Veritone Inc (VERI) 2017 Q4 法說會逐字稿

Good afternoon. Welcome to the Veritone's Fourth Quarter and Full Year 2017 Earnings Conference Call. Joining us for today's call are Veritone's Chairman and CEO, Chad Steelberg; and the company's Chief Financial Officer, Pete Collins. (Operator Instructions)

Please note that certain information discussed on the call today will include forward-looking statements about future events and Veritone's business strategy and future financial and operating performance, including (technical difficulty) of its expected operating performance for the first quarter of 2018. These forward-looking statements are only predictions and are subject to risks, uncertainties and assumptions that are difficult to predict and may cause the actual results to differ materially from those stated or implied by those statements. Certain of these risks and assumptions are discussed in Veritone's SEC filings, including its registration statements on Form S-1 and its quarterly reports on Form 10-Q.

These forward-looking statements reflect management's beliefs, estimates and predictions as of the date of this live broadcast, February 26, 2018, and Veritone undertakes no obligation to revise or update any forward-looking statements to reflect events or circumstances after the date of this conference call.

In addition to the company's GAAP financial results, during this call management will be discussing the company's earnings before interest, expense, depreciation, amortization and stock-based compensation or EBITDAS, which is a non-GAAP financial measure. A reconciliation of the company's EBITDAS to its net loss is attached to the company's press release issued today. The company is providing this supplemental information because management believes it would be an important measure of performance that is commonly used by securities analysts, investors and other interested parties in evaluation of the company and its industry.

Management also uses this information internally for forecasting and budgeting. This non-GAAP measure may not be indicative of the historical operating results of Veritone or predictive of potential future results. Investors should not consider EBITDAS in isolation or as a substitute for analysis of the company's results as reported in accordance with GAAP.

Finally, I would like to remind everyone that this call will be recorded and made available for replay via a link available in the Investor Relations section of the company's website at www.veritone.com.

Now I would like to turn the call over to Veritone's Chairman and CEO, Chad Steelberg. Sir, please proceed.

Welcome, everyone, and thank you for joining us today. After the market close, we issued a press release announcing our results for the fourth quarter and full year ended December 31, 2017, a copy of which is available in the Investor Relations section of our website.

We are proud of our achievements in 2017. Our AI SaaS net revenues grew by 189% year-over-year. Our Media Agency's net revenues grew by over 50%. We made 1 strategic acquisition and we completed 2 common stock offerings. We expanded our customer base globally, now with customers on 4 continents, and diversified our direct sales process to channel partners and system integrators.

We continue to enhance the capabilities and performance of our proprietary AI operating system and Conductor technology, producing more accurate and timely AI-powered solutions across a broader set of customer use cases.

I will discuss some of these highlights in more detail a bit later on today's call. But first, I'd like to have our CFO, Pete Collins, walk us through our financial results and key performance indicators for the fourth quarter and full year of 2017. Pete?

Thank you, Chad, and good afternoon, everyone.

Turning to our financial results for the fourth quarter and full year ended December 31, 2017. Our net revenues in the fourth quarter of 2017 increased 40% to $3.5 million from $2.5 million in Q4 of last year. Our Media Agency revenues grew from $2.2 million in 2016 to $3.0 million in 2017, an increase of 37%. And our SaaS licensing revenues from our AI platform business grew from $0.3 million in 2016 to $0.5 million in 2017, an increase of 61%.

For the full year of 2017, our net revenues increased 62% to a record $14.4 million from $8.9 million in 2016. Our Media Agency revenues grew from $8.4 million in 2016 to $12.9 million in 2017, an increase of 54%. And our SaaS licensing revenues grew from $0.5 million in 2016 to $1.5 million in 2017, an increase of 189%.

Looking more closely at our AI platform business, the total contract value of new bookings received during the fourth quarter increased to $360,000 from $163,000 in Q4 of last year, an increase of 121%. In addition, our monthly recurring revenue under agreements in effect at the end of Q4 2017 increased 276% year-over-year to $173,000 from $46,000 in the fourth quarter last year.

