Amplitude Inc (AMPL) 2025 Q4 法說會逐字稿

Good afternoon, everyone, and welcome to Amplitude's fourth-quarter and full year 2025 earnings 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.

各位下午好，歡迎參加 Amplitude 2025 年第四季及全年財報電話會議。我是投資者關係主管約翰‧斯特雷帕。今天與我一同出席的還有 Amplitude 的執行長兼共同創辦人 Spenser Skates，以及財務長 Andrew Casey。

During today's call, management will make forward-looking statements, including statements regarding our financial outlook for the first-quarter and full year 2026, the expected performance of our products, our expected quarterly and long-term growth, investments and our overall future prospects. 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.

在今天的電話會議上，管理層將發表前瞻性聲明，包括有關我們對 2026 年第一季和全年財務展望、我們產品的預期表現、我們預期的季度和長期成長、投資以及我們整體未來前景的聲明。這些前瞻性陳述是基於當前資訊、假設和預期，並受風險和不確定性的影響，其中一些風險和不確定性是我們無法控制的，可能導致實際結果與這些陳述中描述的結果有重大差異。

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與預期結果存在差異的風險的更多信息，請參閱我們向美國證券交易委員會提交的文件。請注意，不要過度依賴這些前瞻性陳述，除法律要求外，我們不承擔在今天的電話會議後更新這些陳述的義務。今天電話會議中使用的某些財務指標是按非GAAP準則表示的。

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.

我們內部使用這些非GAAP財務指標，以便分析我們的財務和業務趨勢，並用於內部規劃和預測。這些非公認會計準則財務指標有其局限性，不應脫離公認會計準則所編製的財務資訊單獨使用，也不應取代後者。

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.

有關這些非GAAP財務指標的更多資訊以及這些GAAP和非GAAP財務指標之間的調節表，請參閱我們的盈利新聞稿和補充財務信息，這些資料可在我們的投資者關係網站 investors.amplitude.com 上找到。

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

這樣，我就把電話交給史賓塞了。

Good afternoon, everyone, and welcome to Amplitude's fourth-quarter and full year 2025 earnings call. Today, I'm going to cover three things. First, our strong Q4 results and progress in the enterprise. Second, how AI is driving demand for analytics, and our strategy to deliver. Third, a look at our new AI agents in action and a spotlight on customer stories.

各位下午好，歡迎參加 Amplitude 2025 年第四季及全年財報電話會議。今天，我將介紹三件事。首先，我們第四季業績強勁，企業也取得了進展。其次，人工智慧如何推動對分析的需求，以及我們滿足需求的策略。第三，讓我們來看看我們新的人工智慧代理是如何運作的，並重點介紹客戶案例。

Q4 represents one of these strongest quarters in Amplitude history. Our fourth-quarter revenue was $91.4 million, up 17% year-over-year and exceeding the high end of our revenue guidance. Our annual recurring revenue was $366 million, up 17% year-over-year and up $18 million from last quarter. This was our highest net new ARR quarter since 2021.

第四季是 Amplitude 史上業績最強勁的季度之一。我們第四季營收為 9,140 萬美元，年增 17%，超過了我們營收預期的上限。我們的年度經常性收入為 3.66 億美元，年增 17%，比上一季成長 1,800 萬美元。這是我們自 2021 年以來淨新增 ARR 最高的季度。

Non-GAAP operating income was $4.2 million or 4.6% of revenue. Customers with more than $100,000 in ARR grew to 698, an increase of 18% year-over-year. Over 25 AI companies are now included in that $100,000 cohort as well. This quarter was marked by balanced execution. No single deal was over $1 million, yet we had our highest ever number of multiproduct and $100,000 ARR lands.

非GAAP營業收入為420萬美元，佔營收的4.6%。年度經常性收入超過 10 萬美元的客戶數量增至 698 家，年增 18%。目前已有超過 25 家人工智慧公司加入了這 10 萬美元的資助計畫。本季業績執行均衡。雖然沒有一筆交易超過 100 萬美元，但我們獲得了有史以來最多的多產品和 10 萬美元 ARR 的土地。

I want to talk more about AI and our strategy. Over the past year, AI coding assistance from Anthropic, OpenAI, Cursor and others have compressed development cycles dramatically. The velocity at which companies are shipping new products has accelerated. When software is this easy to build, it creates a gap between how fast teams can ship features and how fast they can learn if they are working.

我想更深入地探討人工智慧和我們的策略。過去一年，Anthropic、OpenAI、Cursor 等公司的 AI 編碼輔助功能大幅縮短了開發週期。企業推出新產品的速度已經加快。當軟體開發如此簡單時，團隊發布功能的速度和學習功能是否有效之間的速度就會出現差距。

This shifts the pressure to the right side of the product development loop that you see here, the use and learn side. Understanding how users behave, what works and what doesn't and what actions to take next becomes the bottleneck. The constraint is no longer knowing how to build, it is knowing what to build instead. This is the hardest problem in software today.

這使得壓力轉移到了產品開發循環的右側，也就是你在這裡看到的「使用與學習」環節。了解使用者行為方式、哪些方法有效哪些無效以及下一步應該採取哪些行動，成為了瓶頸所在。現在的限制不再是知道如何建造，而是知道應該建造什麼。這是當今軟體開發中最棘手的問題。

I say that because builders and their AI assistants need a system of context that combines multiple data streams. They need structured behavioral data. They need the correct retention and funnel logic, and they need the right analytical tools exposed in a way that enables AI to reason effectively. The AI then needs to be able to iterate with that system, test hypotheses, refine queries, identify root causes and recommend actions accurately and repeatedly.

我這麼說是因為建築工人和他們的 AI 助理需要一個能夠結合多個資料流的上下文系統。他們需要結構化的行為資料。他們需要正確的使用者留存和轉換漏斗邏輯，並且需要以能夠讓 AI 有效推理的方式呈現正確的分析工具。然後，人工智慧需要能夠與該系統進行迭代，測試假設，改進查詢，識別根本原因，並準確、反覆地提出建議。

This is not something that can be bidecoded over a weekend or replicated accurately with an LLM on a data warehouse. However, it is exactly what Amplitude is purpose-built to do. We have worked with thousands of companies over the past 13-years and amass the world's largest database of user behavior. Our AI can explore patterns, explain changes and guide teams on what to do next more accurately and reliably than any other system.

這不是一個週末就能完成雙向編碼的事情，也不是用 LLM 在資料倉儲上準確複製的事情。然而，這正是 Amplitude 的設計初衷。在過去的 13 年裡，我們與數千家公司合作，累積了全球最大的使用者行為資料庫。我們的人工智慧能夠比其他系統更準確、更可靠地探索模式、解釋變化並指導團隊下一步該做什麼。

Over the past six-months, our Agentic analytics platform has reached a 76% success rate on complex production-grade queries, that is seven times better than a straight text-to-SQL approach. With the new agents we launched yesterday, teams can now move from insight to action in minutes, not weeks using analytics, cohorts, experiments and messaging in one continuous agenetic workflow.

在過去的六個月裡，我們的 Agentic 分析平台在複雜的生產級查詢中達到了 76% 的成功率，比直接的文字到 SQL 方法好七倍。借助我們昨天推出的新代理，團隊現在可以在幾分鐘內（而不是幾週）將洞察轉化為行動，這得益於在一個連續的代理工作流程中使用分析、群組、實驗和訊息傳遞功能。

Through our MCP integrations with Anthropic, Figma, OpenAI, GitHub, Lovable and Slack, we are bringing behavioral intelligence to teams where they already work. Understanding user behavior now becomes as simple as asking a question in a chat window. This puts Amplitude in a unique position. The frontier labs are pushing the boundaries of AI models, and they recognize the complexity of analytics experimentation and behavioral understanding, so they turn to Amplitude.

透過我們與 Anthropic、Figma、OpenAI、GitHub、Lovable 和 Slack 的 MCP 集成，我們將行為智慧引入團隊已經工作的地方。現在，了解使用者行為就像在聊天視窗中提問一樣簡單。這使得 Amplitude 處於一個獨特的地位。前沿實驗室正在拓展人工智慧模型的邊界，他們認識到分析實驗和行為理解的複雜性，因此他們求助於 Amplitude。

As I mentioned earlier, more than 25 of the leading AI native companies, including some of the names you see here, our customers with over $100,000 in ARR with Amplitude. In addition, one of the world's largest frontier AI labs is a seven-figure customer as well. They came to us to replace a manual system built from fragmented internal tools and raw warehouse data. Using Amplitude Enterprise Analytics and Session Replay they can now understand activation, engagement, retention and monetization end to end.

正如我之前提到的，超過 25 家領先的 AI 原生公司（包括您在這裡看到的一些公司）都是我們的客戶，他們透過 Amplitude 實現了超過 10 萬美元的 ARR。此外，全球最大的前沿人工智慧實驗室之一也是我們的七位數客戶。他們來找我們，是為了替換一個由零散的內部工具和原始倉庫資料建構的手動系統。借助 Amplitude 企業分析和會話回放功能，他們現在可以全面了解激活、參與度、留存率和盈利情況。

With Amplitude MCP, they can offer those insights directly within the AI environments, their teams already use, dramatically improving the ability for them to automate development. And it's not just AI companies, companies of all sizes need a system that gives them trusted data, insights and action to successfully deploy AI in the real world.

借助 Amplitude MCP，他們可以直接在團隊已經使用的 AI 環境中提供這些見解，從而顯著提高他們自動化開發的能力。而且，不只是人工智慧公司，各種規模的公司都需要一個能夠提供可信任數據、洞察和行動方案的系統，以便在現實世界中成功部署人工智慧。

So they turn to Amplitude as well. This momentum, combined to one of our strongest quarters across gross bookings and new ARR alongside meaningful improvement in churn. Our go-to-market motion has matured. There is a tighter focus on value-based cases in the enterprise and on expanding with multiproduct deployments. We continue to consolidate the fragmented market.

所以他們也轉向了振幅。這一勢頭，加上我們在總預訂量和新增年度經常性收入方面取得的最強勁季度之一，以及客戶流失率的顯著改善。我們的市場推廣策略已經成熟。企業更加重視基於價值的案例，並致力於透過多產品部署進行擴展。我們將繼續整合分散的市場。

Platform win rates are increasing against point solutions and our newer products are gaining traction. Guides and surveys launched less than a year ago, is our fastest-growing product to date. We are also seeing a large increase in AI native usage as agents connect directly to Amplitude. Over the past few months, the total number of queries triggered by AI agents has increased dramatically.

平台相對於單一解決方案的勝率正在提高，我們的新產品也越來越受歡迎。推出不到一年的指南和調查是我們迄今為止成長最快的產品。我們也看到 AI 原生應用程式大幅增加，因為代理程式可以直接連接到 Amplitude。在過去的幾個月裡，人工智慧代理觸發的查詢總數大幅增加。

In October last year, there were almost none, and today, it is 25%. Agents also drove the vast majority of overall incremental query growth. This tells us that customers are trusting agents with analytics work. It also indicates that our platform offers the accuracy and the context needed in production environments. Taken together, this creates a powerful tailwind for Amplitude as we continue building a durable, scalable company that can unlock the next frontier and software.

去年十月幾乎沒有，而今天，這一比例為 25%。代理商也推動了絕大多數新增查詢量的成長。這說明客戶信任代理商的數據分析工作。這也表明我們的平台能夠提供生產環境所需的準確性和上下文資訊。綜上所述，這為 Amplitude 創造了強大的順風，我們將繼續打造一家持久、可擴展的公司，從而解鎖下一個前沿領域和軟體。

Over the years, we have intentionally expanded beyond core product analytics and into adjacent workflows. We have continued that work and acquired InfiniGrow, an AI-native marketing analytics start-ups that connect spends, behavior and revenue impact. InfiniGrow brings strong AI native engineering talent to Amplitude.

