Good day. Welcome to the Outbrain Second Quarter 2023 Earnings Conference Call. All participants will be in the listen-only mode. Should you need assistance, please signal the conference specialist by pressing the star key followed by zero. After today's presentation, there will be an opportunity to ask questions. To ask a question, you may press star, then one on a touchtone phone. To withdraw your question, please press star, then two. Please note, this event is being recorded. I would now like to turn the conference over to the management team. Please go ahead.
Good morning. Thank you for joining us on today's conference call to discuss Outbrain's second quarter 2023 results. Joining me on the call today, we have Outbrain's co-founder and co-CEO, Yaron Galai, co-CEO, David Kostman , and CFO, Jason Kiviat. During this conference call, management will make forward-looking statements based on current expectations and assumptions. These statements are subject to risks and uncertainties and may cause actual results to differ materially from our forward-looking statements. These risk factors are discussed in detail in our Form 10-K, filed for the year ended December 31, 2022, as updated in our Form 10-Q and other reports, and in subsequent reports filed with the Securities and Exchange Commission. Forward-looking statements speak only as of the call's original date. We do not undertake any duty to update any such statements. Today's presentation also includes references to non-GAAP financial measures.
You should refer to the information contained in the company's second quarter earnings release for definitional information and reconciliations of non-GAAP measures to the comparable GAAP financial measures. Our earnings release can be found on our IR website, investors.outbrain.com, under News and Events. With that, let me turn the call over to David.
Thank you, Imelda. Thank you all for joining us. I'm excited to share with you our financial results and the significant progress we are making across multiple strategic fronts. As you will hear, we are innovating at a great pace, leveraging AI and other technologies. We are pleased to report a solid second quarter, in which we achieved $54.6 million in Ex-TAC gross profit, representing 5% sequential quarterly growth and reaching the high end of our guidance. Adjusted EBITDA of $3.5 million, exceeding the high end of our guidance. In what is still an uncertain but relatively stable macro environment, we continued to focus on driving growth and better performance within our current marketplace while maintaining tight cost controls.
B, making strategic investments, growing our addressable market, both on the advertiser side and publisher side, through a product and technology-led strategy. I want to start with Onyx. Last quarter marked an exciting milestone for our company with the launch of Onyx by Outbrain. Onyx is a new brand-building platform focused on driving high attention from video and high-impact, rich display ads for premium enterprise brands. With our core performance platform, and now with the launch of Onyx, we are proud to be one of the very few advertising platforms that can offer true, full funnel capabilities to advertisers at a global scale on the open web, from building brand awareness and consideration all the way to customer acquisition. Onyx is expected to deliver incremental value to us through premium brand campaigns carrying high CPM, which will be delivered outside of our traditional feed.
This launch means we expect to do a lot more with existing and new customers. It increases our total accessible market by an estimated factor of 3x . This is according to Gartner data that breaks down budget allocation between performance, marketing, brand building, consideration, and loyalty. What is Onyx? What makes it unique? Onyx is a brand-building platform for enterprise brands and agencies. It is built to deliver strong ROAS from video and high-impact display campaigns. While most of the advertising market is focused on ad viewability and video completed views, Onyx goes one step further to maximize user attention, which has been proven to drive business impact. With marketers focusing more and more on outcome and ROAS, attention is gaining momentum as a much smarter success KPI for advertisers.
We partnered with Adelaide, a leader in attention measurement, to allow Onyx to capture Attention Units in real time and use predictive AI to find moments and opportunities to maximize customer attention for every campaign. Yaron will elaborate later on the technology and how we leverage our powerful prediction capabilities developed over 15 years. We launched Onyx in mid-June, and we're off to a great start with more than 25 brands already live or committed to testing. One of the first campaigns we tested was in partnership with Xaxis U.K. for Ford. We used a custom Onyx ad experience called Hybrid that combines both video and display assets into an interactive ad experience. The results exceeded our customers' expectations, with Adelaide putting the campaign attention score 30% higher than the benchmark. We continue to experience higher performance than the Adelaide benchmark on our Onyx campaigns.