Our gross profit in the fourth quarter of 2017 increased 63% to $3.3 million or 93.0% of net revenues from $2.0 million or 79.6% of net revenues in Q4 of last year.

For the full year of 2017, our gross profit increased 82% to $13.3 million or 92.6% of net revenues from $7.3 million or 82.3% of net revenues in 2016. The increases in gross profit and gross margin for both the quarter and year were due primarily to the operating leverage provided by our higher net revenues level.

Turning to our expenses. Total operating expenses in the fourth quarter of 2017 increased 72% to $16.1 million from $9.3 million in Q4 of last year. For the full year, our total operating expenses increased 93% to $60.1 million from $31.1 million in the same period last year. The increase in operating expenses for both periods was due primarily to higher stock-based compensation expense and increased personnel costs, particularly in software engineering, data science and sales and marketing, as we continue to expand our business and enhance our AI platform, including developing new products and functionality.

Our stock-based compensation expense in Q4 2017 was $2.5 million compared with $0.1 million in Q4 2016. For the full year of 2017, it was $16.1 million and in the full year 2016, it was $1.7 million. The fourth quarter increase was primarily due to stock options granted at the time of our IPO and the full year increase was due to those IPO grants along with divesting of some performance-based stock options.

Our loss from operations in the fourth quarter of 2017 was $12.8 million compared with the loss from operations of $7.4 million in the fourth quarter of 2016. For the full year of 2017, our loss from operations was $46.8 million compared with a loss from operations of $23.8 million in 2016. The higher loss from operations was driven by higher stock-based compensation expense and personnel costs, as I noted before.

Net loss attributable to common stockholders in the fourth quarter of 2017 totaled $12.8 million or a loss of $0.83 per share compared with $9.9 million or a loss of $4.12 per share in Q4 of last year. These EPS figures are based on $15.3 million weighted average shares outstanding for the fourth quarter of this year compared with $2.4 million weighted average shares outstanding in Q4 of last year.

For the full year, net loss attributable to common stockholders totaled $64.1 million or a loss of $6.20 per share compared with $30.2 million or a loss of $14.59 per share in 2016. These EPS figures are based on 10.3 million weighted average shares outstanding for 2017 compared with 2.1 million weighted average shares outstanding in 2016.

In addition to the higher stock-based compensation expenses and personnel costs I noted earlier, our net loss attributable to common stockholders for 2017 also included other expenses of $12.9 million related primarily to the write-off of debt discounts and issuance costs upon the conversion of our outstanding convertible notes and the issuance of stock warrants upon our IPO and noncash charges of $4.5 million related to the accretion of preferred stock.

Looking at our fourth quarter 2017 net loss on a non-GAAP basis, our earnings before interest expense, depreciation, amortization and stock-based compensation expense or EBITDAS was a loss of $10.2 million compared with a loss of $7.2 million in the fourth quarter of 2016.

For the full year of 2017, our EBITDAS was a loss of $30.2 million compared with a loss of $21.5 million in 2016. The increased EBITDAS loss for both the quarter and full year was due primarily to our higher headcount levels, particularly in software engineering, data science and sales and marketing. We expect these investments to lead to further enhancements of our aiWARE product and increased net revenues in the future.

Now turning to our cash position. At the end of the quarter, we had cash and cash equivalents and marketable securities of $69.1 million and no long-term debt.

Next, I'd like to shift gears to our key performance indicators, or KPIs, for both our Media Agency and our AI platform business. As I've mentioned on previous calls, in order to track the progress of our Media Agency business, we evaluate 3 key performance metrics: the number of new clients added under master service agreements, the total number of active clients and the average media spend per client. The first 2 of these KPIs, which track the number of new and active clients, provide us with insight regarding our ability to grow the market share of our Media Agency business by winning new accounts as well as keeping tabs on client churn. During Q4, we added 14 net new clients compared with 15 net new clients in the same period last year. We had a total of 57 active clients as of the end of the quarter compared with 39 at the prior year quarter-end, a 46% increase.

