The questions submitted during the Q&A session with Rand Fishkin and Britney Muller at the Online Sales and Marketing Summit were so good! We didn't come close to getting through them all on stage, and a few have stuck with me in the weeks since. So I went back and answered every single one. When you line them up side by side, a pattern jumps out pretty quickly: almost every question is some version of "How do I make a responsible decision when the reporting I've relied on for years doesn't feel trustworthy anymore?"
I've grouped the answers so they build on each other, starting with attribution (since it impacts nearly everything else) and finish with what you can actually do as a result. Some of these answers will feel like relief. A few of them will probably annoy you, and that's fine too (it usually means the answer is worth sitting with a little longer).
Why Your Data Isn’t Enough
The first question I'll tackle asked how to tell their boss that obsessive metric and event tracking on the website isn't delivering the results they're striving for. Unfortunately, there's no getting around the fact that this will require significant education on how the modern web and tracking actually work. The good news is that sources aren't scarce. The bad news is that those sources rarely go viral, because they don't fit the version that people like your boss want to believe (quick fixes you can pay for or decisions you can let the data make for you). Follow Rand Fishkin on LinkedIn, and keep an eye on what we publish at Do You Convert, because we're committed to doubling down on helping people understand that last-click attribution isn't helpful when analyzing budgets and, more often than not, is misleading.
The single best teaching tool I've found for this conversation is a buying journey graphic shared by Eli Goodman, President of Datos. It maps the path to purchasing a relatively simple product, a $1,000 Samsung television, using a composite of observed waypoints across multiple panelists. Buyers bounce between Reddit, search, Samsung's site, Perplexity, ChatGPT, Target, and other product research before the purchase finally happens on Samsung's website. As Goodman puts it, "Samsung's own analytics can only directly see the parts of that journey that happen on its website." Everything else happens without Samsung having any idea the journey took place.
And those are only the digital parts of the journey. The graphic doesn't include walking into a store to look at the television in person, or asking a friend what they think of theirs. We only have data like this because companies like Datos collect clickstream data, tracking a consumer's behavior directly on their device. If a $1,000 TV purchase looks like that, what do you think the path to buying a home looks like over 150+ days?
Another question pushed on this more specifically / technically: don't server-side conversion data and logged-in user signals from Meta and Google still provide fairly strong attribution?
Yes, they can give you more reliable insight into what happened right before conversion. They will never give you the full customer journey, though. Go back to the TV example above, in which digital signals are only some of the signals, and you can only capture those digital signals through first-party and third-party data. To your point, some data sources are better than others, but none of them are good enough to model where you should be spending your time, energy, and resources based on attribution data alone. You need something more, and that still requires human discernment based on experience and close attention to your customers.
Defense, Discovery, and Disguises
I mentioned on stage that Google tends to get credit through last-click attribution, and one attendee asked a great subsequent question about unbranded exact-match keyword impression share in Google Search ads compared to branded terms alone. Paying for an ad on your own brand name inherently means the buyer already became aware of you somewhere else. Brand terms tend to convert the highest because those buyers were already likely on their way to you. It's the most extreme example I can think of when it comes to playing defense with your ad dollars, and it's also one way traditional agencies claim to produce results for you instead of acknowledging what's really happening.
Unbranded terms are completely different - and more valuable because of this difference. They're clearly part of the discovery and awareness-building process, because the buyer isn't typing a specific brand name. They're searching a generic category, and in our industry that category likely includes geography that gets more specific over time (state or metro -> county -> city -> etc). That's the role to evaluate in your search budget, separate (and protected when it comes to your budgeting!) from what you're spending to catch people who already typed your name.
The same logic applies to a question about vendors like OppGen that drive high traffic but low responsiveness and low sales. This falls squarely in the category of "there is no need to do this except to give leadership the illusion we are giving them what they are asking for.” It creates high traffic because it isn't as targeted as it claims. Inevitably, it ends up functioning a lot like branded search, claiming to be part of the process when it really just grabbed someone already interested in you. Neither Do You Convert, Rand, nor Britney would encourage approaches like this, because high-volume traffic (or even “leads”) isn't worth what you're spending on it when you don't have clearer visibility or control. This is why responsive-leads are a great additional way to review your data when making decisions.