This is a powerful testimony for how Onyx is successfully winning user attention and, as a result, driving stronger brand impact for advertisers. Onyx is now available for enterprise brands and agencies in the U.S., U.K., Germany, France, and Italy, and will be launched in other markets later this year. We've run Onyx campaigns in these markets for premium brands like Visa, Porsche, Mattel, iHeartRadio, Opel, Alfa Romeo, and others. Our pipeline is building, and we expect to generate double-digit millions of dollars of revenue from Onyx already in H2 of this year. Just as a side comment, while Onyx is great for brand advertisers, it's also extremely strategic for our relationship with publishers, helping us elevate quality and user experience, which has been one of our key differentiators and factors in winning premium supply deals. Ultimately, it helps deliver even more revenue to our publisher partners.
Moving to the general marketplace advertising results. On the revenue side, we are seeing stabilization in the market, with positive signs of growth in the last two weeks of the quarter and sequential growth over the course of Q2 and stronger start in Q3. From a vertical perspective, we saw year-over-year growth in auto, health, and retail. In addition to the continued innovations we make for conversion bid strategy, one interesting area of progress in Q2 was the increased adoption of Zemanta by our core advertisers. As a reminder, Zemanta is our in-house performance DSP. Unlike traditional DSPs that focus on streamlining media buying for display and video on a CPM basis, Zemanta is connected to most major native advertising SSPs, it allows marketers to run performance-based campaigns across the entire open web.
If this trend with Zemanta continues, we expect this to lead to increased share of wallet with many of our performance advertisers. Moving to the publisher side, where we entered into several new exclusive long-term partnerships in Q2, including TMZ, Washington Times, The Messenger, and others. On the device or platform side, we established a partnership with Disqus, a commenting platform, and started ramping up our placement in Samsung devices through our partnership with upday. With existing partners, I want to highlight our success in renewing partnerships to secure our long-term business growth with the recent renewals of multi-year deals with New York Post in the U.S., Bauer in Germany, CCM Benchmark in France, and Hearst, Italy.
Overall, we're very excited about the launch of our strategic branding platform, Onyx, that we believe will further strengthen our position in the open web as the quality partner for premium publishers and the full funnel open web partner for all types of advertisers. We're encouraged with the 5% sequential growth in Ex-TAC gross profit in Q2 and our profitability, and expect significant acceleration in growth rates in the coming quarters, both in Ex-TAC gross profit and Adjusted EBITDA. I'll now hand it over to Yaron.
Thanks, David. Today, I'll cover 3 areas: an update on Keystone, an AI update, and more detail on what we're doing on Onyx as it relates to attention. First, Keystone. We had a strong quarter of new publisher launches. Since our last earnings call, we went live with Keystone on CNN, Axel Springer's Bild, The Arena Group's TheStreet, and Sankei from Japan, among others. These publishers are advancing various KPIs through the Keystone technology, such as subscriptions, e-commerce, retail, and newsletters. One of the main drivers for this growth has been an important product upgrade from our R&D team. Deploying the Keystone code and integrating it with publisher systems has been a barrier to quickly launching with interested publishers. In Q2, we built and launched a new no-code solution for Keystone, which leverages the unique code-on-page position we already have with publishers.
At the new launches, publishers have been loving this. My second topic is AI. On the last quarterly call, I updated that we built ChatGPT into our advertiser product called Amplify. This follows an automated AI headline generator that we built in-house several years ago. At the time of our last call, I mentioned that about 50% of the automated AI headlines that advertisers were using were then originating from our ChatGPT integration. That was a long three months ago, which is decades in AI terms. Our successful integration and launch since then has now grown to be 100% of the automated headlines that we suggest to advertisers within our product, which are now coming from ChatGPT.
In addition to the speed, scalability, and cost efficiencies resulting from these generative AI capabilities, we're also seeing that the ChatGPT headlines adopted by advertisers are yielding, on average, a 7.5% higher click-through rate than the headlines they've created manually within their campaigns. Better ROAS for our advertisers typically leads to an increase in advertising dollars spent with us. Our initial Gen AI integration was limited to the English language. Over the last quarter, we've added within our advertiser product support for generative AI headlines in Spanish, Portuguese, Japanese, French, Italian, and Dutch. In parallel to these upgrades to our advertiser product, I'm also happy to update that we've now integrated the ChatGPT Gen AI capabilities into Keystone, allowing publishers to quickly scale and better personalize their various business offerings to their own audiences.