Our third KPI for Media Agency business is average media spend per customer, which allows us to analyze not only client spending trends but our ability to grow media spend with our existing clients. During Q4, our average media spend per client was $464,000 compared with $572,000 in the same period last year, which is a 19% decrease. It's important to keep in mind that while this business is mature and provides a solid foundation for Veritone, it can experience volatility in net revenues from time to time as our clients may shift their ad spending based on their specific business trends, which was the case in the fourth quarter of 2017.

It's important to note that our Media Agency business and its success are heavily dependent on Veritone's AI platform.

Moving to our AI platform's KPIs. We track a number of KPIs for this business, which we believe represent the best metrics to evaluate the growth and success of our company at this stage. The first 2 KPIs we track in this business are the number of customers and accounts, which show the growth in our customer base for the platform. We had 57 customers on the platform at the end of the quarter compared with 18 at the prior year quarter-end, an increase of 217%. We had 467 total accounts on the platform at the end of the quarter compared with 23 at the prior year quarter-end, which exceeded our goal of 425 and represented a more than 19-fold increase year-over-year.

The next set of KPIs for this business provide insight into the level and growth of our platform's cognitive capabilities. We had 151 third-party active cognitive engines on the platform at the end of the quarter compared with 41 at the prior year quarter-end, which exceeded our goal of 150 and represented a 268% increase year-over-year. These engines span 16 cognitive categories, such as video object detection and natural language processing compared with 13 at the prior year quarter-end.

The final KPI is the number of hours of data processed on our aiWARE platform. During the quarter, we processed 1.36 million total hours of video and audio content compared with 447,000 in the prior year period, an increase of 204%. During the full year of 2017, we processed 2.87 million hours of video and audio content, which exceeded our goal of 2.75 million and represented a 151% increase over 2016.

As we grow and expand our AI platform, we will continue to provide goalpost to make it easier to track our success. With that in mind, in the first quarter of 2018, we expect to add approximately 8 customers, 110 accounts and approximately 20 active third-party cognitive engines on our platform, including 1 new cognitive category. Also we expect to process approximately 1.9 million total hours of video and audio content during the quarter.

That completes my financial summary. I will now turn the call back over to Chad. Chad?

Thank you, Pete. And now I'd like to share my perspective on Veritone as we shift from 2017 into 2018.

I'll start by saying this: we are in the first inning of the AI revolution. So many people want AI to solve their fundamental, really difficult challenges. I'm talking about the [10xSolutions], cure cancer, stop terrorism, et cetera, et cetera. The reality is, we are all on one of the most profound journeys in human history, where men and women are no longer the sole drivers of innovation and our biological cognitive limitations are being overcome and augmented through artificial intelligence. This AI journey will take time, not only from the technology to mature but also for the marketplace to adopt these potentially disruptive, yet powerful AI solutions.

To address these challenges, Veritone has developed the first AI operating system that delivers meaningful and measurable results to our customers today and is capable of evolving with the marketplace to tackle the [10x] opportunities when the time is right.

We believe that there are 3 fundamental truths regarding AI today: first, that every industry will be radically transformed by AI; second, that businesses must adopt AI to remain competitive; and third, that corporate spending on AI technologies will quickly grow to significant, if not dominant levels, as a result.

Our approach to AI is unique, both in terms of our value proposition to our customers as well as the proprietary science that powers it. Veritone's AI operating system, aiWARE, and the applications and cognitive engines that execute on it, are designed to solve important, mission-critical problems within our target go-to-market industry today.

Our AI solutions harness the power of rapidly expanding set of cognitive engines, machine learning algorithms developed by third parties within the global data science community, which are deployed to Veritone's developer application and orchestrated by Conductor, Veritone's proprietary machine learning orchestration technology, to produce unprecedented levels of accuracy and multi-variant intelligence for our customers.

As artificial intelligence evolves from narrow AI with point solutions today to general artificial intelligence solutions with broader cognitive capacity in the future, Veritone is leading the way in helping our customers adopt, adapt and exploit AI solutions through aiWARE.

