Catch the full webinar: Personalization Without the Bullshit: What’s Actually Lifting Revenue
Everyone has a personalization budget. Almost nobody can tell me, with confidence, what it earned.
That’s what I find walking into large retail organizations, and it was the first thing said out loud on an RTM Nexus panel I joined recently, Personalization Without the Bullshit: What’s Actually Lifting Revenue. Tim Zawislack set the frame in his opening and never had to walk it back. Most personalization running on enterprise sites today is theater. Hero banner swaps. “Recommended for you” carousels. Email subject lines with your first name in them. Then he asked the question every CMO is quietly avoiding: if all of that works, where is the incrementality?
Rebecca Kerper was on the panel too: 25 years at QVC Group, most recently Chief Digital Officer at HSN, now founder of Leadership Growth Group.
The number the industry still quotes to defend that budget is McKinsey’s finding that faster-growing companies drive 40% more of their revenue from personalization. It was published in November 2021, which makes it nearly five years old, and it is still the headline stat on half the slideware in this category. When a market recycles one figure for that long, that isn’t evidence. It’s a folk tale with a footnote.
Which points at the real issue. This was never a data problem. It’s a proof problem.
Enterprise brands spent the last decade buying inputs: CDPs, recommendation engines, dynamic content tools, audience segmentation platforms. The data is there. It’s rich, it’s specific, and there is vastly more of it than any team is currently using.
Ask what it returned and the room goes quiet. Attribution is fuzzy. Baselines moved. The test overlapped a promo. Merchandising changed the rules halfway through. Someone offers a number, and everyone in the room privately discounts it by half.
A personalization program in that state is a gym membership. The charge clears every month. Nobody can point to the muscle.
This isn’t because your analysts are bad at math. Proof requires volume: enough clean, isolated changes shipped fast enough that the signal outruns the noise. A team that ships four personalization changes a year has four data points and a hypothesis. A team that ships forty has a body of evidence. Same traffic, same catalog, completely different confidence at the budget meeting.
So the question stops being what should we personalize and becomes how many times can we find out.
Rebecca brought the best example I’ve heard all year, and it came from a pizza place down the street from her house. She’s a regular, she’s on the text list, and one day her phone buzzes with a breathless announcement: a new two-liter bottle of diet cola.
That’s the state of the art at a lot of brands, just with a bigger license fee. Set generic rules, get generic output. The system worked exactly as designed. The design was the problem.
Tim’s list of the formats still soaking up personalization budget was uncomfortably complete:
Hero banner swaps. Decoration at the top of a page the shopper already decided to visit.
First-name email tokens. A 1998 mail-merge trick wearing a 2026 badge.
Generic “recommended for you” carousels. The same eight best-sellers, rearranged.
I’d add a fourth from my own side of the table: audience segments nobody has touched in years. Technically labeled. No longer safe.
These survive for an unglamorous reason. They’re visible. You can put a banner variant on a slide and a director can see the money working. Deep personalization, in the search box, across the category page, inside the product page, reacting to what someone did ninety seconds ago, is harder to screenshot and vastly harder to build. So budget flows toward what demos well rather than what converts.
Rebecca called the alternative bravery, and she’s right, though I’d argue it’s rarely a courage shortage. Most teams are brave enough. They’re architecturally prevented from acting on it.
Two things stand between an enterprise team and personalization that shows up in the P&L.
The first is the Insight Gap. You can’t see where revenue is leaking fast enough to matter. There is behavioral data sitting in dashboards nobody opens, waiting on tagging plans, SQL, and an analyst with a queue of their own.
The second is the Activation Gap. Even when you know exactly what to fix, you can’t ship it. The change goes into a briefing document, then a design file, then a development backlog, then QA, then a release window. Six weeks later the seasonal moment has passed and the hypothesis has aged out.
The problem isn’t only that you can’t see what’s broken. It’s that the system that shows you the problem isn’t the system that lets you fix it.
Your analytics platform diagnoses. Your CMS can’t prescribe. Your testing tool needs its own implementation. Your personalization vendor needs a script that slows the page it’s trying to improve. Your smoke detector works perfectly. It just isn’t wired to anything that holds water.
Which is why “we need better alignment between marketing, merchandising, and engineering” is the wrong answer to this. Alignment is the workaround teams invent when the architecture won’t let them act. No amount of standing up in a circle grants your team permissions the system was never built to give them. This is the way most teams built personalization in the first place, and it is the part almost nobody revisits.
I made the governance argument on the panel, and I still believe it. At scale, across multiple brands, multiple markets, and org charts built around departmental KPIs the customer has never heard of, cohesion has to start at the top and cascade down. Get that wrong and you don’t ship a great experience faster. You amplify a bad one.
Rebecca’s addition was sharper than my original point. You can have every rule and every governance layer in place, she said, and still get tripped up if team members aren’t clear on their part in the omnichannel picture.
