中？

I think probably the much bigger opportunity is to take Statsig to Amplitude customers. I think Statsig customers, as I mentioned earlier, tend to be much more bleeding edge from an AI innovation standpoint. And so that's where everyone is trying to get their organizations to over the long term. It is a very -- it's like a more -- Amplitude is historically focused on product management and then Statsig is much more tailored towards engineers like it has tons of customization out of the box. It has like all the statistical testing.

我認為更大的機會，可能是把 Statsig 帶給 Amplitude 的客戶。我先前提到，Statsig 的客戶從 AI 創新角度來看通常更偏向最前沿（bleeding edge）。而長期來看，這正是大家都在努力把組織帶往的方向。這是一個非常——可以說更——Amplitude 歷史上更聚焦於產品管理，而 Statsig 則更針對工程師，因為它開箱即用就有大量客製化能力。它也具備各種統計檢定。

Now, like I said, those two personas are merging, but it's early days on that. So, I think the opportunity is as more of our traditional Amplitude customers look and try to build like AI-natives, introduce AI to their software development process, try to build out a harness, eventually try to get to self-improving products, all of those are opportunities for us to bring Statsig. Now, we definitely do see places where Statsig customers are also very interested in Amplitude, but there's a lot more both from a number and ARR basis that are Amplitude.

如我所說，這兩種角色正在融合，但目前仍在早期階段。所以，我認為機會在於：當我們更多傳統的 Amplitude 客戶開始觀察並嘗試打造 AI 原生能力、把 AI 引入他們的軟體開發流程、嘗試建立一套框架，最終嘗試做到可自我改進的產品——這些都是我們導入 Statsig 的機會。當然，我們也確實看到 Statsig 客戶對 Amplitude 也很有興趣，但無論從客戶數量或 ARR 的角度來看，Amplitude 的基礎都更大。

Perfect. And then if I could ask a second one. Can you just contextualize maybe how either your hiring needs have kind of changed year-to-date or where you're seeing the best ROI from AI-driven efficiencies internally within Amplitude?

了解。那如果我可以再問第二個問題。你能否說明一下：截至目前為止，你們的招募需求有什麼變化？或是在 Amplitude 內部，你們在哪些地方看到由 AI 驅動的效率帶來最佳的投資報酬（ROI）？

There's a ton. On the hiring front -- so a few different things. One, like it's been -- I've been just very focused on transforming the entire workforce, getting leaders, getting engineers, getting people in other functions that are AI-native, both by hiring that talent, acquiring it, hiring executives that have that background. And then in addition to that, retraining the -- and re-educating the workforce that we have here, great part, like everyone wants to learn. It's like, yes, people see like, hey, the more I can learn how to use AI, the more relevant my skills are going to be both at Amplitude and at other places in the future.

很多。在招募方面——有幾件不同的事。第一，我一直非常專注於轉型整個人力結構：找領導者、找工程師、找其他職能的人才，讓他們是 AI 原生的；做法包含招募這類人才、透過併購取得人才、招募具備這種背景的高階主管。除此之外，也要重新訓練——並再教育——我們現有的員工；很棒的一點是，大家都想學。大家會覺得：是的，我越能學會如何使用 AI，我的技能在 Amplitude 以及未來其他地方就會越有相關性。

So, everyone is like embracing it, which is great. The few specific areas, I think on -- so yes, that's like an always ongoing thing. Like I just -- we were just adding Angela, which we announced today in marketing. We're always looking at companies and other places to pick up talent. New Grads is another great source of very highly leveraged talent.

所以大家都在擁抱它，這很棒。幾個具體領域，我認為——所以是的，這是一個持續進行的事情。像我們剛加入 Angela，我們今天也在行銷方面宣布了。我們也一直在看一些公司和其他地方，以便延攬人才。應屆畢業生（New Grads）也是另一個能高度槓桿化的人才來源。

One of the funny things I'll tell you guys is during downturns or whatever, a lot of companies pull back on university hiring because it's like the easiest thing to cut. But if you have the confidence to evaluate who is great from that talent pool, you can get some exceptional folks right out of school, which is awesome. So we've been -- had that as a big focus here at Amplitude. So, that's like the primary thing. And then the one specific area is Statsig.