In 2017, we had a number of learnings. Despite the growth in available cognitive engines, we were asked to solve issues for businesses that exceeded the capability of engines on the platform, and even of those within the broader global ecosystem. In most cases, the cognitive gap was a result of third-party engine providers training models that were too broad. For example, being able to identify 100,000 objects within a video clip, and as a result, lacked the accuracy our customers demanded for a specific narrow subset of objects. As a result, we worked with our partners to train more narrowly focused cognitive engines, resulting in higher levels of accuracy. We were able to onboard these new models with minimal to no engineering involvement through our developer application to meet our customers' needs.

As we enter 2018, we are seeing a shift in the machine learning ecosystem. Companies are now building trainable models that will Veritone to develop micromodels to fill in the cognitive gaps very rapidly. I believe this trend will continue as a result. AI solutions will depend on conductive learning to efficiently orchestrate these models in the future.

As I expected even before Veritone was formed in 2014, the number of cognitive engines continues to explode at exponential rates. Their accuracy and speed are increasing while their costs are decreasing, which is fantastic news for Veritone and our customers. It means we can orchestrate those engines to deliver more powerful applications to our customers across a broader set of use cases and industry verticals.

We found in 2017 that keeping data behind firewalls is paramount to many organizations. To address this need, we developed a hybrid version of our cloud-based aiWARE that moves the cognitive engines behind the customer's firewall instead of moving their data to the cloud for processing. As 2018 unfolds, I believe the boundaries between the cloud, on-premise and the Edge will diminish, allowing our customers to flexibly deploy and configure aiWARE to meet their business, security and budgetary requirements.

We are also developing a stand-alone network isolated version of aiWARE, which will enable substantially all of our functionality of cloud-based aiWARE to be performed behind the customer's firewall to serve organizations with stricter security requirements.

In 2017, we saw the need for speed. As the speed of cognition approaches real-time levels, customers are able to apply AI solutions to more valuable use cases. To support real-time cognition, we are working on modifications to aiWARE to support real-time data ingestion, storage, cognitive engine execution, orchestration and use case specific actions. We have already achieved real-time performance levels in a specific use case and we expect to have a more universal solution in 2018.

Conductor, our proprietary AI orchestration system, efficiently selects and combines multiple natural language processing engines, resulting in higher levels of accuracy in any single cognitive engine on the platform. We proved this in 2017 in the media and legal verticals and we expect our performance advantage to increase as we onboard more NLP engines. It's important to note that in many of these new customer opportunities, Veritone competes against a host of Tier 1 technology providers and delivers better results, making us more convinced than ever that the future of AI depends on conductive learning and an AI operating system.

In 2018, we plan on upgrading Conductor and integrating several of the new trainable cognitive engines into the orchestration process, thus closing the loop and allowing aiWARE to evolve and learn more autonomously. Internally, we call this process the virtuous cycle, and if achieved, will push Veritone 1 step closer to human levels of cognition and deliver unprecedented value to our customers. This is difficult science and will require significant breakthroughs, but our preliminary results are promising. We are also working to extend Conductor's intelligent orchestration capability to other cognitive platforms.

In 2017, the majority of our customers and accounts were sold to our direct sales force. In Q4, we had our first win to a channel partner. And as we enter 2018, our sales pipeline is quickly being filled by channel partners, system integrators and technology partners. I expect this trend to continue, which will accelerate our market penetration both geographically and across industry verticals.

Our SaaS revenue in 2017 came primarily from customers within the media and entertainment vertical, with some moderate ramping in Q4 within the legal and compliance market. In 2018, I expect our legal and compliance sales to grow quickly. However, the trajectory will be both unpredictable and lumpy as a result of the industry dynamics.

In 2017, we laid the groundwork to enter the government and public safety markets. In 2018, we will continue to invest in the infrastructure and government certifications required to operate at scale within these markets. We've been very pleased with the results of our initial POCs and expect to start generating moderate revenue in the latter half of this year.

While organic expansion of our customer base and technology continue to fuel our market success and revenue growth, we're continuing to explore inorganic opportunities as they present themselves. We have a very active pipeline of synergistic targets and have signed several letters of intent. We will keep you apprised of any significant development in this area.

I appreciate the continued support of our investors who share Veritone's mission to deliver purpose-driven artificial intelligence. As we continue to increase the capabilities of aiWARE, I believe that our land-and-expand approach to deploying AI solutions to our customers will deliver meaningful and measurable value in the short term and 10xSolutions in the not-so-distant future, resulting in tremendous financial rewards to our customers and our shareholders.