多年來，我們有意地將業務範圍從核心產品分析擴展到相鄰的工作流程。我們繼續推動這項工作，並收購了 InfiniGrow，這是一家人工智慧原生行銷分析新創公司，致力於將支出、行為和收入影響聯繫起來。InfiniGrow 為 Amplitude 帶來了強大的 AI 原生工程人才。

This strengthens our platform as a system of context and expands our ability to bring acquisition, activation and retention into one continuous feedback loop. Yesterday, we launched our global AI agents, specialized agents and MCP. This represents the start of a fundamental shift in how teams work with their analytics data. Historically, analytics has required humans to do most of the heavy lifting, writing queries, building dashboards, monitoring changes, interpreting results and then figuring out what to do next.

這增強了我們作為情境系統的平台，並擴展了我們將獲取、啟動和留存整合到一個連續回饋循環中的能力。昨天，我們推出了全球人工智慧代理、專業代理和 MCP。這標誌著團隊處理分析數據的方式開始發生根本轉變。從歷史上看，數據分析需要人類完成大部分繁重的工作，包括編寫查詢、建立儀表板、監控變化、解釋結果，然後確定下一步該做什麼。

That process does not scale in the world where teams are shipping faster and faster. AI agents change that model. Instead of asking questions one at a time, teams can now delegate work to agents that continuously analyze behavior, surface insights and guide action. Our agents understand events, funnels, cohorts, experiments, session replay and outcomes because they operate inside a context system specifically designed for them.

在團隊交付速度越來越快的今天，這種流程無法擴展。人工智慧代理改變了這種模式。現在，團隊不再需要逐一提問，而是可以將工作委派給代理，由代理持續分析行為、發現洞察並指導行動。我們的代理人能夠理解事件、漏斗、群體、實驗、會話回放和結果，因為它們在專門為它們設計的上下文系統中運作。

Agents make life easier by doing the work that slows teams down today. That is very, very different from bolt-on AI tools from SaaS companies that sit outside the data and try to infer meaning after the fact. The best way to see this and understand this is to look at it in action. I want to show you a quick teaser video, and then I'm going to show you a demo of what we've released. Let's go ahead and roll the video.

經紀人透過承擔目前拖慢團隊速度的工作，讓生活變得更輕鬆。這與 SaaS 公司提供的附加式 AI 工具截然不同，後者脫離數據本身，試圖在事後推斷含義。了解和理解這一點的最佳方法是觀察它的實際運作。我想先給你們看一段簡短的預告片，然後再給你們展示一下我們發布的產品演示。接下來我們開始播放影片。

(video playing)

（影片播放）

It's a great question all product builders should ask themselves now, what will you build? I want to now walk you through what we've launched in AI analytics yesterday. I'm really excited about the future, and I want to show you Global Agent. Global Agent radically changes how our customers interact with their data. Starting your day with a dashboard is dead.

這是所有產品開發者現在都應該問自己的好問題：你要開發什麼？現在我想帶大家來了解我們昨天在人工智慧分析領域推出的產品。我對未來充滿期待，我想向你們展示 Global Agent。Global Agent 從根本上改變了我們的客戶與其資料互動的方式。每天早上查看儀表板的做法已經過時了。

Take a look at this interface, no dashboard, no graphs, no charts, just a chat box and a few simple prompts if customers need help getting started. I can talk to Global Agent like I talked to a colleague. I'm going to go ahead and ask it how's our loyalty program doing? In seconds, it comes back with a summary. Notice, I didn't use any jargon about event totals or taxonomy, just a regular question.

看看這個介面，沒有儀錶板，沒有圖表，沒有圖形，只有一個聊天框和一些簡單的提示，以便在客戶需要幫助入門時提供幫助。我可以像和同事一樣與全球代理商交談。我接下來要問它：我們的會員忠誠度計畫進行得如何？幾秒鐘後，它會傳回一個摘要。請注意，我沒有使用任何關於事件總數或分類的術語，只是一個普通的問題。

It's calling out some pretty concerning numbers. Only 5% of users who view our welcome page actually go on to join the loyalty program. That is low, so I'm going to click in and investigate more. The Global Agent has followed me to a deep dive on this chart. I can keep investigating with another simple question. Break this down by traffic source. Here's the breakdown.

它公佈了一些相當令人擔憂的數據。只有 5% 的瀏覽過我們歡迎頁面的用戶最終加入了會員計劃。這個數值很低，所以我打算點擊進去進一步了解。全球代理商已跟隨我深入研究了這張圖表。我可以用另一個簡單的問題繼續調查。按流量來源細分。以下是詳細分析。

Facebook and Instagram are driving loyalty sign-ups at 5.6% and 5.2%, while Google and direct traffic lagged behind. The Global Agent summarizes it perfectly. Social media converts 10% to 15% better. Since social media outperforms Google, I might shift ad spend, but looking overall, all the rates are low. So before reallocating budget, I'm going to go deeper. Is this a channel problem or an audience problem?

Facebook 和 Instagram 分別推動了 5.6% 和 5.2% 的用戶註冊，而 Google 和直接流量則落後於它們。全球代理人對此總結得非常到位。社群媒體的轉換率比傳統管道高出 10% 到 15%。由於社群媒體的表現優於谷歌，我可能會調整廣告支出，但總體而言，所有廣告費率都很低。所以在重新分配預算之前，我要先進行更深入的研究。這是頻道問題還是觀眾問題？

Let me ask, do new users convert differently than existing customers? Without AI, this kind of analysis takes a lot of time, segmenting users, comparing funnels, pulling insights together, the Global Agent does it in seconds. And here it is, 14% conversion for repeat purchases, 5.4% overall. That's 2.6 times higher. That answers my question. It's an audience issue, not a channel issue.

請問，新用戶的轉換率與現有用戶的轉換率有何不同？如果沒有人工智慧，這種分析需要花費大量時間，例如用戶細分、漏斗比較、匯總見解等等，而全球代理商只需幾秒鐘就能完成。結果顯示，重複購買轉換率為 14%，整體轉換率為 5.4%。那可是高出2.6倍。我的疑問得到了解答。這是觀眾的問題，不是頻道的問題。

I should reallocate my budget towards repeat purchases. Again, simple language, fast answers, deep learning that anyone can use. Analytics is the perfect use case for agents. So I want to show you specialized agents. Our specialized agents work continuously on specific jobs that would usually take dozens, if not hundreds of hours, monitoring dashboards, analyzing session replays, processing feedback, running conversion experiments, legwork now done automatically.

我應該把預算重新分配到重複購買的商品。再次強調，語言簡潔，解答迅速，深度學習技術人人都能掌握。分析是經紀人的完美應用場景。所以我想向你們展示一些專業代理人。我們的專業代理商持續進行特定工作，這些工作通常需要數十甚至數百小時才能完成，例如監控儀表板、分析會話重播、處理回饋、運行轉換實驗等等，這些繁瑣的工作現在都已自動完成。

We're going to be eating our own dog food on this one. I already have a session replay agent set up to monitor our own session replay tool and have it set in addition to sending a slack when it has a strong finding. This specialized agent has been watching hundreds of replays and sent me some summarized findings.

這次我們要吃自己的狗糧了。我已經設定了一個會話回放代理來監控我們自己的會話回放工具，並且設定了當它發現重要問題時，也會向 Slack 發送通知。這位專業分析師觀看了數百場錄影回放，並向我發送了一些總結性分析結果。

Users with multiple saved filters type search terms, but cannot find filters without scrolling through the full list. Power users cannot preview filter criteria before applying, forcing trial and error selection. These are all things we should improve. We could have had someone watch all those replays. We could have talked to customers from hours on end or we could have let these continue to be issues.

使用者儲存了多個篩選條件後，輸入搜尋字詞，但必須捲動整個清單才能找到篩選條件。進階使用者無法在套用篩選條件前預覽篩選條件，只能透過重複試誤進行選擇。這些都是我們應該改進的地方。我們本來可以安排人把所有回放都看一遍。我們可以花幾個小時和客戶溝通，也可以讓這些問題持續下去。

Instead, I get these findings served to me on a daily basis with a full report and a detailed breakdown with key findings, suggestions on what to explore next and even a highlighted brief set of replays of these issues. Okay. We're going to save the best for last. Finally, I want to show you what I'm most excited about, which is Amplitude MCP.

相反，我每天都會收到這些調查結果，其中包含一份完整的報告和詳細的分析，包括關鍵發現、下一步探索方向的建議，甚至還有這些問題的簡要回顧。好的。我們要把最好的留到最後。最後，我想向大家展示我最興奮的東西，那就是 Amplitude MCP。

We're releasing a fast-growing library of expert level workflows that customers can trigger in AI clients like Claude with a simple slash command. I'm going to go ahead and use Amplitude and Claude by typing use slash create dashboard and create a dashboard that tracks our growth conversion performance, hit, I hit enter and it goes to work.

我們正在發布一個快速成長的專家級工作流程庫，客戶可以使用簡單的斜線命令在 Claude 等 AI 用戶端中觸發該工作流程。我接下來要使用 Amplitude 和 Claude，輸入 use slash create dashboard 並創建一個跟踪我們增長轉換率表現的儀錶板，點擊，我按下回車鍵，它就開始工作了。

Instead of me manually creating 15-charts, running the segmentations myself, and piecing together an explanation in a doc, this skill handles it in one-click. With MCP apps, Claude is opening and building Amplitude charts right inside itself. It's done it. So I've now gone to the link it gave me in a perfectly built dashboard with top-level metrics, conversion funnels and segment breakdowns. Amazing.

這項技能只需單擊即可完成，而無需我手動建立 15 個圖表、自己進行細分，並在文件中拼湊解釋。借助 MCP 應用程序，Claude 可以直接在應用程式內部開啟和建立振幅圖表。搞定了。所以我現在點擊了它給我的鏈接，進入了一個結構完美的儀錶板，其中包含頂級指標、轉換漏斗和細分市場分析。驚人的。

Moving on to customers. We had a great quarter for new and expansion deals with enterprise companies, including one of the largest music streaming apps, the Cheesecake Factory Asana, PGA of America, CrossFit, Stewart Title Guaranty Company, Crunch Fitness, WHOOP, Once Upon Publishing and NTT DOCOMO.

接下來談談客戶。本季我們在與企業公司的新交易和擴展交易方面取得了巨大成功，其中包括最大的音樂串流應用之一、芝士蛋糕工廠 Asana、美國職業高爾夫協會、CrossFit、Stewart Title Guaranty Company、Crunch Fitness、WHOOP、Once Upon Publishing 和 NTT DOCOMO。

I'm going to highlight three examples that demonstrate the power of the platform in different ways. Japanese telecom NTT DOCOMO is using Amplitude across more than 1,000 active users to drive efficiencies at scale. As an early design partner for our AI agents, their data platform team uses agents to streamline analysis across existing dashboards.

我將重點介紹三個例子，從不同的角度展示平台的強大功能。日本電信公司 NTT DOCOMO 正在 1000 多名活躍用戶中使用 Amplitude 來大規模提高效率。作為我們人工智慧代理的早期設計合作夥伴，他們的資料平台團隊利用代理來簡化現有儀表板上的分析。

In one project, an agent reduced campaign analysis time by over 90%. Our AI-powered session replay summaries automatically localized into Japanese help UX teams identify issues faster and improve the digital journey for millions of customers. We are now working closely with NTT DOCOMO to shape our agents road map with feedback on collaboration features and AI-powered insights.

在一個專案中，一名代理人將行銷活動分析時間縮短了 90% 以上。我們利用人工智慧技術自動將會話回放摘要在地化為日語，幫助使用者體驗團隊更快發現問題，並改善數百萬客戶的數位體驗。我們現在正與 NTT DOCOMO 密切合作，根據他們對協作功能和人工智慧驅動的洞察的回饋，制定我們的代理路線圖。

Siemens, the $70 billion global technology leader partnered with Amplitude over three-years ago to power analytics across its website presence and broader digital ecosystem. By consolidating onto our AI analytics platform from a series of point solutions, Siemens gained a unified real-time view of user behavior. Recently, the team organizing their annual user conference use Amplitude to identify their overreliance on direct e-mail and organic channels.

