In the third quarter of 2025, DoubleVerify grew revenue 11% year over year to $189 million, within its guidance range, while adjusted EBITDA of $66 million at a 35% margin exceeded expectations on AI-driven efficiency and cost discipline. Growth decelerated from Q2's 21% as the company lapped tough 2024 comps and absorbed softer retail spend, which weighed on activation (up 10%) and measurement (up 9%). Bright spots included a fast start for DV Authentic AdVantage on YouTube (~$8 million ACV in weeks), Meta activation scaling to 56 advertisers, supply-side growth of 27%, and the launch of an AI verification suite and Verified Streaming TV products. DV raised full-year margin guidance to approximately 33% but trimmed its Q4 outlook and full-year growth to about 14% on continued retail weakness.
Good afternoon, and welcome to DoubleVerify Q3 2025 earnings conference call. With us today are Mark Zagorski, CEO, and Nicola Allais, CFO. Today's press release and this call may contain forward-looking statements that are subject to inherent risks, uncertainties, and changes, and reflect our current expectations and the information currently available to us, and our actual results could differ materially. For more information, please refer to the risk factors and recent SEC filings, including our Form 10-Q and our annual report, or Form 10-K. In addition, our discussion today will include references to certain supplemental non-GAAP financial measures and should be considered in addition to and not as a substitute for GAAP results. Reconciliations to the most comparable GAAP measures are available in today's earnings press release, which is available on our Investor Relations website at ir.doubleverify.com.
Also, during the call today, we'll be referring to the slide deck posted on our website. With that, I'll turn it over to Mark.
Thanks, Tejal, and thank you all for joining us today. Q3 reflected disciplined execution and resilient performance across the business. Revenue grew 11% to $189 million within our guidance range, and adjusted EBITDA margin reached 35%, once again above expectations, demonstrating the scalability of our model. We're leveraging automation and AI to drive structural efficiency and profitability, proving DV's ability to deliver strong margins even in a dynamic ad market. During the quarter, market dynamics led to some retail budgets being softer, while growth in our other core verticals, including CPG, remained in line with expectations. Upsell momentum stayed strong, led by early demand for our AI-powered DV Authentic AdVantage solution, which closed roughly $8 million in annual contract value after only its first few weeks in market, fueled by early adoption from global CPG leaders.
We also maintained strong customer retention, with zero churn among our top 100 customers in Q3, underscoring the stability of our largest relationships. Core customer engagement and adoption rates remain healthy, and we continue to execute with discipline. At the same time, social and CTV are adding new growth and diversifying our revenue, strengthening the foundation for 2026. To frame the quarter simply, DV's growth drivers, AI-driven product innovation, margin expansion, and customer success remain firmly in our control, and on those levers, we continue to deliver. Today, I'll focus on three themes shaping our progress this quarter and beyond. First, innovation: how we're harnessing AI and automation to launch new products for the AI era, advance content classification, and drive greater efficiency at scale. Second, diversification: how growth across social, streaming TV, and programmatic is strengthening the durability of our model.
Third, monetization: how we're translating that innovation and diversification into sustained revenue growth, operating leverage, and cash flow. Each of these themes builds on the next, starting with innovation. At the center of innovation is AI, the engine behind our product development, precision, and scale. AI is driving the next major transformation in digital media, fundamentally changing how content is created, distributed, and consumed. Marketers, publishers, and AI agents themselves are beginning to design advertising strategies around this new layer of engagement, and DV is already embedded within it, capturing proprietary data that reveals how this ecosystem is taking shape. Each month, we analyze nearly 2 billion automated agents, crawlers, and bots, giving us unmatched visibility into how declared assistants like ChatGPT, Claude, and Perplexity, as well as undeclared or evasive bots and personal shopping agents, shape media performance.
These interactions represent an untapped opportunity for marketer-to-LLM engagement that DV is driving to enhance and monetize. To meet this moment, this week we launched the DV AI Verification offering, a group of tools built to empower advertisers in an AI-driven world. The suite includes DV's Agent ID Measurement, which, in its first iteration, identifies, measures, and classifies declared and evasive AI activity. It also features DV's AI SlopStopper, which detects and blocks synthetic or manipulated media across the programmatic open web, with expansion to social underway. Within Pinnacle, advertisers were able to view and act on this data in real time, quantifying AI impact and eliminating waste pre-bid. Powering the AI SlopStopper and our broader contextual classification capabilities is our Agentic Classification System, which uses generative AI to automatically build and retrain thousands of models using DV's proprietary data across programmatic and walled gardens.