我跟各位分享一個有趣的現象：在景氣下行之類的時候，很多公司會縮減大學校園招募，因為那是最容易砍的項目。但如果你有信心能從那個人才池評估出誰很優秀，你就能直接從學校招到一些非常出色的人才，這很棒。所以我們在 Amplitude 一直把這當作一個重要重點。所以，這是主要的部分。然後有一個特定領域就是 Statsig。

As you imagine, this is a huge complex product and code base and architecture. And so our -- we've taken our existing experimentation team, and they're now running Statsig, which is awesome, but they also need a lot more help. So, we're adding lots of different roles in hiring on that data science leads for deployed engineers, other engineers who are just familiar with that architecture. We've actually hired one person who used to work at Statsig, pre the OpenAI acquisition and we're continuing to go more there. So, there's a lot we need to do there.

如你所想，這是一個非常龐大且複雜的產品、程式碼庫與架構。因此，我們——我們把現有的實驗（experimentation）團隊調整後，現在由他們來負責 Statsig，這很棒，但他們也需要更多支援。所以我們正在增聘許多不同職務，包括資料科學主管、負責部署的工程師（deployed engineers）、以及其他熟悉該架構的工程師。我們其實已經聘到一位曾在 Statsig 工作的人，是在 OpenAI 收購之前就在那裡的，我們也會持續朝那個方向加強。所以那邊還有很多事情要做。

We've kind of caught the ball, which is good, but now we have to like go maximize it.

我們算是把球接住了，這很好，但現在我們得把它的價值最大化。

Great to see. Thanks, guys.

很高興看到這些進展。謝謝各位。

Our next question will come from Clark Wright from D.A. Davidson, followed by Koji Ikeda from Bank of America.

下一個問題來自 D.A. Davidson 的 Clark Wright，接著是美國銀行（Bank of America）的 Koji Ikeda。

It was great to see the 30% year-over-year increase in customers with over $100,000 in ARR, which looks to be the highest since 2021. Can you potentially break out the adds from Statsig? And what else is helping in terms of the new logo momentum that you're seeing today?

很高興看到 ARR 超過 10 萬美元的客戶數年增 30%，看起來是自 2021 年以來的最高水準。你們能否把其中來自 Statsig 的新增客戶拆分出來？另外，就你們目前看到的新客戶（new logo）動能而言，還有哪些因素在推動？

Sure. So, about 40 customers came from the Statsig business itself that we added. And so if you kind of do the quick math on that, you're still well in almost 23%, 24% growth in customers that are in that greater than $100,000 cohort. And so it's still growing quite nicely and contributing to ARR and to revenue growth. So, that was really good.

當然。所以，大約有 40 位客戶是來自我們新增併入的 Statsig 業務本身。如果你快速算一下，即使扣除這部分，ARR 超過 10 萬美元的那個客群，客戶數仍接近 23%、24% 的成長。所以它仍然成長得相當不錯，並且對 ARR 與營收成長都有貢獻。所以，這點真的很好。

And as Spenser mentioned earlier, what we're seeing back when we're talking to customers, especially as we've gotten introduced them for the first time with their brand-new customers to Amplitude that were formerly Statsig customers is we're finding that they're, one, very appreciative of the fact that Amplitude is shepherding and taking forward the road map and showing confidence in our ability to actually give them a future where self-improving products is a reality.

另外，如同 Spenser 先前提到的，當我們回頭與客戶交流時，特別是我們第一次以 Amplitude 的身分接觸那些原本是 Statsig 客戶、現在成為 Amplitude 全新客戶的公司時，我們發現他們首先非常感謝 Amplitude 正在引導並推進產品路線圖，並且對我們真正有能力帶給他們一個「自我改進產品」成為現實的未來展現出信心。

And they do that through adopting an experimentation mindset, and they're very confident then to move further with Amplitude in other areas. So, that cross-sell expansion opportunity is real. I think we talked about it at the time, there was a multi-hundred million dollar opportunity for us just in the installed base. So, we're pretty excited about it.