市值 700 億美元的全球科技領導者西門子三年前與 Amplitude 合作，為其網站和更廣泛的數位生態系統提供分析支援。透過將一系列獨立解決方案整合到我們的 AI 分析平台，西門子獲得了使用者行為的統一即時視圖。最近，負責組織年度用戶大會的團隊使用 Amplitude 來識別他們過度依賴直接電子郵件和自然管道的問題。

They experimented by reallocating spend into targeted web promos plus paid and organic social. This delivered a 90% year-over-year increase in web traffic and a projected 50% increase in registrations in attendance to their conference. Lastly, we landed one of the largest music streaming apps in the world. We are working with the teams that lead checkout optimization, upgrades, churn prevention and recovery as they seek to understand the revenue drivers for hundreds of millions of monthly active users.

他們嘗試將支出重新分配到有針對性的網路推廣以及付費和自然社交媒體推廣中。這使得網站流量年增了 90%，預計會議註冊人數將增加 50%。最後，我們成功拿下全球最大的音樂串流應用程式之一。我們正在與負責結帳優化、升級、客戶流失預防和復原的團隊合作，因為他們正在努力了解數億月活躍用戶的收入驅動因素。

They will use Amplitude analytics combined with session replay to get a holistic view on these monetization drivers. These stories all point to a common theme from AI start-ups to global enterprises, customers are betting on Amplitude as the AI analytics platform that will help them thrive in this new era. Before I hand it over to Andrew, I want to be clear on how AI is shaping our opportunity.

他們將使用 Amplitude 分析和會話回放功能，以全面了解這些獲利驅動因素。從人工智慧新創公司到全球企業，這些案例都指向一個共同的主題：客戶們都寄望於 Amplitude，將其視為能夠幫助他們在新時代蓬勃發展的 AI 分析平台。在將話題交給安德魯之前，我想先明確一下人工智慧是如何塑造我們機會的。

There is a common misconception in public markets that AI makes analytics either irrelevant or easy to replicate. The exact opposite is true. AI has made software easier to create, but creation is no longer the moat. The real advantage is how quickly a team can learn, iterate, improve and automate. Agentic analytics is the key.

公開市場普遍存在一種誤解，認為人工智慧使分析變得無關緊要或容易被複製。事實恰恰相反。人工智慧讓軟體開發變得更容易，但開發本身已不再是製勝法寶。真正的優勢在於團隊學習、迭代、改進和自動化的速度。智能體分析是關鍵。

It unlocks the bottleneck on the right side of the product development loop and enables teams to learn as fast as they ship. AI is a structural tailwind for Amplitude. It is why I believe the opportunity ahead is massive and why I'm excited about what's to come.

它突破了產品開發循環右側的瓶頸，使團隊能夠像產品發布一樣快速地學習。人工智慧是Amplitude發展的結構性利多因素。也因為如此，我相信前方的機會龐大，我對未來充滿期待。

Now over to Andrew to walk you through the financials.

現在請安德魯為大家講解財務數據。

Thank you, Spenser, and good afternoon, everyone. 2025 was a year of innovation, execution, and we delivered a solid base for our future long-term growth strategy. When we met at our Investor Day last March, we laid out a deliberate road map to capture the enterprise and accelerate multiproduct adoption, while leading the industry in innovation.

謝謝斯賓塞，大家下午好。 2025年是創新和執行的一年，我們為未來的長期成長策略奠定了堅實的基礎。去年三月，我們在投資者日上製定了周密的路線圖，旨在贏得企業客戶，加速多產品採用，同時引領產業創新。

Today's results demonstrate that we haven't just met those goals, we've established a new baseline for durable growth. The enterprise is now our core growth engine. ARR from our enterprise customer cohort is up 20% year-over-year, with higher retention and expansion rates than the rest of our business. This was not by accident or luck, our AI analytics platform has been designed to be enterprise-grade with trust and safety of our customers at the center.

今天的成果表明，我們不僅實現了這些目標，而且還為永續成長樹立了新的基準。企業現在是我們的核心成長引擎。來自企業客戶群的 ARR 年成長 20%，其客戶留存率和擴張率均高於我們其他業務。這並非偶然或運氣，我們的人工智慧分析平台的設計目標是達到企業級水平，並將客戶的信任和安全放在首位。

Our go-to-market team has worked for the past three-years to orient our go-to-market motion to focus on the enterprise, increasing customer value through selling our platform and engaging in longer-term contracts. 2025 was the coalescence of this work to focus on our customers' value and creating durable base for future growth.

過去三年，我們的市場推廣團隊致力於調整市場策略，將重點放在企業客戶身上，透過銷售我們的平台和簽訂長期合約來提升客戶價值。 2025 年標誌著這項工作的最終成果，我們將更加重視客戶價值，並為未來的成長奠定堅實的基礎。

We sustained growth of current RPO greater than 20% throughout the year. And in Q4, total RPO grew 35% year-over-year. Our average contract duration is now above 22-months. In addition to our success in the enterprise, we have also formulated our product and our go-to-market team to embrace our AI platform strategy. By combining niche point product solutions surrounding analytics into a comprehensive platform, we are able to deliver greater value than stitching together point solutions.

我們全年保持了目前RPO超過20%的成長。第四季度，RPO 總量年增 35%。我們的平均合約期限現在超過 22 個月。除了在企業層面的成功之外，我們還制定了產品策略和市場推廣團隊，以貫徹我們的人工智慧平台策略。透過將圍繞分析的細分領域產品解決方案整合到一個綜合平台中，我們能夠提供比拼湊各個獨立解決方案更大的價值。

We also believe that having a platform is essential to the harnessing capabilities of AI to reduce friction in our customers' workflows. In 2025, we did a great job expanding our multiproduct attach rate for our customers. 74% of our ARR is from customers with more than one-product, up 15 percentage points from last year. We still have a great opportunity to expand our multiproduct customers as well.

我們也認為，擁有一個平台對於利用人工智慧的能力來減少客戶工作流程中的摩擦至關重要。2025年，我們在提升客戶的多產品附加率方面取得了顯著成效。我們74%的年度經常性收入（ARR）來自購買了不只一種產品的客戶，比去年增加了15個百分點。我們仍然有很大的機會拓展我們的多產品客戶群。

Only 51% of our ARR comes from customers with greater than three products. Looking at a full platform deployment of five-plus products, that percentage is 20%, doubling year-over-year. We have a massive opportunity to expand with our customer base. We believe our market opportunity expands dramatically with the inclusion of our new AI products that promise to expand adoption and use cases.

只有 51% 的年度經常性收入來自擁有三種以上產品的客戶。如果從包含五款以上產品的完整平台部署來看，這一比例為 20%，比前一年翻了一番。我們擁有巨大的發展機遇，可以拓展我們的客戶群。我們相信，隨著我們新的人工智慧產品的推出，我們的市場機會將大幅擴大，這些產品有望擴大應用範圍和用例。

The progress in selling our platform is best exemplified through improvement of our retention and expansion motion with dollar-based net retention now above 105%, after exiting 2024 at 100%. However, our work is not done. At the beginning of this year, we introduced a new pricing and packaging strategy to our sellers. Let's start with what's not changing. We are not changing our core billing metric of events.

我們平台銷售的進展最好地體現在我們客戶留存率和擴張速度的提高上，以美元計的淨留存率現在已超過 105%，而 2024 年底時為 100%。但是，我們的工作還沒有完成。今年年初，我們向賣家推出了新的定價和包裝策略。我們先來說說哪些方面沒有改變。我們不會改變以事件為核心的計費指標。

We believe this is a great representation of the value our customers receive from our platform, and it is also an appropriate monetization strategy as we center AI engagement on our platform. What has changed is we are centralizing the monetization of our other products, such as experimentation, session replay, guides and surveys to be a percentage uplift on the core platform charge, which is events based.

我們認為這很好地體現了我們的客戶從我們的平台獲得的價值，而且由於我們將人工智慧互動作為我們平台的核心，這也是一種合適的獲利策略。改變的是，我們將其他產品（如實驗、會話回放、指南和調查）的貨幣化集中起來，作為核心平台費用的一定比例的提升，而核心平台費用是基於事件的。

This reduces the friction of adoption of those products by making it easier to understand for our customers and reduces the need to estimate how many sessions or experiments they want to run in the near term. Longer term, this will also encourage greater consumption on our platform as comes no longer fear over using certain parts of the contract or underutilizing others.

這降低了客戶採用這些產品的阻力，使產品更容易被理解，並減少了客戶在近期內估計要執行多少會話或實驗的需求。從長遠來看，這將鼓勵用戶在我們平台上進行更多消費，因為他們不再擔心使用合約的某些部分或未充分利用其他部分。

It's a radical simplification of our pricing that acknowledges our customers' needs for greater cost transparency and certainty on their costs as the volume of data ingested into our platform expands. It also supports our focus of integrating AI into all of our product offerings and expanding customer usage, which can be a tailwind longer term on easier lands and faster platform expansions.

這是我們定價方式的徹底簡化，旨在滿足客戶對成本透明度和成本確定性的需求，因為隨著我們平台接收的資料量不斷增加，客戶的需求也不斷增長。這也有助於我們專注於將人工智慧融入我們所有的產品中，並擴大客戶的使用範圍，從長遠來看，這可以為我們更容易取得進展和更快地擴展平台提供助力。

In summary, as we've transitioned to an AI analytics company, we have created a more durable base of our business focused on the enterprise. We've driven expansion of our platform through innovation, and we're making it easier for customers to get value quickly and encourage expansion. We've done all this while being disciplined in our spending and driving to non-GAAP profitability with record free cash flow.

總而言之，隨著我們轉型為人工智慧分析公司，我們為專注於企業市場的業務建立了一個更穩固的基礎。我們透過創新推動了平台的擴展，我們正在讓客戶更容易快速獲得價值並鼓勵擴展。我們在嚴格控制支出的同時，實現了非GAAP獲利，並創下了自由現金流的新紀錄。

Looking at the rule of 40, which we measure based on free cash flow yield and ARR growth, we've now improved from a rule of 15 in 2024 to over 24% in 2025. We'll continue to focus on driving top line growth through a disciplined manner in 2026. Now turning to our fourth-quarter and full year results. And as a reminder, all financial results that I will be discussing with the exception of revenue, are non-GAAP.

根據自由現金流收益率和 ARR 成長來衡量的 40 法則，我們已經從 2024 年的 15 法則提高到 2025 年的 24% 以上。2026年，我們將繼續以嚴謹的方式推動營收成長。現在來看我們第四季和全年業績。再次提醒大家，除收入外，我將要討論的所有財務表現均為非GAAP財務業績。

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. Fourth-quarter revenue was $91.4 million, up 17% year-over-year versus 9% in fiscal 2024. Fiscal year 2025 revenue was $343.2 million, up 15% year-over-year versus 8% in the fiscal year 2024.

我們的 GAAP 財務表現以及 GAAP 和非 GAAP 業績的調節表，可以在我們網站的投資者關係頁面上的獲利新聞稿和補充財務報表中找到。第四季營收為 9,140 萬美元，年增 17%，而 2024 財年同期為 9%。2025 財年營收為 3.432 億美元，比 2024 財年成長 15%，而 2024 財年成長 8%。

Total ARR increased to $366 million exiting the fourth-quarter, an increase of 17% year-over-year and $18 million sequentially. 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. Platform sales were also particularly strong. 44% of our customers now have multiple products with 74% ARR coming from that cohort.