The Rest of the Airbnb Story
Another wonderful question didn't make it to me while on stage, and I've gone back to it several times since Rand's talk. The attendee pointed out that Airbnb is a very recognizable brand with high repeat usage and referrals, and asked how eliminating digital ads would impact a smaller business. Such an insightful question! Airbnb cut brand and performance marketing 58% in 2020, and the traffic recovery that followed turned it into one of the most repeated case studies in marketing about how advertising wasn’t as necessary as many thought. Thankfully, because Airbnb is a public company, you can read the rest of the story in their SEC filings: by 2023, spending in that category exceeded its 2019 level in dollars, and brand and performance marketing in the first half of 2026 reached $1.09 billion, up 32%.
This is, I think, the weakest spot in Rand's talk, where he essentially hints at ads being mostly worthless or simply a tax. It isn't nuanced enough, especially for our industry, where ads serve an incredibly important role in the 90 days before a purchasing decision. Because our buyer journey is so long, zero-click has a huge impact at the beginning and middle of the journey, and ads also have plenty of time to impact a shopper who hasn’t considered you yet or to reconnect with one who already has. Content created through a zero-click approach can carry through, and you should use it for remarketing in your advertising as well. Since Airbnb was the single example used to make the point about stopping ad spend, and Airbnb has modified so much since then, it's worth pausing to review the whole story before you bring it into a budget meeting or make a dramatic shift.
Which brings us to paid social. Should you keep doing it? Yes. But as the entire team at Do You Convert has been telling builders for years, you can't simply spend more on any one platform and force the results you're looking for to happen. Measuring reach or impressions on paid ads is less valuable than measuring the same metrics on organic content, because on paid you're buying that privilege, so of course it happens. When organic content earns reach and impressions, it's because the algorithm believes the content is strong and is giving it more delivery, and that's a signal worth paying attention to. Remarketing and in-market segments on paid social are extremely important, but when you don't have enough content to consistently update and vary the creative, consumers have already seen it, and that's where diminishing returns set in and why so much attention at this year’s Summit kept going back to content with high context still not being invested in enough.
LinkedIn came up too, with a question about whether the data supports a focus on Realtors or consumers. Even though Realtors are a segment you could focus on, consumers are always going to be on the platform as well so you have to be discerning about your positioning. You can't segment organic content unless you pay to target people with ads, so write knowing both audiences are paying attention. On the ad side, both perform well, although cost per click and cost per impression for consumers will be significantly lower because there are more of them (every digital ad platform charges you more per result when your target audience is smaller). The key on LinkedIn is understanding that you don't have to be there all the time like paid search. After an initial 30-, 60-, or 90-day campaign, you can front-load your spend for a week, then go dark for the rest of the month, and that incrementality approach helps you track effectiveness over time as well.
Create, Curate, Contribute
“How do you begin to think like a creator rather than a marketer?” was a question asked in a few different ways by several people. This one is all about understanding the difference in incentives. Creators (artists) want their creation to be viewed and understood (or at least viewed), and, at the very least, to exist for a purpose. Marketers historically thought this way too. The modern digital marketer is usually focused exclusively on what a piece of content gets you next, an immediate next step connected almost by a piece of string to the one after it, and that becomes a rather transparent manipulation approach. The consumer can feel the difference, and we can too as marketers if we’re honest. Thinking like a creator means making great "art" for the benefit of the audience who will experience it, without focusing too much on a obvious next step, and I hope those lines keep blurring as more marketers adopt that approach.
The same instinct showed up in a question about AI-written listing descriptions. If everyone is using AI to write them, wouldn't writing them with a human touch be a way to differentiate? Britney answered this one from the stage and said she completely agrees with the sentiment. The idea is to use the probability-based nature of LLMs to generate an unlimited number of ideas to start from, then use your human senses and experience to curate the best ones and guide them to completion. Nobody from Do You Convert, and neither Rand nor Britney, is suggesting AI is an easy autocomplete for getting content updated. Be unmistakably human, or you run the risk of losing trust with your customer.
Reddit is where that mindset gets tested, and a few questions on the platform came in. My first tip for builders is to observe and learn as much as you can by watching threads over time to get a sense of the context and activity. Reddit is a great place to learn, and that's why both Britney and Rand suggested spending time in communities you're naturally interested in in your personal life, alongside anything related to home building. Your natural desire to join those conversations will teach you the platform's mechanics as you observe areas tied to home building and real estate. Britney also mentioned using AI as a listening tool to alert her to conversations where she can add value, or at the very least to synthesize what's being said as free market research. Your ultimate goal is to provide value, so don't post ads disguised as organic content or dump content there in hopes of getting picked up in AI search results later.