We've also been busy deploying AI capabilities into our R&D and business operations. As examples, we've deployed the use of Microsoft's GitHub Copilot, which is an AI assistant for coding, across our engineering teams. We've built our own AI bot based on ChatGPT for automatic code reviews and flagging of issues relating to security and privacy compliance within our code. Moving from generative AI to our ad-serving algorithms, which have been AI-based for several years. During the first six months of the year, we've deployed new and improved AI algorithms, which have so far resulted in a 4.3% improvement in CTR potential based on our internal A/B testing.
As you can see, our product and engineering teams are racing to build and deploy these new AI capabilities across our algorithms, our products, and our operations to ensure that Outbrain is at the forefront of the AI revolution. Switching gears to Onyx and attention, I believe that the world of brand advertising on the web has evolved in three main eras. The first was the reach era. During this period, advertisers were focused mostly on buying ad impressions, regardless of how and where those ad impressions happened. To use TV advertising as an analogy, brands were paying for their ads, whether the TV set was turned on or off. The second was the viewability era. During this past decade or so, brands got smarter and no longer wanted to pay for ads that were never even viewable by a human being.
They started demanding viewability as the major metric they bought advertising against. To jump back to the TV advertising analogy, viewability ensures that the TV set is on when the ads are shown, but it doesn't ensure that anyone is actually paying attention to them. Bathroom or beer runs during the ad break will still count as 100% viewable, and the advertiser pays for the full cost. We're now at the very beginning of the third big era of brand advertising on the web, and that is the era of attention. Brand advertisers now want to make sure that their dollars aren't just theoretically viewable, but rather that a human being is paying attention to them. Therefore, in the coming months, we're all gonna be hearing the word attention come up often by many companies in the space.
Now, I wanna explain why Outbrain, with its 15-year heritage of powering newsfeeds for the world's best publishers, has such a strong and differentiated position in this era of attention. Most companies in online advertising are thinking of attention the way they did about viewability. They're typically looking at an ad placement and deciding in a binary way whether that ad placement is viewable or not, or whether it's high attention or not. They're placement-based systems and not user-based systems. Now, Outbrain's huge advantage is that we deeply understand the user's interests of the roughly 1 billion people we serve each month. Through the newsfeeds we power exclusively with code on page, our technology understands what links and ads are likely to be the most engaging to each individual user.
Through this deep understanding of users' interest graphs, we can predict on a user-by-user basis which specific ads will generate attention. The shift of brand budgets to attention will likely be significant amongst all players in the space, but we strongly believe that the biggest winners won't be those that make binary decisions on a placement level, but rather those that can make user-by-user predictions as to which brand ad will generate the highest levels of attention by each user and audience. Our 15-year head start in understanding users' interests and personalization, combined with this industry shift to the era of attention, is the reason we're so excited with the timing of introducing Onyx. With that, I'll hand it over to Jason to cover our financials.
Thanks, Yoram. As David mentioned, we beat our Q2 guidance for Adjusted EBITDA and achieved the high end of the guidance range for Ex-TAC gross profit. From a demand perspective, we experienced a softer start to Q2 with a stronger relative performance in June, where we started to see days of year-over-year revenue growth in the back half of the month. The early portion of Q3 has continued this trend from June, and we're cautiously optimistic about the outlook for the back half of the year. Geographically, Europe has shown stronger signals of demand recovery than the US. Revenue in Q2 was approximately $226 million, a decrease of 10% year-over-year. New media partners in the quarter contributed 12 percentage points, or approximately $30 million of revenue growth year-over-year.
Net revenue retention of our publishers was 78%, reflecting the lapping of the strong level of ad impressions in the prior period, driven by several factors, including the prior period having heightened traffic around the war in Ukraine, as well as being more selective and effective in supply we bid into, as we focus on premium, high-quality supply, driving the best results for our advertisers and on optimizing serving costs. Additionally, we continued to see a headwind from the impact of the demand environment on yields. Our churn remains low by our standards, and our 3 largest churns year-over-year contributed just approximately 4 total points of net revenue retention headwind in Q2. As another data point, our logo retention remained around 95% for all partners that generated at least $10,000.