Again, thank you all for your time today. We look forward to updating you on your progress on future calls. We're now ready to open the line for questions. Operator?

(Operator Instructions) Our first question comes from the line of Mike Latimore with Northland Capital.

On the fourth quarter bookings, was that in the -- mostly in the legal vertical? Or can you describe what went into the SaaS bookings in the fourth quarter?

Mike, it's Pete. So the bookings in the fourth quarter were primarily in the media vertical. The legal vertical is looking like bookings will happen in 2018. But that media -- that Q4 number was really in the media vertical.

Got it. And then what's -- I don't know if you can report this, but what is your sale headcount? And do you expect that to grow this year or do you expect to focus more on, kind of, the channel?

So the AI sales team basically doubled in '17 from the beginning of the year to the end of the year. And I think that for '18, we're looking at slowing down that pace. I think we've got a lot of the right team members in place today. That was a lot of the build we did especially in the fourth quarter. And so as we look out into '18, we're not expecting to see a significant ramp up there. The sales lift in '18 is really going to come through the channels and through the existing infrastructure we've got from a sales and marketing perspective.

Okay. And then this deal you announced with Entercom, is that a first quarter booking? Can you describe that relative to the other media -- big media deals you have, like CBS Radio, I heard that sort of thing?

Yes. It's a Q1 booking. So we're excited about it. It's nice that the work we've done with CBS Radio carried over into basically doubling the size of the deal with the combined Entercom-CBS Radio organization.

Okay. And last, I know the number of cognitive engines that you guys track continues to grow rapidly. Any rough guestimate of where we might be at the end of 2018 in terms of number of cognitive engines in the market?

Well, as you know, we gave a target for the end of the first quarter as increasing by 20. I don't know -- I'm turning here to Chad to see if there's any other color you'd like to share on that at this stage?

Yes. Mike, I think our focus in 2018 is really going to be about the -- not just the number of engines, but actually the quality of those engines, the lift and the breadth in terms of the different cognitive categories that they operate in. Our customers are continuing to push us into new fields in terms of different areas of cognition. So I think this year we'll be looking at having 4 additional categories of cognition and probably staying on pace and watching that -- the number of engines, actually, accelerate beyond our Q1 objectives that Pete has already talked about.

And our next question comes from the line of Darren Aftahi with Roth Capital Partners.

Just a couple, if I may. Chad, on your comments about, kind of, the area of process of narrowing down, kind of, the target of the cognitive engine, I'm sort of curious that the time frame -- or is that sort of a moving target, if you will, but maybe more of a time frame where that's more mature and you can more readily, kind of, address your customers needs? Secondly, on the customer -- on the total account number of out-performance, I know you had iHeart in there. But if you can just talk about any specifics exceeding that metric? And then on your developer application, could you talk about the traction you've seen since you've launched that? And then just lastly on the media spend and to your comments, I'm just curious if that's something where you've seen that media spend per customer kind of shift back in Q1 and normalize or something that's kind of trend throughout 2018 and beyond?

Sure. Why don't I take your first and third question and combine the 2, which was the number of cognitive engines and how they relate to the developer application. So the number of cognitive engines on the platform, again, is very important to us. It really demonstrates the intelligence capabilities of the platform and what type of value we can offer our customers when we start combining them in Conductor and the level of intelligence we can bring. I talked about this on the call a little bit, but we've seen a shift from when we founded the company back in 2014, where the data scientists and the [starts] were actually heavily involved in actually developing the mathematics behind deep learning and other forms of neural networks. Today, we're seeing that shift actually building solutions that can generate new engines -- sorry, new cognitive engines on the fly. So that essentially, no longer do you have data scientists in the middle of that process that can take weeks, if not months, to build a new engine for us. And by moving from, again, that kind of handcrafted modeling to more programmatic in the marketplace, it's truly going to allow us through the developer applications to accelerate the number of engines to be far more nimble in terms of what we can deliver to our customers and fill those gaps. So they go hand in hand. Without developer application, we would have to have engineers on our side onboarding that technology, and too, without moving the marketplace to a more programmatic development of cognitive engines without data scientists reaching that, what I call, that virtuous cycle would be nearly impossible. So 2018, those 2 questions I think are tightly tied together to our success and growth as we expand our customer base. Pete, do you want to take the other 2 questions?