他們是透過採納實驗（experimentation）的思維模式來做到這一點，因此也更有信心在其他領域進一步與 Amplitude 合作。所以，交叉銷售（cross-sell）的擴張機會確實存在。我想我們當時也談過，僅在既有裝機基礎（installed base）中，對我們而言就有數億美元規模的機會。所以我們對此相當興奮。

Got it. And then last quarter, you called out event volume growth being 21% year-over-year. What is that now as you kind of talk about the momentum that you're seeing in all-time highs? And how should we think about the ramp of that metric going forward given agentic workflows and the amount of events that they can process?

了解。另外，上季你們提到事件量（event volume）年增 21%。那現在呢？你們提到目前看到的動能創下歷史新高。考量到代理式（agentic）工作流程以及它們可處理的事件量，未來我們應該如何看待這個指標的爬坡（ramp）？

Yeah, It's definitely growing faster than both ARR and revenue. And it's one of those areas that -- for us, it feels like we've gone through many, many quarters of trying to bring it up and get the entitlements rightsized and everything else. It's definitely a leading indicator for us that, one, we're not going to have the same types of churn issues like in the past. Two, sales has adopted that value-based orientation sale where they're not trying to get everything upfront.

是的，它的成長速度確實快於 ARR 與營收。而且這是其中一個領域——對我們來說，感覺我們已經歷經很多很多個季度，努力把它拉上來、把權益（entitlements）調整到合適水位，以及處理其他相關事項。對我們而言，這絕對是一個領先指標：第一，我們不會再像過去那樣面臨同類型的流失（churn）問題。第二，銷售團隊已採用以價值為導向的銷售方式，他們不再試圖一開始就把所有東西都賣出去。

They're trying to get our customers to value quickly and show them the value of an expansion. And like I said, it's an indicator that we're going to see upsells have a larger meaningful contribution to growth, whereas before it was a tractor and the predominance of our growth with cross-sell. We're just not going to have those same instances if we've got customers who are bumping up against their entitlements and getting value from the investment they've made.

他們的目標是讓客戶快速獲得價值，並向客戶展示擴充（expansion）的價值。而且如我所說，這也是一個指標，顯示我們將看到加購（upsell）對成長做出更大且更具實質意義的貢獻；而在過去，我們的成長主要是靠交叉銷售（cross-sell）這台「拖拉機」在拉動。如果客戶開始碰到其權益上限，並且從他們已投入的投資中獲得價值，我們就不會再出現同樣的情況。

Got it. Thank you.

了解。謝謝。

Our next question will come from Koji Ikeda from Bank of America, followed by Nick Altmann.

下一個問題來自美國銀行（Bank of America）的 Koji Ikeda，接著是 Nick Altmann。

Thank you. Thanks, guys. Thanks so much. I wanted to ask a question on Wave. Love the demo, long-term vision. I mean, it sounds like it's going to be awesome for finding problems and finding solutions, generating code, measuring outcomes. I mean, it looks like the full deal here. And so the question really becomes, if Wave is successful in all the things I think it could be, then why would you need the other products from Amplitude like Statsig and product analytics?

謝謝。謝謝各位。非常感謝。我想問一個關於 Wave 的問題。我很喜歡這個展示與長期願景。聽起來它將非常適合用來找出問題與解決方案、產生程式碼、衡量成果。看起來就是完整的一套。所以問題就變成，如果 Wave 在我認為它可能做到的所有事情上都很成功，那為什麼你們還需要 Amplitude 的其他產品，例如 Statsig 和產品分析（product analytics）？

Seems like you could do it all from Wave.

看起來好像用 Wave 就能全部搞定。

Yeah, totally. Okay. So yes, this is -- I brushed over this architecturally. What Wave does is it takes data from lots of different data sources. So it takes analytics data from Amplitude, experiment data from Statsig.