Ex-TAC gross profit was $54.6 million, a decrease of 8% year-over-year, driven primarily by the revenue decline. A slightly lower year-over-year decline in Ex-TAC gross profit as compared with revenue is the result of largely offsetting impacts of revenue mix changes and better performance on certain media partners, which led to an improvement in margin. For both revenue and Ex-TAC gross profit, currency fluctuations did not have a material impact on year-over-year growth rates for the quarter. Moving to expenses. Operating expenses remained essentially flat year-over-year, increasing slightly to $51.7 million in the quarter as we continued to exercise discipline around spending. There were several offsetting factors driving the flat year-over-year cost.
To adjust to the continued macroeconomic uncertainty, create additional operating efficiencies, and support the company's strategic growth and profitability objectives, we announced at the end of May a reduction in our workforce of approximately 10%. Included in Q2's operating expenses are $2.3 million of severance and related costs related to this cost reduction initiative. Absent this, personnel-related costs declined $3.8 million year-over-year, driven by headcount reduction and FX favorability. For non-comp, we saw an increase in bad debt expense of about $1 million year-over-year, which was driven by several customer bankruptcies, particularly in the programmatic advertising space. We are monitoring closely to mitigate the risk of further losses. As a result, Adjusted EBITDA was approximately $3.5 million in Q2. Moving to liquidity.
Free cash flow, which, as a reminder, we define as cash from operating activities, less CapEx and capitalized software costs, was a net use of cash in the quarter of approximately $2 million. The small net use of cash was driven by the timing of working capital. As noted last quarter, in April, we repurchased $118 million aggregate principal amount of the convertible notes for approximately $96.2 million in cash, including accrued interest, representing a discount of approximately 19% to the principal amount of the repurchased note. We viewed the chance to repurchase a portion of the debt at a considerable discount to be opportunistic, given the strength of our balance sheet, with remaining cash balance that retains the optionality to invest in opportunities that can drive further shareholder value.
We ended the quarter with $218 million of cash, cash equivalents, and investments in marketable securities on the balance sheet and $118 million of long-term convertible debt. In December, the company's board of directors authorized a $30 million share repurchase program, incremental to the $30 million program fully executed in 2022. We began executing the new program, purchasing 1.5 million shares for about $7.1 million year-to-date through June 30th. We continue to believe it's an attractive way to enhance shareholder value under current market conditions. Turning to our outlook. In our guidance, we assume a continuation of the trends we have seen in the first weeks of Q3, with seasonal increases in ad spend that we typically see to finish the year, and additional growth from bringing Onyx to market.
Despite a lower level of visibility into advertising budgets, we have incremental visibility to our EBITDA plan based on the cost actions we have taken earlier this year. With that context, we have provided the following guidance: For Q3, we expect Ex-TAC gross profit of $56.5 million-$59.5 million, and we expect adjusted EBITDA of $7.5 million-$9.5 million. We maintain our previous full year 2023 guidance of at least $237 million of Ex-TAC gross profit and are increasing our previous full year guidance for adjusted EBITDA to at least $30 million. I'll turn it back to the operator for Q&A.
We will now begin the question-and-answer session. To ask a question, you may press star, then one on your touchtone phone. If you are using a speakerphone, please pick up your handset before pressing the keys. If at any time your question has been addressed and you would like to withdraw your question, please press star then two. At this time, we will pause momentarily to assemble our roster. The first question comes from Shweta Khajuria with Evercore ISI. Please go ahead.
Thanks, thanks for taking my questions. Let me try two, please. For Onyx, could you please explain exa- how the product works? You, you mentioned double-digit million dollars in revenue contribution in the back half. That's great. In terms of the mechanics of the product, so where are the ads placed? Why, why do they attract higher CPMs? What would be a good example in terms of why you are driving, you are expanding into a higher TAM? That would be great. The other question I had was for Jason. In terms of headcount reduction, are you, are you good with the headcount you have today, and what are you thinking for the remainder of the year? Thank you.
