In short
Meta attribution changes plus new “agentic ads” (AI business agents) and what to test now.
Guest backgrounds
Jacob (Head of Socials) at Pilothouse; long-time agency operator (6 years) focused on Meta ads and social performance.
Key claims
- Meta introduced incremental attribution: standard attribution shifts toward click-based conversions; “incremental” uses holdout/ghost ads to estimate purchases that wouldn’t have happened without the click.
- Incremental attribution can be used for optimization (ad sets optimize toward incremental conversions), and accuracy improves over a few weeks.
- Meta ROAS may drop because likes/views are no longer credited, but total performance often stays consistent when measured with third-party tools.
- Agentic ads: AI business agents appear at the ad level, answer product questions in an in-Meta chat, then route users to the site for checkout; brands should test and QA (FAQs, sizing, brand voice).
Notable examples
- Pilothouse reports ~40–50% of clients optimizing to incremental; triangulation with Triple Whale shows closer alignment to “true” Meta-driven conversions.
Written by AI. May contain mistakes. Listen to the episode to check what was said.
Chapters
Tap a time to open that second in VOUnderstanding Meta Attribution Changes
2:26 to 3:25
Discussion on the changes in attribution practices by Meta.
“This is the D2C podcast and I am here with Pilot Houses.”
New Incremental Attribution Explained
3:25 to 4:47
Explaining the new incremental attribution feature and its implications.
“Historically, someone's liked to post, they buy a few days later, Meta takes attribution.”
Holdout Groups and Conversion Tracking
4:47 to 6:32
How holdout groups improve conversion tracking and attribution accuracy.
“Yeah, that's kind of the general change.”
Navigating Attribution Windows
6:32 to 7:45
Discussion on controlling attribution windows in Meta's advertising.
“to give us more insight across all channels, not just meta-specific.”
Optimizing for Incremental Attribution
7:45 to 9:09
Insights on optimizing ad campaigns for incremental attribution.
“Like at the, if you just launch a brand new account, the first week that incremental is going to be probably not a hundred percent accurate.”
Aligning Data for Better Results
9:09 to 10:36
Importance of feeding accurate data for better ad performance.
“this is what I care about, net new customers or people that aren't just going to buy otherwise.”
Introduction to Agentic Ads
10:36 to 13:09
Overview of Meta's new AI business agents and their functions.
“Like the weight each purchase across each channel where it actually happened.”
The Future of Ads with AI Agents
13:09 to 14:02
Exploration of how AI business agents may change advertising.
“We're getting like the monthly updates from our agency advantage partnership.”
Evolution of Ads with AI Business Agents
14:02 to 15:10
Explore how AI business agents are transforming ad interactions on platforms.
“These ads don't just click to site, but they click to an agent chat where AI will handle the whole funnel, give you the FAQs, sizing, objections.”
Testing and Iterating with AI Ads
15:10 to 18:12
Learn about the importance of testing AI features in ad campaigns.
“So that's where a lot of the learning is going to come from.”
Show all 13 chapters
Setting Up AI Business Agents for Success
18:12 to 19:56
Understand the steps to effectively implement AI business agents in ads.
“Definitely keep an eye on the comments on your heads.”
Performance Insights from Meta Ads
19:56 to 22:48
Get insights into the current performance trends in Meta advertising.
“We'll have to keep the podcast posted on how that goes.”
Attribution Challenges in Meta Advertising
22:48 to 24:16
Discuss the challenges of attribution within the Meta advertising ecosystem.
“Doesn't it discount top of funnel too much?”
Transcript
Automatic transcript. May contain errors.0:00Before we jump into today's all killer, no filler episode, a quick word about who makes this show possible. The DTC podcast is brought to you by Pilothouse, the performance agency behind some of the fastest growing DTC brands in the world. Creative, media, and customer journey all under one roof. Performance and brand without the trade-off. Every Friday, we hand the mic to a Pilothouse operator to break down what's actually working in their space right now. Want a team that treats your growth like their own? That's Pilothouse. Head to pilothouse.co and now on with the show. What is going on with attribution on Meta these days?
