How Gymshark connects measurement & AI to drive incremental growth at scale

The pressure on performance marketers right now

You are being asked to do more with less. More ads, more channels, more markets, more tools to manage. Less budget, fewer people, and a smaller margin for error.

And then on top of all of that - AI is on the leadership agenda. Not as a future consideration. Now. Your exec team wants to see it embedded, not just one-off experiments.

The buzz is real. But so is the exhaustion. Most marketers are already running AI experiments - reviewing outputs, testing tools, trying to work out what is actually worth it. It is a lot. And for many teams, it has quietly become another source of pressure rather than relief.

So the question is not really should you let AI optimize your budget. Most brands already know the answer is yes. The harder question is: how do you do it in a way that drives real, scalable growth every day, so your teams can stop firefighting and start focusing on what is next?

That answer starts with trust. And trust starts with measurement.

That is what Dom Devlin, Fospha's Chief Product Officer, addressed directly at Shoptalk Spring 2026 — alongside Daniel Green, Head of Digital Marketing at Gymshark, and Jia Marquez, Senior Sales Manager at Smartly. The session posed three questions every retail eCommerce brand should be able to answer before they give AI access to their budgets. This article sets them out in full.

AI is the accelerant. Trusted, total commerce measurement is the fuel.

Here is the central argument, stated plainly: AI budget optimization is not a standalone capability. It is a multiplier. It makes your decisions more frequent and your spend allocation more precise. But it multiplies whatever is underneath it.

You are selling across your site, Amazon, and TikTok Shop. Your media is running across paid social, search, retail media, and emerging channels. The brands pulling ahead are not just running better campaigns. They have a total commerce measurement system that ingests all marketing and sales channels — brand site, Amazon, TikTok Shop, and beyond — delivering a single reconciled view that marketing and finance can trust, and that gives every channel the credit it deserves.

That is what turns AI from a faster way to make decisions into a compounding competitive advantage. Without it, automation accelerates the wrong things. With it, the performance gap between you and less-prepared competitors widens every single day - automatically.

The three questions that follow are how you know whether that foundation is in place. Think of them as a readiness checklist, and be honest about the answers.

Question 1: Are you measuring the total impact of every impression, across every channel, every day?

The first question from the session is about coverage. Are you confident you are measuring clicks, impressions, awareness, consideration, and conversion across every channel and every sales destination, including the ones you cannot pixel?

Think carefully about what "total coverage" actually means in practice.

It means every channel, full-funnel — because the two tools most teams rely on both fail here, in different ways. Last-click under-credits demand generation channels, attributing success only to whatever captured the final click. Platform-native data can't see beyond its own ecosystem — Meta reports what Meta drove, TikTok what TikTok drove, and neither can give you a cross-channel view. When your measurement has those blind spots, upper-funnel channels get starved of budget — and long-term growth stalls even when short-term ROAS looks healthy.

It means every sales destination. The brands relying on DTC-only measurement are missing an increasingly significant portion of their business. 40% of US eCommerce sales now happen on Amazon — that is approximately $1.75 billion in daily revenue on the platform alone. TikTok Shop is growing fast too, with 120% growth in US sales YoY in 2025, and neither channel can be pixelled. If your measurement cannot account for them, a substantial share of the commercial impact of your media spend simply does not appear in your data, and AI automation acting on that data will never see it either.

And it means daily. Traditional Media Mix Models have been valuable for strategic planning, but they were designed for quarterly cadence — not for the decision velocity that AI automation requires. If something goes viral on TikTok Shop, you cannot afford to wait weeks for a model run to tell you to act. You need daily, reactive data to feed AI automation.

Finally, it means ad-level granularity. You buy media at low levels of granularity — campaign, ad set, individual creative. Your measurement needs to reflect that. Understanding how individual creative is performing is essential to build a trusted foundation for AI.

What good looks like: A total commerce measurement system that ingests all marketing and sales channels — brand site, Amazon, TikTok Shop, and beyond — delivering a single reconciled view updated daily at the ad level. Every impression, every conversion, every marketplace in one place.

Red flags: You are relying on platform-reported ROAS as your primary decision input. Your Amazon and TikTok Shop sales are not reflected in your channel measurement. Your last measurement update was a quarterly report.

Fospha's Core — our always-on Daily MMM — is built to deliver exactly this. Updated every 24 hours at the ad level, it measures the full impact of every impression, view, and click across every channel and marketplace.

Question 2: Do you trust your numbers enough to let AI act on them?

The most common thing teams say when asked about AI automation is: "I'm not ready to hand the wheel to AI." That is a reasonable instinct — but it is often based on a misread of what automation actually looks like.

You are not handing control to an algorithm. You set the guardrails. The machine moves the budget within them. You see exactly what shifted and what it returned. And critically — trust is not a switch you flip. It is something you build. Gymshark did not go all in on day one. They started on one channel, validated the signal, and expanded from there. That incremental proof is what gave them the confidence to scale.

When that trust is in place, something changes internally. Marketing and finance stop arguing about whose numbers are right — because they are both working from the same model. Budget conversations get faster. New channel investments get approved. And your team stops spending their mornings in dashboards manually translating data into decisions, and starts focusing on what is actually next.

When a machine is making hundreds of budget decisions every week, the question is not whether it is fast. The question is whether it is validated. Measurement you cannot interrogate is a poor foundation for manual decisions. It is a catastrophic one for automated ones. Because automation does not introduce doubt into bad data. It removes it. The machine acts with confidence, at frequency, on whatever signal you have given it — including biases and gaps you did not know were there.