截至第四季末，總 ARR 增至 3.66 億美元，年增 17%，季增 1,800 萬美元。以下是本季關鍵要素的更多詳情。本季企業的新交易和擴張交易都表現強勁。平台銷售也表現特別強勁。目前，44% 的客戶擁有多種產品，其中 74% 的年度經常性收入 (ARR) 來自該群體。

The number of customers representing $100,000 or more of ARR in Q4 grew to 698, an increase of 18% year-over-year and up 45 customers since the last quarter, representing the largest sequential increase in this cohort in company history. Additionally, the number of customers representing $1 million or more in ARR grew in Q4 to 56, up 33% year-over-year, demonstrating our ability to land significant accounts and grow them over time.

第四季度，年經常性收入 (ARR) 達到 10 萬美元或以上的客戶數量增長至 698 人，同比增長 18%，比上一季增加了 45 人，這是該公司歷史上該群體最大的連續增長。此外，第四季度年經常性收入 (ARR) 達到 100 萬美元或以上的客戶數量增長至 56 家，同比增長 33%，這表明我們有能力獲得重要客戶並隨著時間的推移實現增長。

In period net dollar retention progressed to 105%, and led by cross-sell expansions across our customer base. 58% of Q4 gross ARR was driven by expansions across a broad range of customers with no individual expansion exceeding $1 million. It's still driving meaningful progress in that dollar retention. We will continue to focus on driving net dollar retention higher through our platform strategy.

本期淨美元留存率提升至105%，主要得益於客戶群交叉銷售的拓展。第四季總年度經常性收入的58%來自各類客戶的拓展，但單筆拓展金額均未超過100萬美元。它仍然在提高美元留存率方面取得顯著進展。我們將繼續專注於透過我們的平台策略來提高淨美元留存率。

Gross margin was 77% for the fourth-quarter, flat to fourth-quarter of 2024 and up 1 point since last quarter. We continue to make progress on optimizing our hosting, driving multiproduct contracts and monetizing our services engagements. We will continue to look for opportunities to incrementally improve gross margin over time. Sales and marketing expenses were 42% of revenue, a decrease of 1 point from the third-quarter.

第四季毛利率為 77%，與 2024 年第四季持平，較上一季成長 1 個百分點。我們在優化主機代管、推動多產品合約以及實現服務業務獲利方面持續取得進展。我們將繼續尋找機會，逐步提高毛利率。銷售和行銷費用佔收入的 42%，比第三季下降了 1 個百分點。

We continue to focus on improving sales efficiencies, driving improvements through our changes in processes, coverage and expansion of enterprise customers. At the same time, we are investing in future growth while balancing those incremental investments with efficiency gains. In Q1 FY26, we will have higher sales and marketing expenses as a percentage of revenue, reflecting timing of events and our annual company kickoff.

我們將繼續專注於提高銷售效率，透過流程改善、擴大覆蓋範圍和拓展企業客戶來推動提升。同時，我們正在投資未來成長，同時努力平衡這些增量投資與效率提升之間的關係。2026 財年第一季度，我們的銷售和行銷費用佔收入的比例將會更高，這反映了活動的安排以及我們公司一年一度的啟動儀式。

R&D was 18% of revenue, flat to the fourth-quarter of 2024. We expect to continue to invest in the talent and capabilities of our team to drive greater innovation in the future. G&A was 12% of revenue, down 4 points for the fourth-quarter of 2024. We expect G&A to improve as a percentage of revenue over time. Total operating expenses were $66 million, 72% of revenue, down 3 points sequentially.

研發支出佔營收的 18%，與 2024 年第四季持平。我們將繼續投資團隊的人才和能力建設，以推動未來更大的創新。2024 年第四季，一般及行政費用佔營收的 12%，下降 4 個百分點。我們預期管理費用佔收入的比例會隨著時間的推移而改善。總營運支出為 6,600 萬美元，佔營收的 72%，較上季下降 3 個百分點。

Operating income was $4.2 million or 4.6% of revenue. Net income per share was $0.04 based on 141.5 million diluted shares compared to net income per share of $0.02 with 135.7 million diluted shares a year ago. Free cash flow in the quarter was $11.2 million or 12% of revenue compared to $1.5 million or 2% of revenue during the same period last year.

營業收入為 420 萬美元，佔總收入的 4.6%。根據 1.415 億股稀釋股份計算，每股淨收益為 0.04 美元，而去年同期，根據 1.357 億股稀釋股份計算，每股淨收益為 0.02 美元。本季自由現金流為 1,120 萬美元，佔營收的 12%，而去年同期為 150 萬美元，佔營收的 2%。

In the fourth-quarter, we managed our cash collections and made meaningful progress on shifting contracts with annual payments in advance. For the full year, we had a record free cash flow of nearly $24 million or free cash flow margin of 7%. 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.

第四季度，我們管理了現金回收，並在調整預付年款的合約方面取得了實質進展。全年來看，我們的自由現金流創下近 2,400 萬美元的歷史新高，自由現金流利潤率為 7%。我們對平台的長期價值充滿信心，並已使用並將繼續使用現金來最大限度地減少股權稀釋的影響。

We have already purchased in the open market under our current buyback. And given the strength in our balance sheet and the underlying business, our Board has approved an additional reserve of $100 million to be used for buybacks. 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.

我們已經根據目前的回購計劃在公開市場上進行了採購。鑑於我們資產負債表的穩健性和業務基礎，董事會已批准追加 1 億美元的儲備金用於股票回購。我們的資產負債表狀況依然強勁，這使我們有機會在適當的時候採取更積極的併購策略，以加快我們的研發路線圖。

Now turning to our outlook. As a reminder, the philosophy of how we set guidance is through the lens of execution. We are confident we have the right strategy and the right platform to continue to consolidate the fragmented market. We continue to improve our go-to-market motion and are accelerating our pace of innovation.

現在談談我們的展望。再次提醒大家，我們制定指導方針的理念是透過執行的角度出發。我們有信心，我們擁有正確的策略和平台，能夠繼續整合分散的市場。我們不斷改進產品上市策略，並加快創新步伐。

We have the right monetization strategy to encourage the adoption of our AI tools, and we believe those tools will reduce the barrier to adoption of our full platform, leading to greater monetization opportunities. Our strategy remains consistent with our go-to-market is being aided by our simplification of our pricing and packaging.

我們擁有正確的獲利策略來鼓勵用戶採用我們的人工智慧工具，我們相信這些工具將降低用戶採用我們完整平台的門檻，從而帶來更大的獲利機會。我們的策略與市場推廣策略保持一致，而簡化定價和包裝也為此提供了幫助。

We will continue to focus on gaining new enterprise customers and driving cross-platform sales with our existing customer base. We also believe that with the release of our AI capabilities, our monetization of data ingested in our platform and the cross-sell opportunities of new products gives us the right strategy to align the value of our customers receive with our growth opportunities and grow our business in a profitable way.

我們將繼續專注於獲取新的企業客戶，並利用現有客戶群推動跨平台銷售。我們也相信，隨著人工智慧能力的發布，我們平台所攝取的數據貨幣化以及新產品的交叉銷售機會，為我們提供了正確的策略，使客戶獲得的價值與我們的成長機會保持一致，並以盈利的方式發展我們的業務。

For the first-quarter of 2026, we expect revenue to be between $91.7 million and $93.7 million, representing an annual growth rate of 16% at the midpoint. We expect non-GAAP operating income to be between negative $4.5 million and negative $2.5 million. And we expect non-GAAP net income per share to be between a negative $0.02 and a negative $0.01 assuming basic weighted average shares outstanding of approximately 135.1 million.

我們預計 2026 年第一季的營收將在 9,170 萬美元至 9,370 萬美元之間，中位數對應的年增率為 16%。我們預計非GAAP營業收入將在-450萬美元至-250萬美元之間。我們預計，假設基本加權平均流通股約為 1.351 億股，則非 GAAP 每股淨收益將在 -0.02 美元至 -0.01 美元之間。

For the full year of 2026, we expect full year revenue to be between $390 million and $398 million, an annual growth rate of 15% at the midpoint. We expect our full year non-GAAP operating income to be between $7 million and $13 million. We expect non-GAAP net income per share to be between $0.08 and $0.13, assuming weighted average shares outstanding of approximately 145.9 million as measured on a fully diluted basis.

我們預計 2026 年全年營收將在 3.9 億美元至 3.98 億美元之間，年增長率中位數為 15%。我們預計全年非GAAP營業收入將介於700萬美元至1,300萬美元之間。我們預計，以完全稀釋基準計算的加權平均流通股約為 1.459 億股，非 GAAP 每股淨收益將在 0.08 美元至 0.13 美元之間。

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 a Agentic analytics to the world.

最後，我們正在加快創新步伐，並不斷提升我們能夠為客戶創造的價值。我們有信心打造一個可持續發展且不斷成長的業務，同時將智慧分析帶給全世界。

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

接下來，我們將進入問答環節。接下來該你了，約翰。

(Operator Instructions) Taylor McGinnis, UBS.

（操作說明）泰勒·麥金尼斯，瑞銀集團。

Maybe just first on -- you announced a number of exciting agent offerings this week. And at the same time, you've also seen good traction with third-party agents connecting into Amplitude's platform. So -- and then Spenser, you showed a really good example of being able to extract insight using Anthropic Claude. So I guess how do you see Amplitude's agents and these third-party agents evolving?

或許應該先說——你們本周宣布了一系列令人興奮的經紀人優惠活動。同時，我們也看到第三方代理商接入 Amplitude 平台取得了良好的進展。所以——然後斯賓塞，你展示了一個非常好的例子，說明如何利用人擇原理提取洞見。所以我想知道您如何看待 Amplitude 的代理和這些第三方代理的發展？

Maybe you just talk about the differentiation that you anticipate with Amplitude's agents versus what's being done with the third-party agents today.

或許你可以談談你預期 Amplitude 的代理與目前第三方代理的不同之處。

Yes. So to be clear, they both use the same underlying infrastructure. What will happen with either MCP model context protocol is a way for external products like Claude or OpenAIs, Chat GPT or Cursor to connect into Amplitude and request a set of calls. But that is the same infrastructure that both our global agents and our specialized agents use.

是的。所以說，它們都使用相同的底層基礎設施。無論採用哪種 MCP 模型上下文協議，都會為 Claude 或 OpenAIs、Chat GPT 或 Cursor 等外部產品提供一種連接到 Amplitude 並請求一組呼叫的方法。但這是我們全球代理商和專業代理商使用的相同基礎設施。

And so the way to think about it is there's a whole set of tool calls that are available to these agents. You can say, get me a list of events, get me a retention, get me the list of tools you have like retention and funnels, get me the possible properties for this event. And what we'll do is we'll expose that to an orchestrator that we have that basically interprets a query, whether it's in the chat with Global Agent or whether it's external from MCP.

所以，我們可以這樣理解：這些代理可以使用一整套工具來呼叫。你可以說，給我一份事件列表，給我一份留存率數據，給我一份你擁有的工具列表，例如留存工具和轉換漏斗，給我一份該事件的可能屬性列表。我們將把這個資訊暴露給我們的一個協調器，它基本上可以解釋查詢，無論查詢是在與 Global Agent 的聊天中，還是在 MCP 外部。

And then it'll kind of pull in all the different contexts I talked about and then spit out the answers that you see. So it's the same underlying infrastructure because the nature and type of questions are the same whether you're asking it from Claude or Slack or whether you're in Amplitudes UI.

然後它會把我剛才提到的所有不同背景都考慮進去，然後輸出你看到的答案。所以，無論你是在 Claude 或 Slack 中提問，還是在 Amplitudes UI 中提問，問題的性質和類型都是相同的，因此底層基礎設施是一樣的。

Perfect. Awesome. And then Andrew, maybe just one follow-up for you. If we look at the 4Q numbers, it looked like the upside in the quarter was a little bit lighter than what we've seen in the past. Now ARR continued to accelerate. So was that just a function of the quarter being back-end loaded? Or anything to flag in terms of the quarter may be in any areas coming a little bit below as expected?