Hey, Shweta, it's David. Thanks for your question. Onyx, in terms of financials, we expect it to do between $10 million-$20 million in the second half of the year. The product is a brand awareness and consideration product targeted at enterprise brands, and it's basically placed either in mid-article or end-of-article placements. In a 100% share of voice, it could be from an experience, could be video, and it could be high impact display or a combination of the two. For example, we mentioned on the Ford example, it was a hybrid, hybrid ad where we combined video and display. These are normally CPMs that are double digits for those brand awareness campaigns, so that's, you know, expected to drive higher CPMs in the network. We're very excited about it.
It really grows our offering to a full funnel partner to brands. We're leveraging many existing brand relationships that we have with customers like
... Disney, for example, or others that use us for performance. Now we also have a dialogue with them around brand awareness campaigns, so we're leveraging that. It also really opens the door for us to have much more strategic dialogues with the big holding companies, the big agencies, and enterprise brands that didn't work before with us because they were mostly interested in, in building brand awareness and, and consideration. Very excited about it, both from a product strategy and, and potential financial impact.
Hey, Trevor, I'll take the second one. On the headcount reduction, obviously, you know, we felt it was prudent in this environment to, you know, exercise, you know, discipline around costs as, as we have in the past and improve our cost structure going forward. You know, what we announced was about a 10% workforce reduction and really focused on efficiencies around, you know, having our teams, automation and focusing on the priority growth areas like Onyx, of course. So you know, what we effectively, you know, did with that is we, we, we reduced our annualized cost by about $12 million-$14 million, and so about half of that we expect to benefit from this year in full year 2023. You know, that, that, that brings our headcount now to about 870 people.
You know, we are, we are, you know, hiring, you know, obviously for critical positions now, and, and obviously, you know, some of the strategic positions around Onyx is, is some of our priority hiring right now. You know, I feel like we're back to business as usual following the reduction now.
Thanks, David. Thanks, Jason.
Thank you. The next question comes from Laura Martin with Needham. Please go ahead.
Good morning. Let's start with the bad debt number, Jason. $1 million, I assume it was MediaMath. Could you talk about what you think exposure, additional exposure might be to the bad debt number?
Sure. Yeah. You know, we, we, we did see a higher bad debt expense in, in, in Q1 as well. Maybe I'll talk about it with respect to the full, the full first half of the year. Our total bad debt expense was, you know, between $4 million and $5 million, which for us is, is basically a full year number. Exceptionally high in this, in this first half of the year. You know, really a few drivers, obviously, the, the, the kind of macro pressure on, on, on, on many in the, you know, advertising and programmatic, businesses is, is the main driver there. You know, we did see, you know, a few large bankruptcies driving, you know, close to half of that bad debt expense in, in the first half of the year.
Obviously, there's a few, you know, big programmatic companies, you know, MediaMath and Sizmek are the, the highly publicized ones. You know, the two of those alone accounted for, I think, $1 million or so of our of our bad debts in, in the quarter and in the, in the first half. You know, of the remaining amounts, you know, we don't think there's, there's, there's too, too big of an exposure. Actually, we think, you know, a, a, a good chunk of it is really just aging, and a good chunk of that aging is, is, you know, the, the big agency holdcos, who are just, you know, slow payers. You know, part of it is that they might be collecting slower even from, from their brands.
You know, we're very formulaic, obviously, with our, with our bad debt expense and the aging. You know, we see opportunity to, to reverse some of that expense in the second half of the year. Yeah, I mean, I think in this environment, we, we should expect to have a, you know, higher than normal bad debt expense.
Okay. great, thanks. Then Keystone versus Onyx, are they complementary, or do they attack, or do they go after different supply side capabilities? I know they both deepen your relationship. One is more full funnel, and the other one is sort of a deeper relationship with suppliers. Are you finding the same supply-- the supply side guys are interested in both products or are different, you know, supply side guys interested in these two products? Do they work together or totally independently, these products?
Hi, Laura. Okay.
Okay, David, take it.
I'll start. I mean, they're separate products. They're very complementary. One is totally focused on the advertiser side, which is Onyx, where it's more of a full funnel solution for advertisers, totally new markets for us, that it increases our TAM by 3x. Keystone is a supply publisher-focused product that is a business optimization platform for publishers. Two different products. Yaron, anything else?