Rolling out this technology will enable us to double our classification volume with fewer people and should achieve a four-fold gain in productivity per classification specialist by the end of 2026. It also lets us scale labeling volume by 260% and generate results 2,300 times faster than human labeling, all while maintaining human-level accuracy at lower cost. Bottom line, we're leveraging AI to not only innovate but also to expand margins, doing more, faster, with fewer resources, while simultaneously creating new monetization opportunities as AI agents play a larger role in digital advertising. Just as DV helped define transparency during the rise of programmatic, as well as the emergence of ad-supported CTV, we're now beginning to set the standard for trust and accountability in AI-powered media, positioning DV as the independent benchmark for verifying both human and AI-mediated engagement and content.
Moving to our next growth engine, diversification, our progress in AI-powered innovation is driving customer adoption in social and CTV. Beginning with social activation, both DV Authentic AdVantage and our Meta Pre-screen solutions are off to solid starts, underscoring the demand for transparent, performance-driven solutions in walled gardens. Social within activation is growing at 20% and remains one of our fastest-growing sectors. DV Authentic AdVantage, which launched on YouTube towards the end of September, is a DV-exclusive solution that unifies pre-bid brand suitability, Scibids AI optimization, and post-bid measurement into one seamless automated workflow. Early adoption has been strong, led by major CPG brands, including Kraft Heinz and Haleon. Much like our flagship Authentic Brand Suitability product, which was one of the most successful launches in DV's history, Authentic AdVantage delivers measurable ROI right out of the gate.
In early CPG tests, this solution delivered 24%-34% lower CPMs and 26%-50% higher impression volumes while maintaining or improving brand suitability. Continuing on social activation and turning to Meta, we significantly expanded content-level avoidance on Facebook and Instagram feeds and reels, nearly doubling our ability to filter out content on behalf of an advertiser's suitability preferences across categories and markets. Revenue from Meta activation solutions continues to outpace expectations, with 56 advertisers now live and in the early stages of scaling, up from 26 last quarter. 20 of our top 100 customers now leverage our Meta activation solution, up from 13 in Q2, and usage is beginning to ramp. Today, our pre-bid solution is attached to roughly 6% of our brand-suitable measurement impressions on Meta, representing an upsell opportunity we expect to rise meaningfully as adoption deepens.
On TikTok, we expanded our video exclusion list by 100 times, significantly enhancing advertisers' ability to proactively avoid unsuitable content and reducing their rate of unsuitable content by one-third. Together, these advancements strengthen pre-screen protection on the world's largest social and video platforms and demonstrate our partners' commitment to giving advertisers the tools they need to safeguard brand equity and improve contextual relevance at scale while still driving performance. There's been some debate about whether platform-native AI optimization tools, those black box tools that automate targeting, creative, and attribution, could reduce the need for independent verification. The reality is, while those systems optimize delivery, they don't disclose where ads run or how suitability is maintained. In a sample of AI-run social campaigns, we found brand suitability rates to be roughly two points lower than in non-AI campaigns.
As these closed algorithms scale, advertisers are relying even more on DV for the transparency control that platforms don't provide. In our sample, our pre-bid protection was applied more than three times as often on AI campaigns than on standard campaigns, evidence that advertisers see higher risk in these black box solutions and a greater need for safeguards. The takeaway is clear: as platform AI engines become more sophisticated, the need for an independent, trusted verification becomes even more essential to ensure performance, suitability, and accountability work together. Turning to social measurement, we continue to expand post-bid coverage across the world's largest social media environments, expanding our AI-powered brand suitability measurement to Meta Threads, giving advertisers independent transparency on yet another fast-growing social media platform.
We also extended our brand suitability measurement on Snapchat to shows and publisher stories, adding to our existing coverage of Creator Stories and Spotlight, and giving advertisers greater clarity across more premium inventory. Shifting to diversifying revenue through CTV growth, advertisers continue to describe the streaming landscape as fragmented and opaque. They often do not know where their ads run, the quality of the content they appear in, or even if those ads are viewable and paid attention to by a real human. In some cases, ads are intended for premium full-episode TV experiences and end up in mobile gaming apps like Solitaire or other non-TV environments. This is a problem we estimate impacts roughly 15% of CTV impressions and wastes over $1 billion of media spend each quarter, eroding trust as well as ROI.