完美的。驚人的。安德魯，或許我還有一個後續問題想問你。如果我們看一下第四季的數據，就會發現本季的成長幅度似乎比我們過去看到的要小一些。現在，年度經常性收入持續加速成長。所以這僅僅是因為季度末期資料集中在後半段造成的嗎？或者說，本季有哪些方面需要特別注意，例如哪些領域的表現可能略低於預期？

So first, I'd say, Q4 was a great quarter for new logo ARR. We had a lot of new customers that are getting value from Amplitude and they're starting their journey with us. Those tend to be ones in which you're working throughout the quarter, and there was a large proportion of ARR that was booked later than we've seen in prior quarters.

首先，我認為第四季對於新 logo 的 ARR 來說是一個非常棒的季度。我們有很多新客戶，他們從 Amplitude 獲得了價值，並且正在與我們一起開啟他們的旅程。這些專案往往需要你在整個季度內持續推進，而且與以往季度相比，本季 ARR 的確認時間也相對較晚。

And as we mentioned before, we didn't see a lot of really big expansion during the quarter. So it was one of those areas where you're building a lot of opportunity for future growth with these new customers. And it's always one you're competing for -- when you're going into a new logo, you have to really compete and show value and sometimes those take a little longer as well.

正如我們之前提到的，本季並沒有看到很多真正的大規模擴張。所以，在這個領域，你透過這些新客戶創造了很多未來發展的機會。而且，設計新標誌時，你總是要參與競爭，展現價值，而這有時需要更長的時間。

But we're really pleased with all the new customers that have become Amplitude customers, and we think that, that sets us up very well for expansions in the future.

但我們對所有成為 Amplitude 客戶的新客戶感到非常高興，我們認為，這為我們未來的擴張奠定了良好的基礎。

Billy Fitzsimmons, Piper Sandler.

比利·菲茨西蒙斯，派珀·桑德勒。

I guess maybe to start, can you help us think through the NRR improvement? And how much would you contribute or attribute, I should say, to greater upsells and cross-sells versus maybe better -- more success in kind of mitigating some of the churn in the business?

我想首先，您能幫我們一起思考如何提高淨收益率（NRR）嗎？那麼，您認為在多大程度上，向上銷售和交叉銷售的增加，與在降低業務流失率方面取得的更大成功相比，應該如何看待呢？

Sure. So throughout the year, we've seen our customers increasingly adopting more and more of our applications in the platform. When we started off 2025, we specifically were training our sales team how to sell our platform when we're introducing new capabilities. We acquired new capabilities, and we put those into our platform as well.

當然。因此，在過去一年中，我們看到我們的客戶越來越多地在平台上採用我們的應用程式。2025 年伊始，我們專門培訓銷售團隊，在推出新功能時如何銷售我們的平台。我們獲得了新的功能，並將這些功能也整合到了我們的平台中。

So predominantly throughout 2025, the improvement in net dollar retention was related to our cross-sell capabilities. And as you were alluding to, in the past, we had situations where we were overselling capacity against analytics. And even with some customers increasing data, it wasn't enough really to offset and contribute materially towards net dollar retention.

因此，在 2025 年，淨美元留存率的提高主要與我們的交叉銷售能力有關。正如您所提到的，過去我們曾出現過過度銷售分析能力的情況。即使一些客戶增加了數據量，但這仍然不足以抵消淨美元留存率的損失，也無法對淨美元留存率做出實質貢獻。

Now that we're past most of those capacity-related issues that we created for ourselves. We're starting to see customers and their data ingested into our platform contribute towards net dollar retention improvements as well. And so now as we think forward, and I've said in the past that we have full intention to continue to set up our customers and expand with our customers, introducing new innovation.

現在我們已經克服了大部分我們自己造成的產能相關問題。我們開始看到，客戶及其資料被導入我們的平台，這也有助於提高淨留存率。所以現在，當我們展望未來時，正如我過去所說，我們完全有意願繼續為我們的客戶提供支持，並與客戶一起發展壯大，引入新的創新。

We think that both vectors, both data ingested into the platform, meaning upsells as well as cross-sells will contribute to further improvements.

我們認為，這兩個管道，即平台吸收的兩種數據（追加銷售和交叉銷售），都將有助於進一步改進。

Makes sense. And I guess on that note, if I could sneak in one more. Can you give us a sense of how the role volume upsells will play in the FY26 growth algorithm, especially as you start lapping some of the contract rightsizing from the first half of last year.

有道理。我想說，如果可以的話，我再補充一點。您能否簡要說明一下，在 2026 財年成長演算法中，銷售追加銷售將扮演怎樣的角色，尤其是在您開始消化去年上半年的一些合約調整措施之後？

So one of the things we talked about in the call was introducing our new pricing and packaging that is aligning not only to our enterprise motion but also towards the implementation of our new AI products. And in the past, I would say there were certain times where customers felt very leery about the amount they'd have to pay based on increasing data rates that we're ingesting in the platform.

因此，我們在電話會議中討論的內容之一是推出我們新的定價和包裝方案，這不僅符合我們的企業發展策略，也符合我們新的人工智慧產品的實施目標。過去，有時客戶會對我們平台不斷上漲的數據費率而產生的費用感到非常擔憂。

Meaning that those rates were so high that they wouldn't be able to see the benefits associated with marginal incremental reductions in the cost of that data. Well, our new pricing and packaging structure rewards our customers now for adding more and more data into the platform so that they're paying marginally incrementally less.

這意味著這些費率太高，以至於他們看不到數據成本哪怕只是略微降低所帶來的好處。嗯，我們新的定價和套餐結構現在會獎勵我們的客戶，鼓勵他們為平台添加越來越多的數據，這樣他們支付的費用就會略微減少。

Now that doesn't mean that it's not going to contribute growth to Amplitude as our customers are getting greater value by ingesting more data in the platform. We believe it's fair for us to have some of that fair exchange of value. But if you were going to ask where we are really focused on driving NRR and where that -- the biggest benefit, it will be continued to show from those cross-sell opportunities, that expansion of our products because we want our customers to not fear adding more data.

但這並不意味著它不會為 Amplitude 的成長做出貢獻，因為我們的客戶透過在平台上攝取更多資料而獲得更大的價值。我們認為，我們也有權利獲得一部分公平的價值交換。但如果你問我們真正關注的是如何推動 NRR 成長，以及最大的好處在哪裡，那就是繼續從交叉銷售機會和產品擴展中體現出來，因為我們希望客戶不要害怕添加更多數據。

We want them to take advantage of implementing more data into our platform, and we want that to scale, especially as they look at longer-term contracts with us.

我們希望他們能夠利用我們平台引入更多數據，並且我們希望這種模式能夠擴展，尤其是在他們考慮與我們簽訂長期合約的情況下。

Rob Oliver, RW Baird.

羅布·奧利弗，RW·貝爾德。

A follow-up there, Andrew, on the pricing and packaging question. So obviously, enterprises really like predictability. You guys have never been a seat-based model. So if you can just help us understand in the context of the new pricing model, clearly, it sounds like it's driving more engagement, a cross-sell opportunity, less of a friction experience. But how does the buyer manage that predictability? And I guess the inverse of that would be, how do you get comfortable on the cost side with AI embedded in?

安德魯，關於定價和包裝問題，我還有後續要問。顯然，企業非常喜歡可預測性。你們從來就不是以座位為基礎的模式。所以，如果您能幫助我們理解一下，在新定價模式的背景下，很明顯，它似乎能夠提高用戶參與度，帶來交叉銷售機會，減少用戶體驗中的摩擦。但買家該如何應對這種可預測性呢？我想反過來想，那就是，如何在成本方面接受人工智慧嵌入的事實？

Yes. It's a great question. We spend a lot of time working with our sales team and our customers and showing how, one, the instrumentation with the platform can have -- give the great visibility into the data they're ingesting within it. And we work with our sellers to help them better understand as the marginal incremental data into the platform, grows, how that then translates into the cost that we're going to be charging to our customer.

是的。這是一個很好的問題。我們花了很多時間與銷售團隊和客戶一起工作，並展示如何利用該平台進行儀器化，讓他們能夠更好地了解他們正在從中攝取的數據。我們與賣家合作，幫助他們更了解隨著平台上的邊際增量資料成長，這將如何轉化為我們將向客戶收取的費用。

We're encouraging to have that conversation as part of the sales process. It's a kind of a gentler way of showing and working with the customer on how they are going to adopt Amplitude over a period of time rather than guessing what their data implementation of the platform is going to be. We're working with them very closely on it and showing how the instrumentation works.

我們鼓勵在銷售過程中進行這樣的對話。這是一種更溫和的方式，向客戶展示並指導他們如何在一段時間內採用 Amplitude，而不是猜測他們將如何實現平台的資料部署。我們正與他們密切合作，並向他們展示儀器的工作原理。

Now the piece that I think is really important, and you touched on it, but I think it's -- we did a lot of work with customers to understand whether we had the right billing metric, whether it's something that they aligned to the value proposition. And we've been testing for quite a while. In fact, nearly 20% of new ARR that we booked in the quarter was actually using our new pricing and packaging in a pilot stage.

現在，我認為真正重要的一點，你也提到了，那就是——我們與客戶做了很多工作，以了解我們是否擁有正確的計費指標，以及它是否與他們的價值主張相符。我們已經測試了相當長一段時間。事實上，本季我們新增的 ARR 中，近 20% 是在我們試點階段使用了新的定價和包裝方案。

So we already know that customers like this. We already know that customers look at it as more transparent. They look at it as less friction as you were saying, we also believe it positions us very, very well given that our focus on implementing AI products into our platform is, one, it's reducing the barriers to adoption. Meaning that customers walk away thinking they're getting great value of what they've already invested in Amplitude and are less fearful knowing that they have greater cost predictability and transparency and how that usage is going to trend over a period of time.

所以我們已經知道顧客喜歡這樣。我們已經知道，客戶認為它更透明。正如你所說，他們認為這可以減少摩擦，我們也相信，鑑於我們專注於將人工智慧產品融入我們的平台，這讓我們處於非常有利的地位，首先，這降低了採用的門檻。這意味著客戶離開時會覺得他們已經為 Amplitude 投入了巨大的價值，並且由於他們擁有更高的成本可預測性和透明度，以及一段時間內的使用趨勢，因此不再那麼擔心。

Great. Super helpful. And then Spenser, 1 quick one for you. Just on InfiniGrow, you guys were very early to the AI acquisitions among our coverage, I think been very aggressive on it. And in particular, it looks like to us like this gives you guys a further opportunity to sort of go for that consolidation play that you guys have talked about.

偉大的。非常有用。史賓塞，給你快速提問一個問題。就 InfiniGrow 而言，在我們報道的範圍內，你們很早就開始關注人工智慧領域的收購，我認為你們在這方面非常積極主動。尤其值得一提的是，在我們看來，這似乎給了你們進一步的機會，去實施你們之前討論過的整合策略。

But if you could help us maybe understand what, in particular, what area or what response to what customers need InfiniGrow is going to help address and how that might accelerate that platform opportunity.

但是，如果您能幫助我們了解 InfiniGrow 將在哪個領域或針對客戶需求做出哪些回應，以及這將如何加速該平台的發展機遇，那就太好了。

There were two big things that stood out to us on the InfiniGrow team. I mean, so first, we're just always looking for great talent out there. And so when the right company and the right opportunity comes along and they are aligned with our vision and excited about it, we're going to act. With InfiniGrow in particular, there were two big things that stood about the team.

InfiniGrow 團隊有兩點讓我們印象深刻。我的意思是，首先，我們一直在尋找優秀的人才。因此，當合適的公司和合適的機會出現，並且與我們的願景一致，並對此感到興奮時，我們就會採取行動。就 InfiniGrow 而言，該團隊有兩個突出的特點。

So Daniel, the CEO there as well as the rest of the group, they've been in it on AI analytics and automating workflows for the last few years and have a ton of perspective on how the future of the category is going to be shaped. And I mean we're in uncharted territory. Like we're inventing something new here, AI analytics. And so whenever you get a chance to partner with someone else who's thought about that so deeply, it's a huge deal, and we're going to -- yes, we want to figure out how we can set up a way to work with them.