Just that they obviously, they both come from the same R&D shop, and we obviously look at at synergies and how they can work well together for the benefit of publishers.
Thank you.
Generally, maybe one, one more comment. I mean, both of them are elevating the strategic relationships we have on the two sides of the marketplace. Keystone is, is a very strong strategic tool for publishers looking to diversify revenues on the advertiser side, existing and new. We're becoming a much broader partner. I think companies generally are looking to work with less partners. This positions us in a pretty unique position in the market on the two sides of being able to provide a much broader offering, which is strategic on, on the two sides.
Thank you. The next question is from Ross Sandler with Barclays. Please go ahead.
Hey, guys, just a couple questions. Can you talk about the environment? Sounds like you start to see positive growth rates before the beginning of three Q. you know, what geos, what, what categories are, are driving that, that strength? you know, as we look at, you know, you said $10 million-$20 million, which is, is a, is a good start for Onyx, but if it is truly incremental-
... you know, I would guess if it's incremental budget, I would guess that the, the ramp into 2024 would be more than just like a low single-digit incremental bump to revenue. How do you think about that product scaling as we get into 2024? Then the last question, the Ex-TAC gross margin finally turned back up for the first time in two years. How are we thinking about that? Are we through the kind of worst of the guarantees and macro headwinds around Ex-TAC margin? Thanks a lot.
Thanks, Ross. I'll start with the with the demand trends, and I'll be brief. Yeah, what we, what we saw obviously in Q1 was sequential improvement each month of the quarter in terms of demand, while Q2 was much more mixed. You know, April was mixed, softer May, and stronger June. Q3, off to a better than expected start, you know, in July relative to June. Yeah, relatively stronger trends, you know, particularly in Europe versus U.S. Obviously, the strengthening, you know, euro and the pound versus the dollar has helped there. You know, CPCs, they're still down year-over-year in Q2. Obviously, as we lap easier comps into Q3, you know, we can see an inversion there.
Just maybe anecdotally, we did see an improvement in yields sequentially, Q2 versus Q1. So, again, positive trends, and, and, and really reasons for optimism, going forward. Verticals-wise, you know, we're not overly diverse. You know, we're, we're pretty diversified, and we're not overly concentrated, so auto, health, and retail were stronger. Finance has been weak and remains weaker than other verticals for us.
Maybe, hey, Ross, I'll start with the margin. We saw an improvement of about, I think 200 basis points versus Q1 and about 50 basis points versus last year. We are encouraged by that. We expect, you know, that the yields will not go, that the margin will not go further down, and hope to get it back to prior levels. I think to get it to historical, very high levels, I think it will be dependent on some market recovery, but overall, we are sort of excited about this trend and definitely through the second half of the year. Regarding Onyx, so we are... We're off to a great start. I mean, what we're getting right now is test budgets.
We're looking into budgets of Q4, which are more significant. Definitely next year, our expectation is that we develop really strategic relationships with many of these holding companies and many of these brands, and then you're going into a different order of magnitude of brand dollars and campaigns. I mean, we're not giving a forecast for 2024, but, we expect that if, you know, we hit the numbers for H2, a significant growth of it into 2024 and in coming years. Again, our expectation is it becomes, you know, a few hundred million dollar business over the next few years.
Thank you. The next question comes from Andrew Boone with JMP Securities. Please go ahead.
Good morning, thanks for taking my questions. I wanted to ask about tools to improve yield. How are you guys thinking about key drivers as we get to the back half of this year and into 2024, on key drivers that can step up yield? Secondly, just quantifying that is, as I think about net publisher retention, historically it's been, you know, 108, 110, kind of in that level. Is there a path to get back there? Do you guys think that you guys can recover to those levels kind of as macro stabilizes? Thanks so much.
Andrew, maybe I'll, I'll take the the yield question around here. Yield is, is kind of the end result of everything. It's it starts obviously with the the technology, and we keep investing in the technology and the algorithms specifically. I mentioned in the comments earlier, we've just in the algorithm updates we've done in the first half of the year, we've seen a potential increase of over 4% in click-through rate. That's that's deployed network-wide, and those, those changes compound over time. The the next level is obviously the the level of data and the quality of that. On top of that, it's the the ad base. Into that part of the yield go a few things.