Yes. So Darren as it relates to the account growth in the fourth quarter, so we saw that really coming across in 2 different verticals. It was in the media vertical and a lot of that was in the radio group within the media vertical, and then it was also in the legal vertical. So we had account growth across both of those spectrums. But as you pointed out, the biggest one -- the biggest driver was picking up the iHeart stations. And then on the media spend, the -- if I look back a year ago, in the first quarter, the media revenues were up about 42%. And for the year last year, they were up 54%. I mean, that's pretty significant growth that we haven't forecasted with continue out into the future. We're very confident in the business. It's doing very well. But a 42% or 54% increase year-over-year isn't what we expect on a go-forward basis. So I think that the level that we'd expect in -- throughout '18 would be, sort of, a strong, but more moderate growth rate than what we saw in '17.

And our next question comes from the line of Joe Fadgen with Craig-Hallum.

On here for Chad today. I want to ask about the artificial intelligence platform and revs. I mean, I guess, given some of the underlying metrics, hours of data processed, the bookings that you've received especially in Q3, the accounts and the engines on the platform, I guess, we haven't really seen the revenue from that business grow as fast as we had expected. Can you -- I guess, one, tell us if it's growing, kind of, in line with what you were planning and maybe we were just a little bit aggressive? And if not, can you kind of explain what that disconnect is? Or why there's such a deviation between some of those metrics and the actual revs?

Yes. This is Chad. The biggest thing that we're seeing in the market today is more of a cautious approach for the big companies in terms of their engagement with AI and it's part of our core strategy. We're not out there trying to convince banks and media companies to go spend $50 million, $60 million of consulting services as you see other players in the AI world do. We have a SaaS-based solution that allows us to turn a customer on a number of hours on our platform. And we're in the market to solve the 2 and 3x problems they have today that they can actually demonstrate and measure ROI. And as such, number one, the time to onboard those customers is less, but also the amount of revenue that we're generating from those Tier 1 customers has also lessened. So our goal is, again, similar to what Eric Schmidt used to talk about, which is land the customer, demonstrate tremendous amounts of value, prove your ROI and then expand with that customer as they start to adopt AI across the enterprise. And we're seeing this heavily with NBC, one of our key customers. When they started, if you look at year-on-year growth or even quarter-on-quarter growth with them -- I know, Pete, we don't provide this data, but it really demonstrates -- we started with a very small set of users in a very specific use case and we're expanding that across 11 different use cases they've identified throughout the enterprise and continue to show substantial internal growth with that account. We're seeing that similar type of behavior with CBS now moving into Entercom. With that acquisition, they're doubling nearly the size of their footprint. So again, winning these Tier 1 accounts, we're beating out the other major players that we're competing against. And we expect as AI becomes more prevalent and those customers become more readily used to using AI for augmented intelligence that they will start to grow and expand that revenue stream. I am disappointed that -- again, we're not landing $10 million-plus accounts. We hoped that would be happening last year. We have many of those now in our pipeline for 2018. So I think that those bigger solutions will start to hit. But at the end of the day, we're more than happy to win those accounts by brand, demonstrate value in ROI and grow with our customers at scale.

Okay. And then a follow-up on the M&A. I mean, the acquisition that you announced recently as well as what you have in the pipeline. Just thinking about what it means from a cash burn or cash flow perspective? Is this something where -- whether you're acquiring them and they're operating around breakeven or slightly profitable? Do you think there's significant, like, incremental spend that you'll have to make to get these businesses integrated with you? Or is it largely -- you can just, kind of, dump them right into your platform or your technology without making large investments?

So we believe that due to the heavy investment we've made already in the developer application VDA, that integrating and onboarding these target acquisitions will be extremely cost effective and efficient. We're integrating right now with partners that we are not acquiring and we are seeing fantastic results both in terms of the benefit they're seeing by integrating with Veritone, but also the speed at which their developers can do this independently of us. So I think that will remain consistent, even probably more efficient, once those acquisitions are made with target companies.