所以，Daniel（該公司執行長）以及團隊其他成員，在過去幾年裡一直致力於人工智慧分析和工作流程自動化，並且對該領域未來的發展方向有著非常深刻的見解。我的意思是，我們正處於未知領域。就好像我們正在發明一種新的東西——人工智慧分析。所以，每當有機會與對此思考如此深刻的人合作時，這都是一件大事，我們將會——是的，我們想弄清楚如何與他們合作。

So that's the first one that really stood out about InfiniGrow. The other piece that stood out is they have a lot of familiarity with analysts more on the marketing side versus product management. And particularly as those personas merge over long term and more customers from legacy MarTech tools want to come off and use something bleeding edge like an Amplitude, we want to make sure that we're ready to meet them and serve all their needs and help with that transition.

所以，這是InfiniGrow最讓我印象深刻的第一點。另一個突出的特點是，他們對行銷方面的分析師比對產品管理的分析師更熟悉。尤其是隨著這些使用者畫像在長期內逐漸融合，越來越多的傳統行銷技術工具使用者想要轉而使用像 Amplitude 這樣的前沿產品，我們希望確保我們能夠滿足他們的需求，幫助他們完成過渡。

And again, they know a lot of those buyers better than almost any other company that we've seen in the analytics space out there.

而且，他們對許多買家的了解程度，幾乎超過了我們所見過的分析領域的任何其他公司。

Clark Wright, DA Davidson.

克拉克·賴特，DA戴維森。

You noted the cross-selling opportunities continue to be an area of strength, what is the natural pathway you're seeing in terms of product adoption? And what is the role that agents are going to play going forward to help drive additional cross-selling motions?

您提到交叉銷售機會仍然是一個優勢領域，您認為在產品採用方面，自然的發展路徑是什麼？那麼，經紀人在未來將扮演怎樣的角色來推動更多的交叉銷售呢？

I mean it's great on both fronts. So analytics is the core. We're an analytics platform, something we've been very consistent about. You want to be able to track the core -- the base -- the foundation of the user journey. And that makes every single other part of the platform more valuable. So it makes it easier to do experiments because you can target users as well as measure those more effectively.

我的意思是，它在這兩方面都很棒。所以，數據分析是核心。我們是一個分析平台，我們一直堅持做這件事。你想追蹤使用者旅程的核心——基礎——根基。這使得平台的其他所有部分都更有價值。這樣一來，進行實驗就更容易了，因為你可以更有效地鎖定用戶並衡量這些用戶。

It makes it better do session replay because you can understand, hey, for a group of users that ran into this error, let me see what they did by looking at the session replays. It makes guides and surveys better because you can target guides to specific users based on if you see them confused. So analytics is the core and all of these -- they become more valuable with analytics and vice versa.

這樣可以更好地進行會話重放，因為你可以了解，嘿，對於遇到此錯誤的一組用戶，讓我透過查看會話重播來了解他們做了什麼。它能改進指南和調查，因為你可以根據使用者是否感到困惑來針對特定使用者提供指南。所以分析是核心，所有這些——它們都會因為分析而變得更有價值，反之亦然。

In terms of agents, I think the big opportunity there, and I just showed the session replay one is that these other products are -- while we launched AI analytics yesterday, and that was the main focus. These other products are actually capable of being leveraged by AI. So the session replay specialized agent demo that I shared earlier is a great example where you can watch 1, 2, 3, maybe 10 session replays, but watch 100.

就代理商而言，我認為最大的機會在於，我剛才展示的會話回放就是其他產品——雖然我們昨天推出了 AI 分析，而且那是主要關注點。這些其他產品實際上可以被人工智慧所利用。因此，我之前分享的會話回放專用代理演示就是一個很好的例子，您可以觀看 1、2、3、也許 10 個會話回放，但實際上您可以觀看 100 個。

I'll take you a few hours to get through them. And so to have an agent speed up that analysis, still get all the valuable data, summarize it up and kind of put it back to you. I mean that's you're talking a 100x fold increase in productivity versus what you might other do. Experiment is the same thing. I mean one of the things that people ask us a lot is like, cool, do you have a library of best practices for what sort of web pages or what sort of interactions work and what don't?

我需要幾個小時才能看完。因此，讓代理人加快分析速度，獲取所有有價值的數據，進行總結，然後將其回饋給你。我的意思是，與你可能採取的其他方式相比，生產力將提高 100 倍。實驗也是如此。我的意思是，人們經常問我們的一個問題是，你們是否有一個最佳實踐庫，說明哪些類型的網頁或互動有效，哪些無效？

And our experiment conversion agent will actually suggest those based on best practices of what we know from all the companies that we work with. And so it makes experimentation a lot more powerful, too. So -- and then the really cool moment is when these tie together. So you can start out in analytics and say, okay, cool, give me my unhappiest users and suggest ideas for that I could do to improve them?

我們的實驗轉化代理實際上會根據我們從所有合作公司了解到的最佳實踐來提出這些建議。因此，這也使得實驗變得更有效。所以——然後真正精彩的時刻就是當這些要素連結在一起的時候。所以你可以從數據分析入手，然後說，好的，很棒，請告訴我哪些用戶最不滿意，並提出一些我可以改進他們的想法？

And then it says, wow, okay, all of these users, they're unhappy because they ran into a page that wasn't working, and then you could have a session replay agent come in and say, okay, well, let's look at what was it on that page. It's like, oh, okay, hey, this button isn't formatted correctly and wasn't labeled and so that's probably confusing to users.

然後它會說，哇，好的，所有這些用戶都不高興，因為他們遇到了一個無法正常工作的頁面，然後你可以讓會話重播代理進來，說，好的，那麼讓我們看看那個頁面上有什麼問題。這就好比，哦，好吧，這個按鈕的格式不正確，也沒有標籤，這可能會讓用戶感到困惑。

And then you can -- and then go even further and say, okay, great, let's run. Can you propose a variant, an experiment variant to me that would actually fix it? And then it will propose it and propose another web page and you can run the test. And so you can not know anything about analytics, not know anything about your data taxonomy, not know anything about how to use session replay, not know anything how to do experimentation AB testing and do all the work of all of those projects -- products from the Global Agents or specialized agents interface. So I just -- it's going to be a massive unlock in terms of the usage.

然後你就可以——然後更進一步說，好的，太好了，讓我們開始吧。您能否向我提出一個能夠真正解決這個問題的方案，一個實驗性的方案？然後它會提出建議，並建議另一個網頁，您可以執行測試。因此，如果你對分析一無所知，對資料分類一無所知，對如何使用會話回放一無所知，對如何進行實驗 AB 測試一無所知，卻要完成所有這些專案的所有工作——全球代理商或專業代理商介面的產品。所以我覺得——就使用方面而言，這將是一次巨大的飛躍。

We're obviously most focused on analytics right now, but I'm really excited about some of the other things and a lot so it's funny. We already got some comments on Twitter that are like, hey, why does it only watch 100 sessions at a time? Why can't you watch 1,000 or 10,000 like, all right, we're working on it. We're working on it. So.

我們現在顯然最關注的是數據分析，但我對其他一些事情也感到非常興奮，這很有趣。我們已經在 Twitter 上收到了一些評論，例如，為什麼它一次只能觀看 100 個會話？為什麼你不能像這樣觀看 1,000 或 10,000 部影片呢？好吧，我們正在努力。我們正在努力。所以。

Appreciate that. And then, Andrew, there's a reference to increasing win rates versus point solutions. Is that an output of the go-to-market changes as well as the pricing and packaging updates? Or are there any other factors that is helping to drive improvements in that metric?

謝謝。然後，安德魯，這裡提到了提高勝率與點解決方案之間的關係。這是市場策略調整以及定價和包裝更新的結果嗎？或者還有其他因素有助於推動該指標的改善嗎？

I'd say the pricing and packaging is relatively new. So I wouldn't attribute that necessarily increasing win rates. I think that the biggest thing is, one, our sales team has just worked really hard at demonstrating value of our platform to our clients, and that's really resonating. And the other is you really have to credit our product team for creating just really great products that work well together.

我認為定價和包裝都是相對較新的。所以我不認為這必然會導致勝率上升。我認為最重要的是，我們的銷售團隊一直在努力向客戶展示我們平台的價值，而這種努力真的引起了客戶的共鳴。另一方面，我們真的應該感謝我們的產品團隊，他們創造了許多非常棒的產品，而且這些產品還能很好地協同工作。

A lot of people claim they have a platform, but the reality is it's a bunch of products that's stitched together, doesn't look really well. When you have a platform, you have workflows that are instrumented well and it's easy to interact with the different modules in the product. And that's the way I would characterize our platform today.

很多人聲稱自己擁有一個平台，但實際上它只是一堆產品拼湊在一起，看起來並不美觀。擁有一個平台，就能擁有完善的工作流程，並且可以輕鬆地與產品中的不同模組互動。這就是我對我們當今平台的描述。

And every time that customers are adopting more than one product, it's because they're -- that integration, those workflows seamlessly across our platform are coming through as real value. I mean I talked to a number of customers myself with the sales team, and they always come back and say, we're just so far ahead of what everybody else is even representing an analytics platform to be.

每當客戶採用不只一種產品時，都是因為他們——這種集成，這種跨我們平台無縫銜接的工作流程，正在轉化為真正的價值。我的意思是，我親自和銷售團隊與許多客戶交談過，他們總是回饋說，我們遠遠領先其他所有人對分析平台的認知。

On that, like if I just go through the last 30 buyers I've talked to in the last month, all they want to do is be educated about analytics. And sorry, how AI is going to transform analytics and the whole platform. And they see it coming. They see tons of automation ahead and they're like, hey, teach me how I can be relevant. And so when we can offer that to them by, one, providing a view on how the future unfolds.

就這一點而言，如果我回顧上個月我接觸過的 30 位買家，他們唯一想做的就是學習分析的知識。抱歉，人工智慧將如何改變分析和整個平台。他們預見了這一天的到來。他們看到了未來大量的自動化趨勢，然後他們會說，嘿，教我怎麼能保持競爭力。因此，當我們能夠透過以下方式為他們提供未來發展方向的觀點時，我們就可以這樣做。

And then two, offering them the products, tools and services that actually enable them to be successful and relevant, they want to spend a ton of time with us. And so the competitive question, especially against the smaller point solutions is kind of going away. It really is just is now the right time and can you help me get to this future fast enough and teach me.

其次，如果我們提供他們真正能幫助他們成功和保持競爭力的產品、工具和服務，他們就願意花很多時間與我們在一起。因此，競爭問題，尤其是與規模較小的點解決方案之間的競爭問題，正在逐漸消失。現在真的正是時候了，你能幫我盡快實現這個未來並教我嗎？

George McGreehan, Bank of America.

喬治·麥格里漢，美國銀行。

This is George McGreehan on for Koji. Taking a big step back, one for me on kind of the big picture. Where can we expect Agentic queries to grow to become in the mix from 25% today, maybe over the next 12- to 24-months?

這裡是喬治‧麥格里漢，為您報道小次郎。退後一步，從宏觀角度來看。我們預期 Agentic 查詢的佔比會從目前的 25% 成長到什麼程度？也許在未來 12 到 24 個月內會達到這個比例？

Yes. I mean I -- none of us have a full crystal ball, but my expectation is the vast majority are going to be done Agentically, where you're just going to have agents that run over your data all the time. They're looking at dashboards. They're looking at KPIs, they're trying to find underlying root causes of why things are changing. They're creating suggestions for your product. They're reviewing session replays.