0:33Jacob:Meta has introduced this new feature called incremental attribution. A lot of people aren't even aware of this feature because it just got lost in the mix. Agentic ads, the rollout of Meta's AI business agents. Tell me about that. Definitely need to be taking advantage of these things and testing them because it's just really like putting your ads on steroids essentially.
1:03If you're still making decisions based on messy, scattered data, bouncing between ad platforms, spreadsheets, and dashboards, you're not alone. Most e-commerce teams are stuck piecing together a story they don't fully trust. Triple Whale changes that. It's a complete intelligence platform built for modern operators who are done guessing. Triple Whale centralizes all your data, equips you with measurement tools you can actually trust, and layers in best-in-class AI to turn that data into clear recommendations, showing you what's driving growth, what's hurting profit, and exactly what to do next.
1:35It even takes action for you. With Triple Whale, you can quickly spot trends, catch issues before they impact revenue, and uncover opportunities you would have otherwise missed. It's about moving faster, making smarter decisions, and operating with real confidence. That's why brands like True Classic, Way, Press Juicery, and thousands of the fastest growing D2C companies rely on Triple Whale every day. They're saving time, increasing efficiency, and unlocking insights that simply weren't possible before. And if you want to hear directly why users love it, here's your chance. Triple Whale is hosting The Whaleys, a one-day summit designed for e-commerce operators who want to make more confident decisions and scale faster.
2:12You'll hear from industry leaders and experts sharing practical, data-driven strategies you can actually implement. If you're ready to stop guessing and start growing, you won't want to miss it. Visit triplewhale.com slash whaleys. That's T-R-I-P-L-E-W-H-A-L-E dot com slash whaleys to get your ticket today. It's all killer, no filler. I'm Eric Dick. This is the D2C podcast and I am here with Pilot Houses. What's your title again these days, Jacob? I know you are our longest standing team member. I know you're leading the socials division. What's your actual title?
2:46Jacob:Yeah, head of socials. Head of socials. I should have known that. Nice and simple. And I've been having these conversations with you for around six years. And Meta is this ever-changing landscape. And we are here talking about some of the changes that have kind of set in over just the last few months around AI and specifically around attribution. Let's start it off there. What is going on with attribution on Meta these days? Yeah, I mean, a lot's going on. Some big changes in the last two months, essentially, where Meta has introduced this new feature called incremental attribution. What this is, is Meta starting to adjust how they count attribution for ads.
3:31Jacob:Historically, someone's liked to post, they buy a few days later, Meta takes attribution. Someone sees an ad, they don't even engage with it. They buy a few days later, Meta takes attribution, as does every platform, right? But Meta has realized that pretty much everyone is seeing the need for third-party reporting. You know, let's unlock the true value of every ad that we have. Meta wants people using Meta. So they've shifted how they attribute ads to focus strictly on click, click actions. And ultimately, this is a good thing for advertisers. You know, we hear from 80 % of our incoming clients like, oh, I don't trust Meta.
4:11Jacob:Meta's ads over attribute. This is them sort of combating that with more meaningful attribution. It's acknowledging that measurement has become harder. Meta's unleashing some new tools to help close that loop, which, yeah, is great. It's a real mature move from an ad platform. I've been part of ad platforms for decades, and they're almost always trying to claim more credit, not actually identify. It takes it from a game of credit to a game of truth in a way or attribute of really understanding. And so you're saying that they're only going to count something towards someone's conversion if they've clicked on it?
4:47Jacob:Yeah, that's kind of the general change. That's the change to their standard attribution now. You can still view purchases, but that theoretically should be going away. Um, click conversions historically have counted people liking a post or clicking on the video to watch it. Um, now it's like clicking to website essentially only counts much more intent from like correlation of an ad versus the actual causation of an ad. And now there's this new column that you can add called incremental attribution. And those specific ones are purchases that wouldn't have happened without the click on the meta ad.
5:29Jacob:So it's even taken it one step further where now you can say, hey, I have 100 conversions in the last day. 70 of them are incremental. Those are the ones that wouldn't have happened. And they do this by having a holdout audience, essentially, holdout groups where they show them ghost ads so they don't actually see the ads. But Meta's tracking them as if they did see the ad. And they're saying, hey, would these people have bought anyways? people you know okay we sent our ghost ads to 100 people 30 of them still purchased even though they didn't see a physical ad so they know hey 70 of this brand's ads in this campaign are incremental the other 70 are because of our ad interesting this is like super powerful as a agency or anyone who's really running the ads because it's just giving you that extra insight and you're not needing to unlock this full puzzle, you're still going to use, like we're still using other third-party reporting to fact check it, to give us more insight across all channels, not just meta-specific.