Tools Worth Testing (and Tracking Worth Questioning)
Several questions were about getting started with AI tools, with Claude coming up the most. If you want to start with Claude, Anthropic has its own training program that's easy to work through in small segments, and that's a great place to begin. I also highly recommend Britney Muller's Actionable AI for Marketers class. At Do You Convert, we're launching an introduction to advanced AI techniques in early November, and it will be free for all 2026 Summit ticket holders in the marketing track.
One marketer shared that their company recently adopted Claude with limited access, and they've been restricted from Claude CoWork, which is the part they were most excited about. I appreciate the excitement, because CoWork unlocks a lot of opportunities. The biggest benefit is that it lets you write more and keep it as reference material for larger, ongoing projects, which matters because a single chat window can degrade the AI's effectiveness over time. As of October 6, Anthropic transitioned Claude CoWork to run completely in the cloud for all Pro and Max plans. That's good news for those limited by their IT team, because you no longer have to give the tool access to your local machine. It's almost like Claude gives you a small, persistent computer in the cloud to do CoWork-style work with, and the bonus is cross-device continuity (launch a task on the web, check its progress from your phone). Talk with your IT team about gaining access to that cloud environment through a Pro or Max plan.
Meta Muse connects directly to that same idea. It's the first mass-market opportunity to easily engage with a persistent online session of an LLM, in the form of an agent that helps you solve basic tasks. The biggest concern with anything related to Meta, given its social media background, is privacy. The current framework claims to be a completely separate system, and I wouldn't bet it lasts long. Still, it's incredibly intuitive. It isn't trying to get us to Mars or extinguish all human life on Earth, but it can help you get a dinner reservation, clean up your inbox, or do other helpful, time-saving tasks. One recommendation: the setting that lets Meta train on your interactions with Muse is on by default, so go in and turn it off.
The last AI question was about measuring whether GEO or AI visibility efforts are working, since ChatGPT and other platforms don't give us the reporting we're used to from Google. Begin with acknowledging that consumer use of these tools is simply not as high a percentage of overall searches as people believe. Both Rand and Do You Convert have shared a lot of data on this over the past year, and it's a small fraction, so over-indexing on it is not a good investment. If we compare this moment to the history of the internet and website importance, we're much closer to 1999 than 2009. Second, ranking doesn't exist in how LLM tools work for search, because each one is heavily personalized to each individual user. In Rand's words, tracking this properly would require running the same search thousands of times every time you wanted to update your data, per LLM! The reason you keep hearing about it is that it's an easy way for people to make money from builders when there's no clear, definitive way to track results or bottom-line impact. Even worse, it's easy to manipulate the tracking to make it look like you're getting results you aren't.
Test, Don't Torch
The next question to answer came from an attendee who admitted they've thrown a lot of spaghetti at the wall and don't know what's sticking where. They asked whether they should start completely over. Acknowledging that you're testing too many things at once is a great place to start, but I don't recommend starting over. Declaring bankruptcy on everything you're doing right now can feel like a reset, and it also throws away all the data and wisdom you could pull from remembering where you've been. You can start testing incrementality immediately by turning things off and noticing the impact, then iterating on your next step forward instead of starting from scratch. This is also a benefit to our team working with over 100 home builders across North America is we get to see incrementality tests almost daily, and we can overlay patterns to help determine the best current tactics achieving results.
The final question asked something every marketing leader eventually gets asked by an owner or CFO: if we're removing traffic from our KPIs, is there one marketing KPI we can show in a high-level report that proves marketing is doing its job? If you made me pick, I'd reach for one of two data points: responsive lead count or appointments. Both show proof of life, a reality that something really did occur. That's far harder to misinterpret or confuse than traffic or lead counts without the responsive qualifier. A responsive lead is a real person engaging beyond filling out a form, and an appointment is a concrete opportunity for your sales team to close the sale.
Before your next budget meeting, pick one investment you've been defending with attribution data alone and give it a real test. Turn it off or flight it down, watch your responsive leads and appointments, and bring the Samsung TV journey with you to explain why you're measuring it that way. Feel free to ask your DYC coach or strategist what other builders have seen from the same kind of tests if you’re nervous or not sure you need to try it. Start building a marketing plan that's honest about the fact that modern attribution data is compromised, and you'll be making better decisions while everyone else is still arguing over last click or spam leads.