是的。我的意思是，我們誰也沒有水晶球，但我預期絕大多數工作都會透過代理商完成，也就是讓代理商持續處理你的資料。他們正在查看儀錶板。他們正在研究關鍵績效指標，試圖找出事物改變的根本原因。他們正在為你的產品提出建議。他們正在回看比賽錄影。

They're constantly trying out and tweaking new experiments. So I mean, yes, like I don't want to put a number out there, but I think the vast majority. I think what we're seeing generally is if you look at query growth from direct usage of the Amplitude dashboards, it's increasing in line roughly with the size of our business. If you look at Agentic query use, it's skyrocketing, as you saw on that chart in the last few months.

他們不斷嘗試和調整新的實驗方案。所以我的意思是，是的，我不想給出一個具體的數字，但我認為絕大多數人是這樣。我認為我們總體上看到的是，如果你觀察直接使用 Amplitude 儀表板的查詢成長情況，你會發現它的成長與我們業務的規模大致同步。如果你看一下 Agentic 查詢的使用情況，你會發現它在過去幾個月裡呈爆炸式增長，正如你從圖表中看到的那樣。

And so the amount of leverage, I think the same thing happened to coding in the last two-years, where if you look at it, the best software engineering teams, the majority is produced by the majority of lines of codes are produced by agents. It's mostly humans editing, interpreting them, stitching them together. And kind of giving high-level direction and that's where the best software engineers are.

因此，槓桿作用的程度，我認為過去兩年編碼領域也發生了同樣的情況，如果你觀察一下，你會發現最好的軟體工程團隊，大部分程式碼行都是由代理人編寫的。大部分都是由人們進行編輯、解讀和拼接的。並且能夠提供高層次的指導，而最優秀的軟體工程師就在這裡。

I think the same thing is going to happen in analytics and data analysts, where the vast majority of the data munging of the tool and figuring out what query means what thing? And how do you get to a -- how do you do a segmentation, understand the root cause, like that's all going to be automated by agents. And our goal is to be the first company to do that in a big way.

我認為同樣的情況也會發生在分析和數據分析師身上，他們大部分時間都在處理工具中的數據，弄清楚哪個查詢代表什麼？那麼，如何進行細分，了解根本原因呢？這些都將由代理自動完成。我們的目標是成為第一家大規模實現這一目標的公司。

[Nate Ross], KeyBanc.

[Nate Ross]，KeyBanc。

This is [Nate Ross] on for Jackson Ader. So implied non-GAAP operating margin for 2026 is roughly 2.5%. I guess we were wondering what specific possible sources of upside do you guys see for that number?

替補傑克森·阿德爾上場的是內特·羅斯。因此，2026 年的隱含非 GAAP 營業利潤率約為 2.5%。我們想知道，你們認為這個數字具體有哪些可能的上漲空間？

Well, I'll tell you, first and foremost, we've been on this path where we're increasingly driving revenue growth faster than we're driving expense growth. And rooted within that is efforts on changing our go-to-market, changing our processes, modernizing our own application architectures, doing the basics of running the business in a very efficient way such that we continue to go drive growth with leverage.

首先，我要告訴你，我們一直走的是這樣一條路：我們不斷加快收入成長速度，而加快支出成長速度。而這一切的根基在於我們努力改變市場策略、改變流程、實現應用架構現代化，以非常有效率的方式做好業務營運的基本工作，進而持續利用槓桿效應來推動成長。

Those same things are not just a one-time event. You continue to focus and learn and understand how you can drive and deliver services more effectively. And so we look at the path we have in front of us with respect to our growth opportunities, what our pipelines look like, how we're managing our cost to serve with the expectation that sales and marketing will continue to improve on their efficiencies, G&A will continue to drive efficiencies as well. So it all kind of culminates in the plan that we put together for 2026.

這些事情並非一次性事件。你會繼續專注於學習和了解如何更有效地推動和提供服務。因此，我們檢視擺在我們面前的發展道路，包括我們的成長機會、我們的業務管道、我們如何控制服務成本，並期望銷售和行銷部門能夠持續提高效率，一般及行政費用部門也將繼續提高效率。所以這一切最終都匯聚成了我們為 2026 年所製定的計畫。

Great. Perfect. And I guess, one more follow-up. So regarding customers' analytics budgets. Have you guys noticed any trends or changes specifically with AI affecting these budgets? Like has the current AI landscape affected customers' propensity to invest in analytics in any way?

偉大的。完美的。我想，還需要一個後續問題。所以，關於客戶的分析預算。你們有沒有註意到人工智慧對這些預算有什麼影響，或者說有什麼趨勢或變化？目前的人工智慧格局是否以任何方式影響了客戶在分析領域的投資意願？

Yes. So I'd call it two things. I mean one, it becomes -- it's the bottleneck, right? So you remember that lumen I showed at the start, where it's like, okay, you're shipping all the software. Is it good? Are we even going in the right direction? So the comparative value of the analytics piece becomes a lot greater and more higher urgency. When you have like a year-long road map, it's okay if it takes a while to measure the success of it.

是的。所以我會稱它為兩件事。我的意思是，第一，它就成了瓶頸，對吧？所以你還記得我一開始展示的流明嗎？它就像是，好吧，你要交付所有的軟體。好嗎？我們走的方向對嗎？因此，分析部分的相對價值變得更大，緊迫性也更高。如果你制定了一個長達一年的計劃，那麼需要一段時間來衡量其成功與否也是可以接受的。

But when your iteration cycle is measured in weeks or days like it is with the best of the best companies now, it's like, yes, you need to know if you're going in the right direction all the time. And then I think the other thing is the buyers in addition to that being the pinch point for and the big need in terms of the next big step in product development, they know -- they intuitively all know that this whole space is going to get reformulated with AI.

但是，當你的迭代週期像現在最好的公司一樣以周或天來衡量時，你需要時時知道自己是否朝著正確的方向前進。我認為另一點是，除了這是產品開發下一步的瓶頸和巨大需求之外，買家們也知道——他們憑直覺就知道，整個領域都將因人工智慧而重新定義。

And so they -- again, they're just desperate for education and someone to show them the way. This is not a case of like -- I think one of the differences between selling SaaS and selling AI is, in the SaaS world, it's very much like, okay, talk to your customers, get a list to prioritize features from them and build it and you go back and sell it to them. In this AI world, they don't know.

所以他們——再說一遍，他們只是渴望接受教育，渴望有人為他們指明方向。這不是像——我認為銷售 SaaS 和銷售 AI 的區別之一在於，在 SaaS 領域，更像是，好吧，與你的客戶交談，獲得一份優先級的功能列表，然後構建它，再把它賣給他們。在這個人工智慧時代，他們一無所知。

Like they are like, is this model capable of this? Can it automatically look at a session replay for me? Can it analyze the root cause of a breakage in my funnel? And how -- what's the best way to make that happen? And so they're looking at us for all those questions. And this is where sharing the vision of what the future is as well as being close to the leading edge of the technology is super important.

他們就像在問：這個型號能做到嗎？它能自動幫我查看比賽重播嗎？它能分析出我的銷售漏斗中出現問題的根本原因嗎？那麼，如何才能最好地實現這一點呢？所以他們都在等著問我們這些問題。因此，分享對未來的願景以及緊跟技術前沿至關重要。

Ian Black, Needham.

伊恩·布萊克，尼德姆。

This is Ian Black on for Scott Berg. With the new pricing and packaging, are you planning on separately monetizing your AI agents?

這裡是伊恩·布萊克，他代替斯科特·伯格為您報道。採用新的定價和包裝方案後，您是否計劃單獨實現人工智慧代理的獲利？

So most of our AI agents are embedded within our core platform. And so what you see there is we're getting access to customers to utilize more of the platform. That exemplifies all the power of our modules together. So there's a high propensity that customers who are utilizing our AI agents are both ingesting more data into the platform as well as expanding into other modules.

因此，我們的大部分人工智慧代理都嵌入在我們的核心平台中。因此，您可以看到，我們正在獲得客戶的存取權限，以便他們能夠更多地利用該平台。這充分體現了我們所有模組協同工作的強大威力。因此，使用我們人工智慧代理的客戶很有可能會將更多資料導入平台，並擴展到其他模組。

Now we're also going to introduce new products with continuing -- we've done really well at innovating and some of those products will come out with more fee charges as well. So we're not worried about the ability for us to monetize our AI capabilities. We're actually very excited about the opportunities as we expand the use cases and usage of our platform.

現在我們還將推出新產品，並繼續進行創新——我們在創新方面做得非常好，其中一些產品也將收取更多費用。所以我們並不擔心能否將人工智慧能力貨幣化。隨著我們平台的使用場景和用途不斷擴展，我們對由此帶來的機會感到非常興奮。

Congratulations on the good quarter.

恭喜你本季業績出色。

John Gomez, BTIG.

John Gomez，BTIG。

This is John Gomez on for Nick Altmann. With the shift to Agentic democratizing the end user and product analytics, can you just talk about whether you're seeing new users or new lines of business leverage Amplitude and how that's shifting to go to market. So just any commentary on new end users and how that's impacting how you think about the go-to-market strategy would be helpful.

這裡是約翰·戈麥斯，他代替尼克·奧特曼進行報道。隨著向代理模式的轉變，終端用戶和產品分析變得更加民主化，您能否談談您是否看到新用戶或新的業務線利用 Amplitude，以及這種轉變如何推動市場發展？所以，任何關於新終端用戶以及這如何影響您對市場進入策略的思考的評論都將很有幫助。

It's not really new end users. I mean it's the same. You're talking about product teams. You're talking about marketing teams, engineering and data teams. And so it's the same people trying to leverage the data. What they're really -- again, what they're really desperate for is education. And so when we can show them Global Agent, specialized agents, MCP, AI feedback, AI visibility, what we're doing with our next products and Assistant and LLM analytics, there's always a whole bunch they want to grab on to and say, okay, great, teach me, how to use this, make it successful, everything else.

其實並不是新的終端用戶。我的意思是，它們是一樣的。你們說的是產品團隊。你指的是行銷團隊、工程團隊和數據團隊。因此，還是同一批人在試圖利用這些數據。他們真正渴望的——或者說，他們真正迫切需要的——是教育。因此，當我們向他們展示全球代理、專業代理、MCP、AI 反饋、AI 可見性、我們正在開發的下一代產品以及助手和 LLM 分析時，他們總是有很多東西想要抓住並說，好的，太好了，教我如何使用它，如何成功使用它，以及其他所有的一切。

So that's the biggest difference. It's one where our go-to-market is -- it's about training, getting them to be able to educate, to be able to share the vision, to be able to demo these products and make customers successful.

這就是最大的差別。這是我們的市場策略——透過培訓，讓他們能夠進行教育，能夠分享願景，能夠展示這些產品，並幫助客戶取得成功。

Lucas Cerisola, Morgan Stanley.

盧卡斯·塞里索拉，摩根士丹利。

I'm Lucas Cerisola here for Elizabeth Porter tonight. So with the uptick in new app development that we've seen over the past few months, could you walk through your expectations for balancing this potential new demand from smaller customers with your move-up market as you evolve your go-to-market strategy?

我是盧卡斯·塞里索拉，今晚為您帶來伊麗莎白·波特的報道。鑑於過去幾個月來我們看到的新應用開發數量有所增加，您能否談談您在製定市場進入策略時，如何平衡來自小型客戶的潛在新需求與高端市場需求？

Yes, we're doing both. I think one of the things we see on the start-ups and newer customers is they're very bleeding edge, and so they're trying to push the capabilities of us. And so we've always had a motion where we've taken innovation that we've done with them and bring it to the enterprise in a deliberate way. So we're going to continue to do that.

是的，我們兩者都在做。我認為我們在新創公司和新客戶身上看到的一點是，他們非常追求尖端技術，因此他們試圖擴大我們的能力範圍。因此，我們一直以來都有一種做法，就是將我們與他們一起取得的創新成果，有意識地引入企業。所以我們會繼續這樣做。

I think there's actually massive opportunities, particularly with the rise of Vibe Coded apps. There's going to be Vibe-Coded analytics, too, that needs to go along with all those applications. So it's early, but there's a big opportunity for us there, too. Again, the core thing you want in terms of understanding your customers and knowing if you're going in the right direction and building the product is the same, whether you're -- the newest start-up that was just founded yesterday or your 100-year-old long-lasting business.