6:36Jacob:How does this funnel play? But that's a big shift that a lot of agencies or brands are saying, hey, our meta ROAS has dropped. But then you look at the full picture and it hasn't. Yeah, it answers the oldest question about marketing that we've talked about on the podcast a lot, which is I've got 50 % of my budget works and 50 % doesn't, and I don't know what's what. And so it seems, so how much budget do you have to run to get a proper holdout test? And how long, how long does that take? Because I know people have been, you know, operating manual holdout tests to, you know, prove things or not prove them for their brand for years.
7:12How does it get built into a campaign now?
7:15Jacob:It's a good question. And I don't have like the full answer, but it happens automatically in the back. So you're not, you're not choosing holdouts or anything like that. Meta is just doing it. It's not optional. You can still see, you know, again, your, your full purchases, which is now less because it's only like the meaningful click, but then you can add this other column where meta is, is showing the holdout or the incremental. And I think, you know, for it to become more and more accurate, it does take a couple of weeks. Like at the, if you just launch a brand new account, the first week that incremental is going to be probably not a hundred percent accurate.
7:54Jacob:They're still running holdout tests. And I don't know if meta will ever release how they're specifically doing it, the actual like algorithm behind it. But just like an ad goes into learning phase, that incremental conversion is going to be more and more accurate as it gets data and settles in. And then what's going around with, what's going with conversion windows at this point? Cause I remember many years, it was four years ago where like we were, you know, there are times where, you know, optimizing towards the one day click really helped you focus on the bottom part of the funnel, really helped you knock some things loose.
8:26Do you even have access to control your attribution window at this point?
8:29Jacob:You do. Yeah. So you still have all those options at the moment. It's just that you have one day, seven day, 28 day, and then you have one day view. It's just that, that default of the seven day click again is now not counting certain actions it used to count. So you can still play around with these attribution settings to unlock a story for sure. But the incremental attribution is the new one in that same list where you're selecting them. And you can view those independently of the other ones. And you can also optimize to incremental attribution. So you can now tell your ad sets, this is what I care about, net new customers or people that aren't just going to buy otherwise.
9:18Jacob:I want these ads to optimize to everyone that is going to be incremental. So it's not just a reporting feature, it's also an optimization feature. Very cool. And we're at this point experimenting this or with this setting or we've moved a lot of campaigns to it? We are shifting quite a few campaigns to it and it has been working. I'd say like 40, 50 % of our clients now are kind of optimizing to incremental. Hasn't worked in every case. And just like anything, there's always going to be a big story around why that would be. Maybe it's a brand with a huge following. Maybe you're looking for second or third touch purchases anyways, and you don't care about the whole first touch thing.
9:59Jacob:But the audits coming in, like a lot of people aren't even aware of this feature because it just got lost in the mix of, you know, metas releases. Yeah. But it's a really important one. And what you're finding then when you do triangulate against with an MTA or with, with triple whale or something, you are finding that the, that when the revenue isn't, it's just being attributed differently. It's not going down when the ROAS on meta goes down in most cases. Yeah, exactly. We're finding that the incremental conversions on meta align a lot more closely with what you would see on like triple whale, meaning that these are true conversions attributed to meta and triple whales trying to do the same thing, right?
10:44Jacob:Like the weight each purchase across each channel where it actually happened. Yeah. It behooves you to get your signals right. And I was just chatting with my co-founder about the ads for D2C. We've just been through this exercise, actually not for D2C, for our agency community business. We went through this exercise where we were feeding it kind of the wrong information. We were giving it the seed audience as people who'd signed up for agency. But the problem is like 80 % of people aren't qualified for agency who sign up. And so we were like mistakenly just giving it this whole list of leads where 80 % of them weren't the right people.