Her example was the merchant. The job used to be find the product, buy the product, get it to the store. Then it was buy the product and figure out allocation between store and web. Now there’s TikTok Shop, inventory to allocate across places that didn’t exist five years ago, and a buyer who has to know the customer as well as the product, because the search terms they hand the ecommerce team decide what’s findable at all. Being a product expert stopped being the whole job.
Her warning to leaders was the part I’d underline. Expertise turns into a blocker when you know the old version of the job too well.
Something I’ve watched shift alongside that: retail and ecommerce used to compete inside the same company, quietly, over budget and credit. That’s breaking down. The channels are starting to collaborate on strategy, and when they do they magnify each other instead of cannibalizing. The customer was always moving across them. The org chart is finally catching up.
The brands producing defensible numbers stopped decorating and started personalizing where the shopper is actually deciding.
Tapestry built a conversational assistant trained on conversations with hundreds of its own store associates. Shoppers who engage with it convert at four times the normal rate and add to cart at five times the typical rate (eMarketer, June 2026). Zenni Optical rebuilt search and discovery with Algolia and reported a 9% conversion rate lift alongside a 34% increase in search revenue between February and September 2023, measured against the same window the year before. Ulta runs personalization off 47 million loyalty members, with 95% of sales flowing through that program (eMarketer, June 2026), rather than off segments somebody drew in a workshop three years ago.
Notice what none of those is. None of them is a banner.
Notice something else. A search platform, a loyalty program and a conversational assistant are three different systems, and not one of them personalizes anything by itself. Each produces signal. What separates these brands from the ones running identical tools with nothing to show for it is the layer above: whether anybody can act on that signal this week instead of next quarter. That layer is where we work, and it is the layer most stacks simply don’t have.
The pattern underneath is simple, and it’s the one every good store associate runs. Someone walks in, you ask what brings them in today, they say they’re shopping for a wedding, and the entire visit reorganizes around that answer. Nobody finds that creepy. It’s the opposite of creepy. It’s being paid attention to.
Rebecca made the digital version of that point better than I did. A shopper searching “black dress” has told you something, and whether they’re dressing for a wedding, a funeral, or a Tuesday changes every result that should follow. Most sites do nothing with the difference. They return a hundred black dresses and a banner about free shipping, which is the digital equivalent of a greeter shouting the same slogan at everyone who walks past.
Two honest caveats, because a proof argument that ignores its own weak points isn’t proof either.
First: engagement multiples aren’t incrementality. Shoppers who choose to open a conversational assistant are already further down the path than shoppers who don’t. Some of that 4x is the tool working. Some of it is self-selection. Any team quoting a number like that to a CFO should expect the question and should have run a holdout to answer it. Rebecca made the same point about conversion rate itself. It might be incremental, or it might be the number you were going to get anyway. That this is a hard question is the entire reason the proof gap exists.
Second: speed without governance amplifies mistakes. Give a fragmented organization faster publishing and it will ship an incoherent experience faster. Shared component libraries, brand rules, regional overrides, and role-based access aren’t bureaucracy here. They’re what makes velocity safe at scale.
Which brings it back to the two that don’t talk to each other. Closing the distance between them is the thesis Fastr Workspace was built on. Fastr Optimize shows where the site is leaking revenue without tagging plans, SQL, or an analyst queue. Fastr Frontend makes the fix live: sitewide experimentation, real-time on-site personalization, full templates, no third-party scripts, no development ticket, no replatforming. And because the change and the measurement sit in the same workspace, you can show what a given experience actually earned instead of arguing about attribution for a quarter.
That is the whole answer to the proof problem. Not better attribution modeling on four changes a year. More clean, isolated, measured changes, shipped while the insight is still true.
AI is what compresses the distance between those two moves. It surfaces, prioritizes, and ships at a pace no team could staff for. It does not decide. That part is still yours, and it should be.
Nearly every consumer goods category is commoditizing. Whatever you sell, there’s a listing next to yours on a marketplace from a brand with twelve consonants in its name at a price you can’t match. Tariffs won’t change that. Those challengers aren’t going anywhere.
What legacy brands still own is the choice itself. The consumer decides where to spend their time and their money, and the brands winning that decision right now are leaning hard into strategy, taste, brand image and creative as the differentiators. Personalization is how that taste reaches one shopper instead of an audience segment. Done well, it’s the difference between a brand and a product listing. Done as theater, it quietly erodes the exact thing it was supposed to protect.
The strategy was never the weak link. Neither was the measurement.
What broke is an execution model assembled from tools that were never designed to hand work to each other, where the system that finds the problem has no authority to solve it, and the system that could solve it never hears about the problem in time.
The brands pulling ahead in 2026 aren’t better at data. Everyone has the data. They’re faster at turning it into a shipped change, and speed at that level isn’t a matter of effort or headcount or a better quarterly planning process.
It’s a matter of what you built.
Rebecca, Tim and I got into org design, KPIs and the trust recession in more depth than fits here. The full panel is worth the forty minutes.
Want to see where your own site is losing the sale? We’ll show you your own site, not ours. 20 minutes. No slides.