對，完全同意。好。所以是這樣——我在架構層面稍微帶過。Wave 的作用是從許多不同的資料來源取得資料。因此它會取得來自 Amplitude 的分析資料（analytics data）、來自 Statsig 的實驗資料（experiment data）。

We're eventually -- we're planning to make it agnostic long term, so it can take data from any analytics thing. If you're using Google Analytics or Adobe or something else, it doesn't matter, and then translate that insight. So, you still need a place to get that data. Like it's not like it can just look at a product and figure out what people are doing it. It actually needs to have that data from some area.

長期來看，我們最終——我們計畫讓它成為中立（agnostic）的，因此它可以接收任何分析工具的資料。如果你使用 Google Analytics、Adobe 或其他工具，都沒關係，然後再把那些洞察轉譯出來。所以，你仍然需要一個地方來取得那些資料。它並不是說可以直接看著一個產品就推斷人們在做什麼。它實際上需要從某個地方取得那些資料。

And so it's a nice build where it's like, hey, use Amplitude, use Statsig. The more data sources you put into this thing, the better the output that you see. One of the big learnings from the AI boom is that the power of massive scale of data, it just gets you better and more accurate and more insightful results like that is just a straight -- that -- like you can -- the scaling laws look like you can grow that almost infinitely. So, Amplitude Analytics actually as well as the experimentation and everything else we have, play a really important part in being the collection points for that data. Again, though, the goal is to be agnostic, so we can just plug into whatever system, your data warehouse, your own internal thing, other tools, third-party tools and kind of build it on top of that.

因此這是一個很好的組合：像是，嘿，使用 Amplitude、使用 Statsig。你把越多資料來源接進來，你看到的輸出就會越好。AI 熱潮帶來的一個重要學習是：大規模資料的力量會讓結果更好、更準確、也更有洞察——這就是一條直接的——也就是——你可以看到，規模法則（scaling laws）看起來幾乎可以無限成長。所以，Amplitude Analytics 以及我們的實驗能力與其他一切，在作為這些資料的蒐集點方面扮演非常重要的角色。不過再說一次，我們的目標是中立（agnostic），因此我們可以接入任何系統：你的資料倉儲、你們自己的內部系統、其他工具、第三方工具，然後在其上方建構。

I think another thing is that because we have that data, that gives us the ability to have much greater insight into the right things to build. If you're a start-up starting out for the first time and you don't have the massive multiple petabyte data set that we have, it's like, okay, how do you even know if what you're recommending is best practice or what leads to something good? And so there's a lot of feedback loops that we have because we have this data set, we know, okay, hey, here's what a great e-commerce app looks like. Here's what a great social media app looks like. Here's what a fintech app should look like.

我認為另一點是，因為我們擁有那些資料，這讓我們能對「應該打造什麼」有更深入的洞察。如果你是一家剛起步的新創，第一次開始做，並沒有我們這種多 PB（petabyte）等級的龐大資料集，那就會變成：好吧，你要怎麼知道你所推薦的是最佳實務，或什麼會帶來好的結果？因此我們有很多回饋迴路（feedback loops），因為我們擁有這個資料集，我們知道：好，嘿，一個很棒的電商 App 長什麼樣子。一個很棒的社群媒體 App 長什麼樣子。一個金融科技（fintech）App 應該長什麼樣子。

Here's the typical workflows for sign-up that work well. Here's what message customization should be so and so on. And so because, like, we're one of the few companies out there, there's no open source equivalent data sets for it. And so having that allows us to develop a much higher quality, better version of Wave than kind of anyone else out there. So the other good part is it's not like a -- it's an alpha.

哪些註冊（sign-up）的典型工作流程效果最好。訊息客製化（message customization）應該怎麼做，等等。而且因為——我們算是少數幾家公司之一，市面上沒有對應的開源資料集。因此擁有這些資料，讓我們能開發出品質更高、版本更好的 Wave，勝過其他任何人。所以另一個好處是，它不是——它目前是一個 alpha 版本。

So, there are customers using it. We're using it internally. There's a number of start-ups. There's a few enterprises that are using it. And so it's spitting out real things that frankly, you look at this and you're just like how did AI come up with this?