11:19And just by switching it to this very, you know, using the server to server, switching it to this like offline action where we've approved someone to be, okay, this is going to be a good fit for agency. it just again that's very that's 101 like just making sure you're you've got your pixel on the right action um but it's amazing how profound a difference it made in the weeks before that change in the weeks after that change so i can imagine by using these attribution models this better incremental attribution model you're just going to be feeding it more real data because people clicking or liking or commenting on an ad are not the same intent level at all in the same category of intent probably as someone who's who's going to your website clicking through on one of your value propositions or whatever.
12:01So hopefully by feeding it higher quality signal, you get higher quality results in the long run.
12:05Jacob:Yeah, a hundred percent feeding it the right like signals, like you said, and then also ensuring your creative is aligned to the user intent of those people. That's really the other key, like Meta's ranking engineering agent, REA, essentially has like doubled the model accuracy, meaning their AI is like two times better at predicting who's going to buy now versus what it was two years ago, which is a huge improvement. So it really relies on like our ads, our creative, our hooks, the first three seconds of our video speaking to those people. If yes, Meta is going to go find the right people because it's like these are the people interacting.
12:46Jacob:Let's take signals from those hundred people. Let's find more of them. Boom, boom, boom. And it's like a self-learning system. And if you're feeding it incorrect data, it's going to take longer for the engineering agent to like get there, essentially. Second topic we have today is Agentic Ads, the rollout of Meta's AI business agents. Tell me about that. Yeah. Yeah. Again, this is really new. We're getting like the monthly updates from our agency advantage partnership. You know, so we do typically get kind of get the info ahead of time a lot. And so we've been testing this. It's not in every account yet, but you should start to see it popping up.
13:29Jacob:Essentially what this is, is moving to autonomous agents that talk, troubleshoot and build right on the ads. So these are the agentic ads, customer facing. You can go in, you can customize it, feed it data, make it learn. But it's essentially like a business assistant. And you're going to start seeing these pop up on ad units where, you know, hey, do you have a question about this product? Hey, we think this is the perfect fit and here's why. And it's just going to be doing this on top of your actual, you know, ad unit that you're delivering. These ads don't just click to site, but they click to an agent chat where AI will handle the whole funnel, give you the FAQs, sizing, objections.
14:13Jacob:Of course, it's extracting this data from your funnel, your website, right? Keeping people on Facebook, getting them what they need, educating them. and then when they're clicking to site, they're ready to convert. So it's just the next evolution of ads. They're called AI business agents and you can go into the ad level and essentially flick them on. You'll see it at the ad level. And then there's also a menu in the ads manager where you can start to customize it, make sure it's set up properly, all these things and starting to see that roll out. Does it work or is it just putting speed bumps on the runway, as they say, If you're just putting another step between getting them to your landing page, is that a good thing?
14:54Jacob:Yeah, I mean, I think it will be. I think, again, it's only been out a few weeks, so I don't have like full confidence in saying, hey, we're seeing X percent upticks by running these. But definitely need to be taking advantage of these things and testing them. And hey, now with incremental conversions, we could have two campaigns, you know, one with the agentic AI agents and one without. let's see what the incremental is on both of them you know is it working or is it just converting people that are already aware of who we are maybe just had a few extra questions hey maybe they would have found those answers anyways on the website the idea of it makes sense to me for sure as people get more comfortable interacting with ai and just getting them that info up front one other thing is like the qa of this we haven't it seems okay from what we're seeing but But if that agentic ad is feeding someone the wrong size or FAQ, how do you manage that as a brand?
15:53Jacob:So that's where a lot of the learning is going to come from. But it also will do things like iterate on your hooks. It'll iterate and change the questions it's asking people based on what's working. So much like an ad and these AI enhancements have already been released, meta AI business is also going to keep enhancing itself and self-learning from itself. So it spits you out on the Shopify page? Like it doesn't actually handle the checkout at this point? Doesn't handle the checkout, but it basically starts a chat like right in meta. And then, you know, there's the option to click to site. I think you can like start the checkout, but ultimately it will redirect you to the site for the payment and everything.
16:38Yeah. I can imagine the conversion rate at that point after you've been, you know, properly gaslit by AI as it likes to do. I imagine the conversion rate is quite high at that point when you do get people there. Yeah.