One is just getting the advertisers selling, macro has to do with that. Obviously, we're excited there, and Onyx I think is a great leg with selling new types of advertisers and new campaigns. The other side is what they run with us. The more kind of algorithm food the algorithm has, the better the results are. That was the example I gave of the AI capabilities that we've built into our Amplify products, which take an advertiser campaign. Once we've sold it, they take the advertiser campaign, and through AI, they recommend a variety of new headlines and variations, which when the advertisers accepted those, what we've seen is more than a 7% improvement in click-through rate, which translates to yield.
The, the yield result, I'd say, is kind of the, the, the result of all the activities we do operationally, business, technology, and, and algorithm-wide.
On the retention, I'll take that 1, Andrew. So just, yeah, I mean, obviously, you know, we've said the last couple of quarters that, you know, the biggest headwind had been the demand and yield, and that remains a headwind. Just to note, we did, you know, grow Ex-TAC sequentially, about 5% from Q1 to Q2, despite revenue kind of continuing to see these headwinds. One thing, you know, we noted, you know, on the call this time was just ad impressions, you know, were year-over-year headwind in Q2. Several factors, no material difference in churn, still at, you know, we feel pretty, pretty low levels for us.
The lapping of the war in Ukraine, well, particularly in Europe, there was pretty high page views last year in Q1 and Q2 from that, was a headwind. We've also just, you know, we made changes and we continue to make changes in our supply bidding to optimize for market ROAS and quality and serving costs. You know, that might have a negative impact on revenue and ad impressions, but we're driving higher, we think, tech and profitability because of it, that's the goal. You know, I think in half two, as the comps ease a little bit, we'll expect to get to a much more normal split between retention and new.
That could be around 100 in NRR in Q3, and, you know, even higher growth. That's the, that's the plan for H2.
Thank you.
Thank you. The next question is from Ygal Arounian with Citi. Please go ahead.
Hey, guys. Good morning. You have Max More on for Ygal. I guess my first question would be on the converts you bought back. Just wondering what's the plan for the rest of the converts still on the balance sheet? Just more generally, how we should think about capital allocation going forward. Then my second one would be, you know, on generative AI. You know, you gave a nice update on the ChatGPT integration, but just, I guess, thinking more down the road, how should we think about more Gen AI integrations from here, where you think it could impact and, you know, over the next couple of years, how you think it could impact the advertising space for you guys?
Thanks, Mike. It's David. Regarding capital allocation and the convert, we took the opportunity to repurchase half of the note at a very attractive price. We did that in April of this year. We're continuing to look, to look at that opportunity, but right now I think we're at a good, good place on the convert. We did announce a new buyback of $30 million earlier this year. We still have significant dry powder in there, and, you know, we intend to continue to buy back some shares, not, not at very high levels. We expect it to be moderate. We think the, again, share price is still an attractive return on capital cost, but we also wanna make sure that we have enough cash in hand to do acquisitions.
The environment for acquisitions is, is becoming a little better, both on the private side and, and, you know, public side. We, we definitely wanna make sure we also have enough cash for that. I think we are right now at, at a good balance with the buyback at moderate amount.
Hey, Yaron here. I'll take the Gen AI question. I, I think generative AI is gonna have a bunch of impacts on media in general and, and advertising specifically. In media, obviously, every publisher we're talking to is looking to incorporate generative AI tools or assistance. Some obviously are going more extreme and are talking publicly about about generating content with Gen AI. I think others are looking at this as assistant to their human editors, which is approach I really like. That's, that's on the media side. I think we're gonna see much more efficiencies in creating creating content and using kind of assistance with generative AI as, as draft starters and things like that.
On the advertising side, as I mentioned, we're using generative AI for headline creation or suggestions, and we take the seed that is provided to us by the advertiser, which is the message they want to get out. Then, with generative AI, we can generate 200 different variations, which they can think about, and they're not gonna deploy, you know, big teams to create those variations. When they get those recommended automatically, it makes them very efficient at looking at them, reviewing them, and approving them. Again, we're seeing direct positive impact in our advertising systems with click-through rate and better return on ad spend with those advertisers that are adopting that.