所以，已經有客戶在使用。我們也在內部使用。有一些新創公司。也有少數企業客戶在使用。因此它會產出真實的成果，坦白說，你看到這些就會想：AI 怎麼會想得出這些？

This is crazy. I'm convinced that whoever wins this space, that's going to be a multibillion-dollar business, if not more. And so our thing is like let's run forward with that as fast as possible. I think we're well positioned in the opportunity because we're the leader in analytics and a few other areas. Yes.

這太瘋狂了。我深信，誰能在這個領域勝出，那將會是一門數十億美元的生意，甚至更多。所以我們的做法就是盡可能快地往前衝。我認為我們在這個機會上處於很好的位置，因為我們在分析以及其他幾個領域都是領導者。是的。

And let's go build that business as quickly as we can.

我們就盡快把那門生意做起來。

Got it. Thanks, Spencer. All from me thank you so much.

了解。謝謝你，Spencer。我這邊就到這裡，非常感謝。

Of course, Koji.

當然，Koji。

Nick Altmann from BTIG, followed by YC Wong from Citi.

接下來是 BTIG 的 Nick Altmann，然後是花旗的 YC Wong。

Hey, Awesome. Just to build off Koji's last question, I kind of wanted to ask the inverse on Wave of, like, it seems like there's more incentive to adopt the broader platform with Wave. And I know it's still very early, but how are those kind of conversations going with customers? Like are you having more sort of multi-product or platform adoption customers as they kind of look at Wave and this vision of the self-improving product? And then the follow-up there is just how should we think about Wave being monetized more so in the near term?

嗨，太棒了。延續 Koji 最後一個問題，我想從 Wave 的反面來問：看起來使用 Wave 會更有誘因去採用更廣泛的平台。我知道現在還很早，但你們跟客戶的這類對話進展如何？例如，當客戶看待 Wave 以及這個「自我改進產品」的願景時，你們是否看到更多多產品或平台式採用的客戶？另外一個追問是：我們應該如何看待 Wave 在短期內的變現方式？

Is it kind of indirectly in the sense of it gives customers more incentive to adopt the broader platform, and that's how you sort of plan to monetize it? Or is it kind of a stand-alone SKU?

它是否是以間接方式變現，也就是讓客戶更有誘因採用更廣泛的平台，並以此作為你們的變現路徑？還是它會是一個獨立的 SKU？

Yeah, so you're exactly right, which is the more data sources you feed to this thing, the better. And so we've already -- I've already seen multiple customers who have gotten on Session Replay as well as one that signed up for AI Feedback specifically because, hey, the stuff fed to Wave makes it a lot better. And so you're absolutely right, where like it drives like the whole platform play where it's like, okay, you have all these individual point things and then you just -- there are more data sources. Session Replay in particular, is very, very powerful.

對，所以你說得完全正確：你餵給它的資料來源越多，它就越好。因此我們已經——我已經看到多位客戶同時上了 Session Replay，也有一家是因為「嘿，餵給 Wave 的內容會讓它更好」而特別去訂閱 AI Feedback。所以你說得完全對，它會帶動整個平台策略：你有這些單點產品，然後——資料來源越來越多。尤其是 Session Replay，真的非常、非常強大。

Like as you imagine, viewing the exact state of UI and where a user clicked has a lot of value for how it can be better. So, that's been awesome to see. And again, early, there's a handful of customers on it. But as we grow it out, I think that will drive more adoption. And I also don't think like -- to my point earlier to Koji, it's like, the goal is to be agnostic with it.

如你所想，能看到 UI 的精確狀態以及使用者點擊的位置，對於如何把產品做得更好非常有價值。所以看到這點很棒。而且再次強調，現在還早，只有少數客戶在用。但隨著我們把它擴大，我認為會帶動更多採用。另外我也不認為——就像我先前對 Koji 說的，我們的目標是讓它保持中立、可相容。

We want to build the most bleeding-edge thing. And so if we plug in other sources too, all the better. On the monetization front, we'll charge for it. We absolutely will charge for it. I mean, you think about the value that this creates.