16:49Jacob:A hundred percent. It's also like handling, like you can bring it beyond just the ads and push it into your like messenger inbox. So you're handling like DM inquiries, kind of like chatbots is all, you know, they've always existed, but you can run that through this same, same AI business agent essentially, and kind of customize it as you go. So it'd be interesting how it evolves because it's just really like putting your ads on steroids essentially. And how much of a, does it do all the learning about your brand from what you have available on your site? Or is it part of setting it up that you've got to provide all the proper brand documentation, brand voice, that kind of thing?
17:28Jacob:It's a bit of both. It pulls what it can from the site. Like you're putting your destination URL in the ad and it's sort of pulling from that. You have your Metapixel across your website so it can track people's interactions, but you can feed it data. So the FAQs, for example, you can go in, make sure you upload all those. And then, hey, now it has 100 % accurate info. It's not going to pull the wrong thing. So it's a bit of both for sure. And the actual customize option isn't available yet for many people, even for us. Like it's mainly just the, the feature to turn it on is starting to roll out.
18:09Jacob:So definitely want to keep an eye on it. Definitely keep an eye on the comments on your heads. Um, make sure again, the whole like sentiment is good with it. Uh, but it's a, it's a really cool enhancement that I think will improve conversion down the road for sure. Very cool. I have some steps for it, but yeah, let's, let's hear them. So basically you add it at the ad level, full functionality is not quite there yet. Eventually you will be able to access the assistant directly in your business suite or your ads manager sidebar. Step one, upload your site, PDFs, brand voice guidelines to train your agent.
18:48Jacob:So this is the piece that's coming out still. step two you can set up garb rails so you can set like limits on it you can set a tone of voice is it stern is it is it you know a funny like is it humor um you can also set like human human loop triggers uh and then step three is deploying it so like turning on agentic placements it's i think it's only available in advantage plus campaigns which would make sense but you just turn those on at the ad level. So you do have some control over it, but ultimately turning it on is pretty easy. So these are some two big changes here. The one thing that we can talk about maybe at a future one, we'll get Jeff on as well, because Jeff has just launched a DTC ad account on his Claude through the MCP.
19:38So I'm interested to see what kind of results. We actually just launched our first campaigns for subscribers. And initially we're getting the sort of exact same results we're getting on the managed account. So I'm interested to see if it's just a new account getting some low-hanging fruit or if it's something that can continue because it's an interesting experiment. We'll have to keep the podcast posted on how that goes.
20:00Jacob:100%, yeah. It's evolving rapidly. I think the best thing you can do right now is empower yourself, your employees, everyone to experiment. Make sure you have holdout tests and ultimately don't just commit to one format or one tool because, like you said, Claude, you can go in, you can set up the CLI line and basically start pushing live builds right into meta, QA it in meta, test it against here. Don't just assume it's going to be better, but also you can assume that there's efficiency gains, right? So it's just like there's Manus, you know, Meta's AI chatbot, which they're trying to match with Claude.
20:43Jacob:So eventually I'm sure you're going to be able to push things through there just as efficiently as Claude. So just be having a pulse on all these things, test them out, but test them efficiently, QA them. And then on the other side, Meta is also trying to keep people in Meta, right? They want people using Meta. So they're releasing things almost too fast where it's like, you do need to be keeping a finger on all these things. Yeah. The thing about, I was actually chatting with a friend of mine who I met in Vegas who runs Hazel, which is an AI platform for sort of, uh, that ties all your platforms together is, is the idea with that.
21:18And he was just talking about the limitations of MCP specifically and how it's good for like single call commands where you like launch this campaign. What are my results here? So, but it's nothing, you can't really rely on it to synthesize data that you might be able to, uh, right now anyways, you know, otherwise in Claude and that it's got some limitations when it comes to a lot of like analysis and synthesis, but it's only, it's such early days.
21:42Jacob:Yeah. It'd be not long probably. And that'll just be improving as well. So it's all improving on itself for sure. How are results overall right now with, with meta across accounts? Are we having regular Q2 results? Yeah. I was actually looking the other day on like the master kind of report and pretty consistent for Q2. Like I said, some of our accounts have seeing platform ROAS drops in meta simply because they're not counting like the likes and the things that they used to count. But then you look at the ecosystem and it's either consistent or upticking. So yeah, I think overall like performance has been pretty good.