At the same time, advertisers still rely on manual, time-consuming, and error-prone workflows to manage do-not-air brand suitability lists, leading to misplaced ads and missed optimizations at scale. We've said before that DV has not fully monetized its CTV exposure, and we're now addressing that opportunity head-on with three streaming TV-specific product launches this quarter and with more to come in 2026. On the measurement side, this week we announced the launch of DV Verified Streaming TV Measurement, a market-first capability that provides impression-level transparency across digital video campaigns, helping advertisers ensure that ads are delivered in high-quality, TV-like environments, not in outstream players on blog pages or in gaming apps, which too often pass as TV inventory in reseller channels in the open market and private marketplace.
We're also extending our Verified Streaming TV capabilities into activation, launching pre-bid Verified Streaming TV segments across leading programmatic platforms such as The Trade Desk, Teads, StackAdapt, Microsoft Curate, and Index Exchange, allowing advertisers to target authentic streaming inventory in open market and PMP buys and avoid wasted delivery to rogue environments. Additionally, in activation, we've launched pre-bid do-not-air lists for streaming TV within AVS, modernizing what was once a manual spreadsheet-based process into one that automatically enforces brand compliance policies across streaming platforms at scale. Finally, we announced a new deal with entertainment database IMDb, leveraging authoritative metadata and popularity insights licensed from IMDb to enhance show-level transparency and classification for streaming TV. This partnership will help fuel agentic streaming TV contextual solutions that we'll be launching in early 2026.
Thank you, Mark, and good morning, everyone. Our third-quarter results reflect continued double-digit year-over-year revenue growth, solid profitability, and strong cash generation. We delivered approximately $189 million in total revenue in the third quarter, up 11% year-over-year and within our guidance range.
Adjusted EBITDA was $66 million, representing a 35% margin and above the high end of our guidance range, driven by cost discipline, operating leverage, and AI-driven efficiency gains across the organization. As we outlined last quarter, Q3 revenue was essentially flat on a sequential basis, driven primarily by tougher year-over-year comps as we lapped our strongest quarter of 2024 and further driven by softer retail spend. We expected second-half revenue growth to moderate, consistent with our full-year outlook for double-digit revenue growth, strong profitability, and the scaling of new activation and measurement products focused on social, CTV, and AI heading into 2026. Last quarter, we noted that approximately a third of our first-half 19% year-over-year revenue growth came from new advertisers, with Moat wins from last year contributing approximately one percentage point.
Through the first nine months of 2025, revenue increased 16% year-over-year with a similar contribution pattern to the first half. Approximately a third of revenue growth came from new advertisers, with moat wins from last year contributing approximately one percentage point. The majority of our growth continues to come from existing customers expanding their use of DV solutions. In the third quarter, total advertiser revenue grew 10%, driven by increased volumes. Media transactions measured, or MTMs, increased 12% year-over-year, while measured transaction fees, or MTFs, decreased 4% year-over-year, reflecting product and geographic mix and excluding the impact of one introductory fixed-fee deal. Activation revenue grew 10% year-over-year in the third quarter. ABS, which accounted for 54% of activation revenue, grew 12% year-over-year, driven by new logo activations, upsell to existing customers, and expanded usage among current users.
73% of our top 500 customers have now activated ABS, up from 68% in Q3 last year, demonstrating the continued adoption of this premium product. Non-ABS activation revenue grew 8%, reflecting solid demand for social activation solutions, partially upset by softer spend from retail advertisers. Measurement revenue grew 9% year-over-year with momentum in social, partly upset by weaker retail spend. Social measurement grew 9% and accounted for 48% of total measurement revenue, while international revenue grew 2% and accounted for 27% of total measurement revenue. Excluding the suspension of the one CPG customer at the start of the year, social measurement revenue would have grown 22% in Q3 and 21% year-to-date. Revenue from Rockerbox was in line with expectations and is on track to achieve our expected 2025 revenue contribution of approximately $8 million.