我們想打造最前沿的東西。所以如果也能接入其他來源，那就更好。在變現方面，我們會收費。我們絕對會收費。畢竟你想想它能創造的價值。

Now, you go from analytics or data tooling where it's like you have to manually go in, collect an event or look at -- ask a particular question, get a result out, think about how to apply to the business. And now you're having a whole flow that does it for you, hey, I've already seen this user is having friction here like the docs example I made is like, hey, we see most search queries are failing. Why is that? Well, they're single characters, and we're not waiting until someone types a complete word, so they get this error when they're in the middle of the typing and it feels bad. And it's like, okay, yes, you should resolve that and make that better.

你會從分析或資料工具那種模式——必須手動進去、蒐集事件或查看——提出某個特定問題、得到結果、再思考如何應用到業務——轉變成一整套流程替你完成：例如，我們已經看到這位使用者在這裡遇到摩擦；就像我舉的文件（docs）例子：我們看到大多數搜尋查詢都失敗了。為什麼？因為他們只輸入單一字元，而我們沒有等到使用者輸入完整單字，所以在他們打字到一半時就出現錯誤，體驗很差。那就變成：好，對，你應該解決它，把它做得更好。

And it's not just that. It's like that times hundreds of things all across all surface areas of your product. One of the lessons is that like behavioral data and product surface areas are so large, it is impossible for any team to stay on top of them. And so the fact that this thing is looking all the time for how it can be better is it's magical. Like it's crazy what it can do.

而且不只是這樣。這會是那種情況乘以數百件事，遍佈你產品的所有接觸面。其中一個教訓是：行為資料與產品接觸面太龐大，任何團隊都不可能完全掌握。所以這個東西能一直在看、一直在找如何變得更好，真的很神奇。它能做到的事情很誇張。

So, I think whatever company goes to win that is going to be multiple billions in revenue, if not more, and we want to aggressively go after it. And yes, customers are willing to pay for that. Now again, early days, we're in alpha. So, we haven't figured out exactly how we're going to monetize it, but we absolutely will charge for that capability. That's like -- that's one of the great -- people are talking about, hey, there's all this money going to AI, where does it actually come out?

所以我認為，無論哪家公司能贏下這個領域，營收都會是數十億美元等級，甚至更多，而我們想積極進攻。而且是的，客戶願意為此付費。當然現在還是早期，我們在 alpha 階段。所以我們還沒完全想清楚要怎麼精準變現，但我們絕對會為這項能力收費。這也是——大家一直在談：有這麼多資金投入 AI，那實際上錢會從哪裡賺回來？

And this is one where you can draw the line really directly. It's like, look, the customer experience is getting better. They're spending more. There's more revenue. There's less friction, there's less downtime, like the whole thing is just better, like great use from an application standpoint.

而這就是一個你可以非常直接畫出因果線的例子。也就是：客戶體驗變得更好。他們花得更多。營收更多。摩擦更少、停機更少——整體就是更好，從應用角度來看是很棒的用例。

Great.

很好。

Thank you so much.

非常感謝。

For sure.

沒問題。

Our next question will come from YC Wong from Citi, followed by Arjun Bhatia from William Blair.

下一個問題來自花旗的 YC Wong，接著是 William Blair 的 Arjun Bhatia。

Hey, good evening. Thanks for taking your question here. Spenser and team, great to see the fast expanding AI platform, here you have like every quarter. Like, I want to touch on Agent Analytics, which now seems to measure like agent themselves, right? I mean, where the market that we see is already multiple vendors out there trying to measure prompts, measure latency, hallucination, like, all the stuff that you can see, but what is the customer problems that the Agent Analytics could solve that the current observability platform cannot? And then how do you view the market opportunity of that problem?