Last piece, which I think is gonna be interesting on Gen AI, we're, we're keenly looking into and playing with in the lab but haven't released anything. It's obviously on the image side. I think everything that has to do with creatives, and this is gonna impact, I think, in a larger way, the advertising industry. We're gonna have a lot of creatives that come through generative AI. We're not yet comfortable with the hallucinations that are happening in that space, so we're still tinkering with that, but haven't launched anything yet. Obviously, I think over time, generative AI and on the image and creative side is gonna be interesting as well.
Just wanna add to Yaron's comment, related to Onyx and Brand Studio. His last point around creative is important. I mean, we use the Brand Studio, which is our in-house studio, to work with existing creatives, repurpose them, and create high impact displays, create a unique ability to generate attention with, with, with creative. In that area, we're looking quite intensively how to use AI to improve these processes to generate better, better return on, on those creatives faster iteration. Brand Studio is, is an area where AI, we're happy into experimenting with AI around the Brand Studio for, for Onyx.
Okay, thanks, guys. Super helpful.
The next question comes from Dan Day with B. Riley Securities. Please go ahead.
Yeah, morning, guys. Appreciate you taking the questions. Just to clarify on Onyx, is this something that your publisher partners need to opt into? Are you only able to serve these impressions on publisher partners, or, or is it something kind of more like a traditional SSP that maybe could be extended across the open web, regardless of whether, the, the publisher itself is a core content recommendation partner or not?
Hey, Dan, David. They, they do not need to opt in. We're using it in mid-article placements, which we have and continue to acquire, whether it's code on page or through a waterfall or through header bidding. This is one area. End-of-article placements are, are ones that we have, and we, we don't need to, to have them opt-in. A lot of the supply is, is totally in our control, and publishers are excited about the, the offering. It does improve user experience. It's better quality and higher yield.
Great, thanks. Appreciate the commentary on the new Keystone wins in the quarter. You mentioned a couple of pretty large publishers in there. Just whether you can talk about the recurring revenue from that becoming material at some point. As far as pricing the product, you know, it's tough days for publishers right now. Do you feel like you need to price it at a discount in this environment to drive adoption first, and then, you know, down the line, you could increase prices as you prove out the product? Or has that not been a concern and your pricing as expected?
Yeah. This is Yaron here. Thanks for the question. On the, on the pricing side, we're obviously working with these publishers, and we work with them at a, at a very large capacity in our existing business. We know that the best way long term to create value is to first ensure that they're seeing tremendous business value from Keystone. They're driving their other business KPIs, and with that growth, we're, we're confident there's plenty, plenty for everyone to have. You know, we don't break out the Keystone revenue from the rest of the business, but Keystone is really a great way to help us both with retention of publishers. It, it's, it's very strategic for them.
It starts getting us involved with all their business objectives, not just with native advertising. We also expect it'll help us improve the NRR. You will see the results baked into those, and it definitely should also help us win new publishers who are looking to diversify their revenue streams but don't really have any great technology to support those efforts.
I want to just add the, to the previous answer about the incrementality. Onyx is the incremental on, on the two sides of the marketplace. On the advertiser side, these are incremental budgets that we never had access to, which really increased our... And very significantly, and on the publisher side, these are outside of the feed, so they're not placements that are in the feed. These are either mid-article or end-of-article, 100% share of voice, high-impact display or video. They are incremental. The end-of-article could theoretically have some impact on, on the feed itself, but it's only going to be served when it's a, a higher RPM generating.
Great. Appreciate the answers, guys. Thanks.
Thank you. This concludes our question and answer session. I would like to turn the conference back over to the management team for any closing remarks.
Thanks, operator. We're happy to have delivered strong results for Q2. We continue to focus on disciplined execution. As you heard, we're excited with the opportunities that recent developments in AI have introduced. We're staying on the forefront of this revolution by implementing AI capabilities across our products, algorithms, and operations. We're also very excited with the launch of Onyx, our innovation for brand advertisers focused on attention. Thanks for your time. We look forward to updating you again next quarter.
Thank you. The conference has now concluded. Thank you for attending today's presentation. You may now disconnect.