我認為實際上蘊藏著巨大的機遇，尤其是在 Vibe Coded 應用程式興起之後。此外，還需要有 Vibe-Coded 分析功能，該功能需要與所有這些應用程式配合使用。現在說這些還為時過早，但對我們來說，那裡也蘊藏著巨大的機會。再說一遍，無論你是昨天才成立的最新創業公司，還是擁有百年歷史的老牌企業，你想要了解客戶、知道自己是否走在正確的道路上以及打造產品的核心都是一樣的。

And so for us, it's like we're here to serve all of them. And again, they're very keen on learning the bleeding edge of what's happening in AI analytics. And so if you're able to teach them that, then it doesn't matter your size.

所以對我們來說，我們來這裡就像是為了服務他們所有人。而且，他們非常渴望了解人工智慧分析領域的最新進展。所以，如果你能教他們這一點，那麼你的體型大小就無關緊要了。

Got it. That's super helpful. And then could you speak to the seven-figure deal pipeline in 2026? And then are there any specific verticals in which you see outsized growth already?

知道了。這太有幫助了。那麼，您能否談談 2026 年的七位數交易計畫？那麼，您認為有哪些特定的垂直領域已經出現了超常成長？

I mean we're seeing -- like as I said on the call, we're seeing a lot of AI companies use us. We have 25 over $100,000. And then we have a seven-figure contract with one of the largest foundational model labs out there, who's been a customer starting last year. And so that's very, very exciting because they obviously know what's going on when it comes to what's possible, and they see a future world where we're a really big part in that.

我的意思是，就像我在電話會議上說的那樣，我們看到很多人工智慧公司都在使用我們的技術。我們有 25 家公司年收入超過 10 萬美元。然後，我們與業內最大的基礎模型實驗室之一簽訂了一份七位數的合同，該實驗室從去年開始就是我們的客戶。所以這非常令人興奮，因為他們顯然知道未來有哪些可能性，並且他們看到了一個我們將在其中扮演重要角色的未來世界。

YC Wong, Citi.

黃永昌，花旗銀行。

Congratulations on a pretty strong close to the year here. I guess maybe I just want to touch on, Spenser, you talked about Amplitude being one of the largest database of user behavior. Just given the rapid progress of agentic capability across our data platform players called Snowflake and Databricks, where we are seeing also customer consolidation towards maybe bigger data platform players.

恭喜你以非常出色的成績結束了這一年。我想稍微提一下，史賓塞，你剛剛提到 Amplitude 是最大的使用者行為資料庫之一。鑑於 Snowflake 和 Databricks 等數據平台廠商的代理能力發展迅速，我們也看到客戶正在向規模更大的數據平台廠商整合。

Curious if you're seeing any -- I think customer blur of like your analytics use cases from Snowflake and Databricks versus Amplitude?

我很好奇你是否注意到——我認為客戶在區分 Snowflake 和 Databricks 與 Amplitude 的分析用例方面存在一些模糊地帶？

I want to make sure, YC, I want to make sure I understand what you're saying, you're saying do we see competition from Snowflake and Databricks because they have a lot of data too --

YC，我想確認一下，我理解你的意思了。你是說，Snowflake 和 Databricks 也會構成競爭，因為它們也擁有大量資料。--

Well, it's not just data. They are thinking about doing the application side as well. I mean if you think about Vibe Coding and then you think of application building, you can make it easier to build. I'm just curious if you see anything blurring between customer talking about just use cases between a customer you can use a data platform like Snowflake to build it, they have Cortex versus what you will see with Amplitude?

其實，這不只是數據的問題。他們也在考慮開發應用程式方面的內容。我的意思是，如果你想想 Vibe Coding，然後再想想應用程式構建，你就可以讓構建變得更容易。我只是好奇，您是否注意到客戶在談論用例時，例如客戶可以使用像 Snowflake 這樣的資料平台來建立應用程式（他們有 Cortex），而使用 Amplitude 則會出現類似的情況，兩者之間是否存在任何模糊之處？

So one of the big things that we see is that customers always want the most advanced and bleeding edge capabilities. Like I heard this great analogy the other day where software is very much like sushi. So it's fine that the gas station at 7-Eleven offers it, but Jiro in Japan is not -- probably not going out of business. In fact, it's going to create more demand for him.

我們發現的一個重要現像是，客戶總是想要最先進、最前衛的功能。就像我前幾天聽到的一個很棒的比喻，軟體很像壽司。所以，7-Eleven 便利商店的加油站提供這種產品沒問題，但日本的 Jiro 便利商店卻不提供——這大概不是因為 Jiro 要倒閉吧。事實上，這反而會讓他更受歡迎。

And so from our standpoint, what we think about is how can we offer the most advanced and robust system for analytics. So if you look at the benchmark that we -- with hundreds of evals that we released, where we got a 76% accuracy rate, if you look at the Cortex or you look at DataBricks Genie, I mean they're going to be in the 10% or sub that.

因此，從我們的角度來看，我們思考的是如何提供最先進、最強大的分析系統。所以，如果你看看我們發布的基準測試——我們進行了數百次評估，準確率達到了 76%，那麼如果你看看 Cortex 或者 DataBricks Genie，我的意思是，它們的準確率會在 10% 甚至更低。

We're working on releasing full metrics on that. And that's because the text-to-SQL is only really one part of it. The other -- there's two other big parts. The first is the context layer. So what data sources are you bringing together in the right way, analytics data, session replay data, data from interactions and guides and survey data from other sources and interpreting those in the right way and then giving an LLM agent, the right set of tool calls so that they can iteratively query, okay, hey, what's my onboarding funnel?

我們正在努力發布這方面的完整指標。這是因為文字轉 SQL 只是其中的一部分。另外——還有兩個主要部分。第一層是上下文層。那麼，你該如何正確整合哪些資料來源呢？分析數據、會話回放數據、互動和指南數據以及來自其他來源的調查數據，並以正確的方式解釋這些數據，然後為 LLM 代理提供一組正確的工具調用，以便他們可以迭代地查詢，例如，“我的新用戶引導流程是什麼？”

Where is the biggest drop on it? Why is the biggest drop on it? What's the biggest difference between users who went to the next step versus the previous step, right? So that's like just in that example, that's four-queries that you're going to have to do in a row all correctly. And to do that, you need to prop the LLM in a very particular way.

哪裡落差最大？為什麼跌幅最大的是它？對吧？進入下一步的用戶和進入上一步的用戶之間最大的差異是什麼？所以，就拿這個例子來說，你需要連續正確地執行四個查詢。而要做到這一點，你需要以一種非常特殊的方式來支撐LLM。

You need to give the right tool calls, you need to give it the right context. And so we have thought really deeply about because we have the largest repository of user behavioral data in the world, we have thought very deeply. We have seen what millions of analytics queries, what good looks like for millions of analytics queries and translated that into an agent that does the same.

你需要呼叫正確的工具，你需要提供正確的上下文。因為我們擁有世界上最大的使用者行為資料儲存庫，所以我們對此進行了非常深入的思考。我們已經了解了數百萬次分析查詢的結果，以及數百萬次分析查詢的良好表現，並將這些結果轉化成一個​​能夠執行相同操作的代理。

And so because, again, you're going to need to give it all that context and then be able to iteratively query a data system, the differences in accuracy are really, really stark if you were just to roll your own or use a Genie or Cortex versus using an Amplitude. And when you're an analyst, that difference between 76% and 10% is a massive difference in terms of your ability to leverage agentic analytics.

因此，因為你需要提供所有這些上下文，然後能夠迭代地查詢資料系統，所以如果你自己開發或使用 Genie 或 Cortex 與使用 Amplitude 相比，準確性方面的差異真的非常非常明顯。而對於分析師來說，76% 和 10% 之間的差距，在你利用智慧分析的能力方面，是巨大的差距。

No, that's helpful color. Maybe one for Andrew. The profitability definitely came in well ahead of expectations here. Maybe you guys are leveraging some agents internally that helps to drive better sales efficiency. But curious to see going into next year, like what can we expect, especially on the free cash flow that outperformed probably saw an expansion by 4 points.

不，那是很有幫助的顏色。或許給安德魯一個。此次獲利情況絕對遠超預期。或許你們公司內部有一些代理商，有助於提高銷售效率。但很好奇明年會怎樣，我們可以期待什麼，特別是自由現金流方面，表現優異的自由現金流可能增加了 4 個百分點。

Like curious to see what your expectation into next year? And any other moving parts that we should be aware of or headwind from to be mindful from the strong performance this year?

很好奇你對明年有什麼期待？除了以上因素外，還有哪些變動因素或不利因素會影響今年的強勁表現，我們需要注意嗎？

I think what you're seeing is that the efforts we've been doing on sales and marketing, on our cost to serve and our G&A and operating more effectively as a company, is not just a one effort, one activity. There's multiple. And certainly, we're introducing agentic capabilities into our own workflows within the company, and that's certainly contributing to it.

我認為你們看到的是，我們在銷售和行銷、服務成本、一般及行政費用以及提高公司營運效率方面所做的努力，並非僅僅是一項努力或一項活動。有多個。當然，我們正在公司內部的工作流程中引入代理功能，這無疑對此有所貢獻。

But there's so many things structurally we've done to the business to create greater durability that that's ending in greater abilities for us to drive efficiencies. I'll give you one example. We've talked a lot about our ability to go drive increasing contract duration to our customers and that our RPO has been growing rapidly. Well, if you don't have to renew your installed base every year or that installed base percentage goes down, because you're executing more and more longer-term duration contracts with your customers.

但是，我們在業務結構上做了很多事情，以提高業務的持久性，最終使我們能夠提高效率。我舉個例子。我們已經多次談到我們有能力推動客戶延長合約期限，而且我們的RPO業務也一直在快速成長。如果你不必每年都更新你的已安裝用戶群，或已安裝用戶群的百分比下降，因為你與客戶簽訂了越來越多的長期合約。

Then the sales team has more time to dedicate towards selling new and expansion deals rather than working on renewals. And so it's just a great example of a strategy we put in place that's going to accrue benefits for a longer period of time.

這樣一來，銷售團隊就有更多時間專注於銷售新合約和拓展新合同，而不是忙於續約。因此，這正是我們所製定的一項能帶來長期效益的策略的絕佳例證。

Willow Miller, William Blair.

柳樹磨坊主，威廉‧布萊爾。

I'm Willow on for Arjun Bhatia. Andrew, in terms of guidance, the full year revenue range seems a bit wider than normal at $8 million. Can you help us understand the reason for this? And what scenarios are contemplated at the low and high ends of the range?

我是Willow，替Arjun Bhatia報道。Andrew，就業績指引而言，全年營收預期範圍似乎比平常寬一些，為 800 萬美元。您能幫我們了解一下原因嗎？那麼，在範圍的低端和高端分別考慮了哪些情況呢？

I think when we approach we approach our guidance, we approach it with what we think we can go execute in the period. And I wouldn't read too much into that other than we have a breadth of different opportunities that we're going after both with our product set, with improvements in our targeting enterprise customers. So I wouldn't read too much into it.

我認為，當我們制定指導方針時，我們會根據我們認為在這段時間內可以執行的事情來製定方針。我不會對此過度解讀，只是我們正在透過我們的產品組合以及改進我們針對企業客戶的策略，去追求各種不同的機會。所以我覺得沒必要過度解讀。

And that will conclude our fourth-quarter earnings call. Thank you for your time and interest, and we look forward to seeing you on the road this quarter as we attend conferences hosted by Baird, Citizens, KeyBanc, Morgan Stanley and others. Take care.

我們的第四季財報電話會議到此結束。感謝您抽出時間並對此感興趣，我們期待在本季度與您在貝爾德、Citizens、KeyBanc、摩根士丹利等機構主辦的會議上見面。小心。