嗨，晚安。謝謝讓我提問。Spenser 和團隊，很高興看到你們的 AI 平台每一季都在快速擴張。我想談談 Agent Analytics，現在看起來是用來衡量代理（agent）本身，對吧？我的意思是，我們看到市場上已經有多家供應商在嘗試衡量提示詞（prompts）、延遲、幻覺等各種你能看到的指標，但 Agent Analytics 能解決哪些現有可觀測性平台無法解決的客戶問題？以及你們如何看待這個問題的市場機會？

Yeah. So, I mean, I think, first, to the extent this replaces most traditional interfaces and the market opportunity is as large, if not larger, than what's going on, on traditional user interfaces with Session Replay and analytics. In terms of our unique positioning, what we offer, which I shared a little bit in the customer story about Economist is that you can connect what's individually happening within a session to the long-term impact to your business. So, you can say, okay, hey, you got a successful answer back from the bot, did that lead to you spending more or signing up or keeping your subscription? Conversely, if you ran into a problem and you got frustrated, did that lead to some negative long-term outcome?

是的。我想，首先，如果這在很大程度上取代了傳統介面，那它的市場機會會跟傳統使用者介面上的 Session Replay 與分析一樣大，甚至更大。就我們獨特的定位、我們提供的能力而言——我在關於 Economist 的客戶案例中也稍微分享過——你可以把單一工作階段（session）內發生的事情，連結到對你業務的長期影響。所以你可以說：好，機器人回覆成功了，那是否帶來更多消費、更多註冊，或是讓你續訂？反過來，如果你遇到問題、感到挫折，那是否導致某些負面的長期結果？

And that loop is really, really important. A lot of the engineering-specific observability products we've seen in this space are just kind of stand-alone. It's like, okay, they'll just show the traces and that's kind of it. And you have no idea if it's actually leading to different results down the line. And so that's why we see both traditional enterprises that are transforming their businesses like the economists as well as a lot of AI-natives.

而這個閉環非常、非常重要。我們在這個領域看到的許多偏工程導向的可觀測性產品，基本上都比較像是獨立工具。就像：好，它們只展示追蹤（traces），大概就這樣。而你完全不知道這是否真的會在後續帶來不同的結果。因此我們才會同時看到像 Economist 這樣正在轉型的傳統企業，以及許多 AI 原生公司。

I mentioned one of the largest foundational model companies. They also are looking at, as you imagine, they'll have a lot of tooling there, but they want to know, okay, is this leading to someone to becoming a subscriber to upsell and all of that sort of stuff long term. And so being able to connect that journey end-to-end is what we uniquely offer.

我提到過一家最大的基礎模型公司之一。他們也在評估，如你所想，他們那邊會有很多工具鏈，但他們想知道，好，這是否會在長期帶動某個人轉為訂閱用戶、進一步升級銷售（upsell）以及諸如此類的事情。因此，能夠把那段旅程端到端串起來，正是我們獨特能提供的。

That sounds like a more TAM expansion opportunity there.

聽起來這是一個擴大 TAM 的機會。

Absolutely. Absolutely. I didn't cover as much today. We demoed it more on the Q1 earnings call. But yes, it's actually one of the things my Chief Commercial Officer and I are very excited about.

完全是。完全是。我今天沒有講太多。我們在第一季財報電話會議上做了更多展示。但沒錯，這其實是我和我們的首席商務長都非常興奮的一件事。

Yeah, Definitely look forward to hearing more, including Wave. I have a quick follow-up for Andrew as well on the guidance. Like Amplitude growth has definitely been accelerating for the past year and more, right? Even adjusting for the Statsig business this quarter, I think it's still accelerated.

是的，非常期待聽到更多內容，也包括 Wave。我也想就財測對 Andrew 追問一個簡短問題。像 Amplitude 的成長在過去一年甚至更久確實一直在加速，對吧？即使把本季的 Statsig 業務因素調整掉，我認為成長仍然是加速的。

But the implied guide that I'm looking for Q4 shows about 2- to 3-point decel. Could you kind of help us double-click on the largest step down on the Q4 guide? Is it more just seasonality or incremental conservatism?