And our next question comes from the line of Sameet Sinha with B. Riley.

A couple of questions. So I'm just looking at the SaaS business, the revenue per hour, that came down. You're expecting number of hours to grow sequentially as well. Is there a specific use case that gained favor in the fourth quarter and that's going to continue into Q1? Secondly, if you can talk about your channel strategy. As you reduce headcount hiring in your direct sales team, what sort of leverage do you expect and if you can also shed some light on the economics there, that would be helpful? And last question is, talking about the cash burn, it seems like it came in a little higher than you had usually delivered. So if you can just you talk about the strategy going forward.

Sameet, it's Pete. Let me start with your third question there about the cash burn. So the challenge for us is to be able to anticipate the revenues and then bring in the resources that match up with that in order to hit the cash burn. And we did end up for the year and for the quarter going over what we had anticipated. In '18, we're expecting that we're going to bring the burn rate down on a year-over-year basis for all of '18 versus '17. But -- and what we're doing is really refining the method that we're using for analyzing what our revenues are going to be and then making sure we're bringing in the right resources to match up with that. So we do expect to bring that run rate -- that burn rate down in '18 versus '17.

Sameet. To address the other question, can you repeat it, just so that all the listeners can hear it again?

Sure. So I was just looking at the number of hours of data processed. So if you look at the average revenue per hour, I saw that number come down sequentially and year-over-year as well. So I'm just trying to understand is there any specific, sort of, use case that grew significantly because the number of hours expanded pretty materially in the fourth quarter. So is there any specific, sort of, usage or application that customers are using you more for that's led to that average coming down? Maybe it's a low-priced application that was used. And similarly, do you expect a similar, sort of, dynamic even in Q1 considering that you're guiding towards a higher number of hours metric for Q1?

So I've talked about this before and I'll reiterate, which is, I think, using the revenue per hour is a poor metric to evaluate our business. The reason why I state that is because it would be like trying to compare Google, which -- how much content they scan in terms of their search index and using that as the basic denominator for then against how much revenue they generate. Our business, in terms of the amount of content that we process for our customers, will continue to accelerate, which really is an asset that benefits us in a couple of ways. One, once it's in our platform, customers then have a much higher stickiness and propensity to use that same asset and use different engines against it and different applications to analyze it and drive value. So winning that customer, building that sticky relationship and expanding with them is critical to us. So that's why I primarily don't believe that's a good metric. Further, I think that our strategy really isn't about just revenue per hour of content, if you will, analyze. We are looking at various acquisitions in various different industries, which have tens of millions of hours of content, but today are using 0 cognition. So with those acquisitions coming onboard, you're going to see that denominator inflate by nearly 3 to 5x with almost 0 cognitive processing initially on the platform. So I think we should move away from that as a metric. I think it's important to look at 2 key things: number of customers and the uplift of all 3 vectors; number of customers and accounts, the size of that asset relationship that we're building, which is measured in hours of content; and three, the breadth of the cognition that we can make available to our customers. We will have some additional metrics this year with regards to the number of applications on the platform that can deliver both revenue and more intelligence for those customers. But at this point in time, we're still not providing guidance for metrics there.

Got it. If you can just talk about the number of categories that you plan to increase. You mentioned you're going to increase by 1 in Q1 and then probably a higher number through the rest of the year. What are these specific categories? If you can -- any thoughts on the sizing of these opportunities as well?

We're not providing any guidance yet on those. I think they're building more reactive. We have identified over 60 different cognitive categories that are available to us with engines in the marketplace for us to onboard, we're using the market as a primary driver for which ones we put on the platform. There is still engineering efforts when we add a new cognitive category, unlike when you are adding a new engine to an existing category, where the developer app really makes that seamless and engineering-free on our side. So we'll come back to you as we start to get better clarity in terms of what our customers are looking at.

And our next question comes from the line of Rob Stone with Cowen and Company.