22:25Jacob:There was a few dips, you know, a couple like last month essentially, obviously lots going on in the world. in the US, people being a little more careful with their spending. But honestly, there's always something going on. So it's hard to attribute it specifically to that, but pretty normal results from what I've seen. Who's the big winner? Is it the retention team that would end up counting the conversion in most cases that Meta is relinquishing by not counting the clicks on on posts and videos and such? Yeah. I mean, theoretically, if they're not showing up in the incremental or they're just liking a post and then buying, it's probably organic, honestly, like it's just going to show up unless they're clicking maybe a Google ad or watching a YouTube video or, you know, unless they are interacting through another channel, if it's just a quick view of a video and they buy, it's probably just going to show up as organic in your third-party reporting now.
23:24Jacob:It's interesting. Doesn't it discount top of funnel too much? Like, isn't there, I guess, I guess they're ultimately trying to get them to engage with your page on their platform anyway. So even if it's the job of the impression to get them to a click, I just, I just, there's platforms for so long, we're, I've been trying to push the value of impression-based conversion. So it's just interesting to hear them move back on it a little bit fundamentally. Yeah, no, it is interesting, but it is a consistent complaint everyone's always had about meta is that they over attribute. You don't even have a choice a lot of the time to view an ad, right?
23:59Jacob:You're just scrolling and all of a sudden there's an ad in front of your face, something you're already aware of potentially. So it's kind of like a middle ground. They're not discounting it completely. Like I said, you can still view the view purchases, but the default one, the seven day click isn't including some of those actions now. Good to know. Well, thanks again for the updates. And yeah, man, stay in touch about it. We got playoffs starting now. Who are you rooting for with the Canucks out? I mean, probably Canada team, maybe Montreal. Montreal, baby. There we go. Team of destiny. And thanks for having me on, yeah.
Read the full transcript
24:32All right. Talk to you later, Jacob.
24:39Thanks for listening to today's episode. If you're not getting the DTC newsletter, you can subscribe for free at directtoconsumer.co. And if you want to learn more about Pilot House's all killer, no filler services, take off to pilothouse.co. I'm Eric Dick, and this has been the D2C Podcast. We'll see you next time.
From the publisher
Subscribe to DTC Newsletter - https://dtcnews.link/signup
Meta attribution has changed, and most brands are still reading performance the same way they were a year ago.
Jacob, Head of Socials at Pilothouse, walks through what’s different now. Click-only attribution, incremental measurement, and how those shifts affect the way conversions show up in your dashboard.
If you’ve noticed numbers feeling off lately, this will help you understand why and what to actually pay attention to.
For DTC founders and marketers spending $50K–$500K/month on Meta who want to understand what’s actually driving conversions.
- What click-only attribution removes from reporting
- How incremental attribution works in practice (without the fluff)
- Why conversion numbers feel different even when performance hasn’t changed much
- Where Meta data and third-party tools start to diverge
- How Pilothouse is thinking about reporting and decision-making now
Who this is for:
DTC operators, performance marketers, founders scaling paid social
What to steal:
- Check incremental conversions alongside total before making changes
- Look at channel performance in context of total revenue, not in isolation
- Use Meta data as one input, not the final answer
Timestamps:
00:00 Intro
02:28 Meta’s New Attribution Shift
05:03 How Incremental Attribution Works
08:11 Conversion Windows and Optimization
10:15 Why Better Signals Improve Meta Performance
13:02 Agentic Ads Explained
16:02 AI Business Agents and Customer Conversations
18:05 Setting Up Meta’s AI Agents
19:28 AI Ad Automation and Testing Strategies
21:14 Q2 Meta Performance Trends
23:02 Why Meta Reduced Over-Attribution
Subscribe to DTC Newsletter - https://dtcnews.link/signup
Advertise on DTC - https://dtcnews.link/advertise
Work with Pilothouse - https://www.pilothouse.co/?utm_source=AKNF609
Follow us on Instagram & Twitter - @dtcnewsletter
Watch this interview on YouTube - https://dtcnews.link/video