但我看到第四季隱含的指引顯示大約會減速 2 到 3 個百分點。你能否幫我們更深入地看一下第四季指引中最大幅度下修的原因？這主要只是季節性，還是額外的保守假設？

What I would tell you is that we always take a look at what -- when we're building our guidance, what we believe is very strong occurrence to occur. And I mentioned some of the factors earlier about pipeline, how well that pipeline has developed, where we're seeing good demand from our customers.

我會告訴你的是，我們在制定指引時，總是會檢視——在建立指引時——我們認為非常有把握會發生的情況。我先前也提到一些因素，包括管線、該管線發展得如何，以及我們從客戶端看到的良好需求。

Usually, Q4 is our strongest quarter from a new ARR perspective, and it's because that's the way we built our comp plans. That's the way enterprise selling cycles run typically in a calendar-based company. So, I would just tell you that our guidance is based upon what we know is out there as far as our pipelines, our RPO, and it's what we're comfortable with.

通常第四季從新增 ARR 的角度來看是我們最強的一季，因為我們的獎酬方案就是這樣設計的。在以曆年為基礎的公司裡，企業銷售週期通常也是這樣運作的。所以我只能說，我們的指引是基於我們在管線、RPO 等方面所掌握的現況，以及我們覺得舒適、可承擔的水準。

Got it. Congrats, guys. Thank you.

了解。恭喜各位。謝謝。

Arjun Bhatia of William Blair by Willow Miller.

接下來是 William Blair 的 Arjun Bhatia，由 Willow Miller 代為提問。

Team, thanks for taking our question. Can we hear your updated thoughts on the 20%-plus revenue growth target given the strong growth this quarter and the strong third quarter guide? I'm curious to hear how you're thinking about it now, considering Statsig and now Wave?

團隊，謝謝讓我們提問。在本季強勁成長以及第三季強勁指引之下，能否談談你們對「營收成長 20% 以上」目標的最新看法？我也想了解，在納入 Statsig、以及現在又有 Wave 的情況下，你們目前是怎麼思考的？

Yeah. I mean, I think Statsig is an accelerant to our long-term plans, which is part of why we -- Vijaye and I agreed Amplitude would be the best home for Statsig long term. I think the -- so we put up $19 million in organic growth last quarter in Q2. And so it's just -- we're just touching on that 20%. It's like the annual number is $410 million.

是的。我的意思是，我認為 Statsig 會加速我們的長期計畫，這也是為什麼我和 Vijaye 一致認為，從長期來看 Amplitude 會是 Statsig 最好的歸宿。我想——所以我們在上一季第二季的有機成長是 1,900 萬美元。因此——我們其實就是剛好觸及那個 20%。年化的數字大約是 4.10 億美元。

So if you divide that out, it's like we're just shy of that 20% growth target when you annualize the quarterly numbers. To me, as I've always said, 20% is kind of bare minimum, frankly. We want to be making sure to continually hitting and exceeding that 20%. Long term, we're aiming a good deal higher. We want to get to 30% and then beyond that as we continue to grow the business. Obviously, a lot of work between here and there, but that's what we're very focused on doing.

所以如果你把它拆開來看，把季度數字年化，我們其實距離 20% 的成長目標只差一點點。對我來說，正如我一直說的，坦白講 20% 算是最低門檻。我們希望確保能持續達成並超越 20%。長期來看，我們的目標要高得多。隨著業務持續成長，我們希望達到 30%，並在此之上再往上走。顯然從現在到那裡還有很多工作要做，但這就是我們非常專注在推進的方向。

Great to hear. Thank you.

很高興聽到。謝謝。

Thank you, Willow. That will conclude our second quarter earnings call. Thank you for your time and interest. We look forward to seeing you this quarter on the road as we attend conferences hosted by KeyBanc, Citi and Piper Sandler. Thank you.

謝謝你，Willow。這將結束我們第二季的財報電話會議。感謝各位的時間與關注。我們期待本季在外出行程中與各位見面，我們將參加由 KeyBanc、Citi 與 Piper Sandler 主辦的研討會。謝謝。

Thank you all.

謝謝大家。

Thank you.

謝謝。