Question for Pete to start out with. You talked about managing the burn rate, lowering in '18 versus '17. Can you give us any color on how we might expect the quarterly cadence of expense growth to lay out for the year? Is there a seasonal pattern there?

I wouldn't say it's necessarily a seasonal pattern there. I mean, the -- there's a balance there though between the revenues and the expenses. So -- but there isn't anything that's necessarily seasonal about that, Rob.

Okay. Your gross margins are holding up really well as you continue to grow the business and add accounts and categories. And you mentioned the developer kit, the fact that you're moving strategically towards the engines and even Conductor of being handled on a more automated basis. How should we think about the impact of that changing the technology on your gross profit per revenue dollar opportunity going forward?

I don't believe that, that will impact our gross profit at all, but it really will enable us to do is accelerate the number of applications and engines without having to increase our operational burn from an OpEx standpoint in terms of engineers and developer engineers.

Okay. In the press release, you note the recent acquisition and how this would help you add some cognitive categories. Can you provide any more specific color there?

Atigeo really was a data science acquisition as I think we discussed. They really bolstered 2 aspects of our team. One is, we picked up our Chief Data Scientist and his team up in Seattle, which are really starting at the far end of the spectrum in terms of broader general intelligent solutions and bringing them from that angle all the way back in. And then we also picked up some of the data scientists that are actually focused specifically on more real-time solutions for our customers. And that's really where their focus today at the shorter end of the spectrum in terms of delivering real-time value that will impact revenue in the next couple of quarters, which is moving. Today, what takes anywhere between 5 and 20 minutes for cognition on our platform to get that all the way to a sub-several-second-type time frame.

Great. Last question is for Chad. So you talked about signing up customers without an overly burdensome ask in terms of millions of dollars, improving the ROI and then expanding within the customer base. Do you have any sense by now if there's a pattern with existing accounts that you've worked with long enough? How long it takes the customer to get around that loop and realize that they want to go to the next larger scale or wider scope?

I still unfortunately think we're dealing with a lot of small numbers in terms of that, but we're seeing a pretty wide spectrum. As I think I mentioned with NBC, they started with an initial single point solution in the media side of their business. It took them less than, call it, 60 days to sort of identify that there were other opportunities to expand the technology and other business areas. They actually produced the analysis for us without our sales team doing that and then that led us to supporting them in that effort. There's other companies that have done that at the get-go. Even before we even signed them onto the platform, they have identified areas of use that are outside of our standard specifics. So again, I think there's a very wide spectrum. I can say that I'd say 80% plus of our customers are at least exploring, expanding beyond initial use cases with us in terms of using the developer applications to onboard new custom engines or even build applications to integrate directly with their current business processes.

So if that's a sort of thing -- just thinking about it logically, if they're expanding the use cases and engines and so forth, it seems like it should lead to, kind of, a step function in the run rate of revenue per customer. Is that -- and I'm not trying to pin you down to a specific target here, but just general -- directionally, do you expect that, that sort of step-up expanded application by existing customer is something that should become more apparent during the course of 2018?

Yes. I do. And I think I mentioned that in the call. Our goal is to show up to our customers today with aiWARE deliver a 2 to 3x type solution to them in literally near real time, let them measure that ROI, get comfortable with the platform and using artificial intelligence. And then again, migrating with them towards that [10x] opportunity. I fundamentally believe that for industry -- industries and businesses to compete in the future, they're going to have to readily adopt artificial intelligence to do so in all aspects of their business. To really being able to be that single source OS for them is a critical part of our strategy. Do I believe that we will start tackling more meaningful problems for our customers with higher revenue and better return for their shareholders? Absolutely, in 2018.

And at this time, this concludes our question-and-answer session. If your question was not taken, you may contact Veritone's Investor Relations team at VERI@liolios.com. And I'd now like to turn the call back over to Mr. Steelberg for his closing remarks.

Thank you for joining us on today's call. We want to thank our employees, partners and investors for supporting us as we pursue our mission to building the AI operating system of the future. We look forward to updating you on our progress on our next call. Operator?

Thank you for joining us today for Veritone's Fourth Quarter and Full Year 2017 Earnings Call. You may now disconnect, and everyone, have a great day.