Finally, supply-side revenue grew 27% in the third quarter, driven by growth on existing platforms and new platform and publisher partnerships. Moving to expenses, cost of revenue increased 14%, primarily due to growth in activation revenue, which carries increased partner costs tied to revenue-sharing arrangements, as well as higher data and hosting costs driven by increased usage. In Q3, we delivered an 82% margin on revenue-less cost of sales, and we expect to maintain margins between 80%-82% in Q4. R&D expenses increased as we continue to invest in AI capabilities, engineering talent, and product development, including the integration of Rockerbox and continued improvement of DV Authentic AdVantage. Sales and marketing expenses and G&A included costs related to the Rockerbox acquisition and other strategic initiatives. As noted last quarter, hiring remains disciplined as we realign resources towards growth priorities and continue to optimize for efficiencies.
Adjusted EBITDA of approximately $66 million in the third quarter represented a 35% margin, exceeding expectations driven by cost discipline, operating leverage, and AI-driven efficiency gains across the organization. GAAP net income reflected the impact of higher tax expenses, which is largely driven by the tax impact of our lower share price and by higher stock-based compensation costs. Looking ahead to 2026, we're implementing an updated equity incentive plan that is projected to reduce annual stock-based compensation costs by 20%. This quarter, we also introduced an adjusted EPS calculation to provide an additional metric to evaluate the business. We delivered net cash from operations of approximately $51 million in the quarter. Capital expenditures were approximately $12 million in the quarter as compared to approximately $6 million in the same quarter last year, as we accelerate investments in new solutions across social, streaming TV, and AI.
In terms of capital allocation, in the third quarter, we repurchased 3.3 million shares of DV Common Stock for $50 million. As of November 7, $90 million remained available and authorized for additional repurchases. Through September 30, we deployed $132 million to repurchase 8.4 million shares, more than offsetting the anticipated full-year 2025 stock-based compensation costs. We also deployed $82 million net of cash to acquire Rockerbox as part of our M&A strategy to diversify our product offering from protection to performance. In addition to investing into the business, we will continue to evaluate M&A opportunities and buybacks, including beyond the current authorization, as part of our capital allocation strategy to maximize shareholder value. In the first nine months of 2025, we delivered net cash from operations of approximately $138 million compared to approximately $122 million in the same period last year.
Capital expenditures in the first nine months of 2025 were approximately $28 million compared to approximately $20 million in the same period last year. In the first nine months of 2025, cash generated from operations after funding capital expenditures totaled approximately $110 million as compared to adjusted EBITDA of $168 million. We ended the third quarter with approximately $201 million in cash and cash equivalents. Our strong cash generation, combined with disciplined capital allocation and share repurchases, continues to enhance long-term per-share value. Turning to guidance, we're updating our fourth-quarter outlook to reflect ongoing retail softness in a key seasonal period. We expect revenue to range between $207 million-$211 million, representing 10% growth at the midpoint. We expect adjusted EBITDA to range between $77 million-$81 million, representing a 38% margin at the midpoint and continued strong operating leverage.
While Q4 is our easiest comparison for existing customer growth, it is also our toughest for new customer growth as we lap a period of outsized advertiser, publisher, and platform additions. For full-year 2025, we expect to deliver approximately 14% year-over-year growth at the midpoint and are raising our adjusted EBITDA margin guidance from approximately 32% to approximately 33%, reflecting margin expansion even as revenue growth normalizes to approximately 10% in the back half of the year. We also expect our full-year 2025 margins of approximately 33% to be a base case for full-year 2026, supported by continued cost discipline, AI-driven efficiency gains, and the inherent operating leverage in our model. For the fourth quarter, we expect stock-based compensation to range between $25 million-$28 million and weighted average diluted shares outstanding to range between 163 million-165 million shares.
Looking beyond 2025, upside from the 10% base case revenue growth we're expecting for the second half of 2025 will be driven by the pace of adoption and scaling of our social activation products, our CTV solutions, alongside the ramp of our new AI offerings. As Mark mentioned, our medium-term goal is to grow social, streaming CTV, and AI verification solutions from less than 30% of total revenue today to approximately 50% while continuing to expand our other key sectors. This evolution will create a more diversified and resilient growth profile and position DV to capture a larger share of the fast-growing digital advertising ecosystem. In closing, our results show consistent double-digit growth, disciplined operational execution, and strong profitability. Our balance sheet remains robust with over $200 million in cash and no long-term debt, supporting innovation, strategic partnerships, and share repurchases.
DV's business model continues to demonstrate resilience and scalability, and we remain confident in our ability to create long-term value for our shareholders. With that, we will open the line for questions. Operator, please go ahead.