Most performance marketing teams face an uncomfortable choice today. Platform-native data is fast and easy to act on — but it is structurally biased toward bottom-funnel channels, does not cover marketplaces, and is produced by the same platforms with a commercial interest in the outcome. Quarterly MMMs carry scientific weight, but they arrive too late, and many operate as black boxes where teams are asked to trust outputs they cannot interrogate.

Neither builds the confidence required for AI automation at scale. Teams that cannot interrogate their measurement do not use it for the big calls. Measurement becomes a cost center — producing reports that inform conversations rather than driving outcomes.

What does transparent measurement actually look like?

There is a direct line between measurement you cannot trust and growth you cannot capture. When teams do not trust their numbers, budget decisions stall. New channels never get the investment they deserve. And when AI automation enters the picture, unresolved trust issues compound automatically, every day.

Think of it like football. The goal scorer grabs the headlines, but any good manager knows it takes the whole team to win. If you only paid players based on who touched the ball last, you would blow your budget on strikers and lose every game in midfield. Most measurement tools do exactly this: they credit the last click and ignore everything that created the opportunity.

Fospha's model works differently. It runs through a clear, step-by-step process — from GA4 validation to click measurement to impression modeling — so every channel gets the credit it actually deserves, and every stakeholder can see exactly how we got there. Not a black box asking for blind trust. A model you can interrogate, explain, and stand behind in a board meeting. When you know how the model works, you can stop second-guessing the outputs and start acting on them with confidence.

This is what we call Glassbox - Fospha's commitment to full transparency at every modeling layer.

For a full step-by-step walkthrough of the model — and why last-click systematically under-credits the channels that matter most — read our model spotlight on how Fospha measures the full funnel.

What good looks like: Marketing and finance working from the same numbers, with a model anyone on the team can interrogate and explain.

Red flags: Your measurement outputs live in a deck that gets shared but rarely challenged. Finance is working from different numbers than marketing. You have never had a conversation with your measurement provider about how the model actually works.

Question 3: What is the gap between your insight and your action, and what is it costing you?

Every brand has this gap. Teams log into dashboards, manually translate insights into budget changes, repeat across platforms, every day. That gap is where growth leaks.

A creative that was working yesterday is still running today, pulling budget it no longer deserves. A channel that showed saturation signals last week is still receiving the same allocation this week. A TikTok campaign went viral on Thursday. By Monday, when the manual process catches up, the moment is gone. The information existed. The action just came too late.

This gap, between what measurement tells you and what actually gets done, is where competitive advantage is won or lost. And it compounds. Every day that insight sits unacted on is a day the budget is misallocated.

The brands pulling ahead have closed this gap entirely. They feed trusted, daily measurement directly into execution platforms so AI optimization happens automatically, without a human copying numbers between tools. The result is not just speed. It is hundreds of small, intelligent budget adjustments every week, each one grounded in the most current signal available, each one building on the last.

That compounding is what creates durable performance advantage. Not one brilliant campaign decision. A relentless cadence of good ones, made faster than any manual process can match.

What good looks like: Daily measurement signals flowing directly into execution platforms, with AI-driven budget adjustments governed by guardrails your team defines. No manual step between insight and action.

Red flags: Budget changes happen weekly at best. Your team spends significant time translating measurement outputs into platform-level adjustments. The gap between what your data tells you and what your campaigns are doing is measured in days.

This is exactly what Gymshark and Smartly achieved together — feeding Fospha's trusted daily measurement directly into Smartly's cross-channel budget allocation, so that insight and execution are connected.

What all three look like in practice: Gymshark's results

Gymshark has been working with Fospha for few years now. The approach Daniel Green, Head of Digital Marketing at Gymshark, described at Shoptalk was deliberate rather than dramatic. They started in smaller markets first, testing to validate the signal before committing larger budgets. That testing was the foundation of trust, not a cautious half-measure. Once the data proved itself in smaller markets, they scaled with confidence.

The setup: Fospha's daily total commerce measurement feeding Smartly's cross-channel budget AI optimization, removing the platform-native data that had previously been constraining their decision-making.

The results reflect what happens when all three foundations are in place simultaneously:

These are not marginal gains from a single optimization. They are the compounding result of giving AI automation a complete, trusted, connected measurement foundation to act on — consistently, at daily cadence, across every major channel.

To close the session, Daniel reflected that data automation is something powerful, and Dom's response captured the room's feeling: it's the kind of thing you wish you'd done sooner.

The brands who get this infrastructure right now are the ones that will compound their advantage fastest. The ones that do not will find themselves increasingly outpaced by competitors who can optimize at a velocity no manual process can match.

The AI Optimization Readiness Checklist

Three things need to be in place before you can automate with confidence  and each one builds on the last.

1. Complete measurement: Total commerce coverage across every impression, channel, marketplace, and sales destination, every day. Full-funnel and ad-level. No gaps, no bottom-funnel bias. This is the foundation everything else requires.

2. Trusted science: Measurement that every stakeholder can interrogate. Not a black box. Not quarterly rigor that arrives too late. Transparent, continuously validated, and explainable to finance as well as the growth team. If you cannot show your CFO how the numbers are produced, you are not ready to automate against them.

3. Connected execution: No manual step between insight and action. Measurement data flowing directly into the platforms where budget decisions are made, with AI automation that moves fast enough to act on daily signals. This is where the compounding begins.

Miss any one of them, and your AI is amplifying the wrong signal. Get all three right, and automation stops being a risk and starts building compounding competitive advantage — every single day.

The performance gap between brands that have built this infrastructure and those still running fragmented measurement stacks is already opening. The good news: it is a solvable problem. And the brands that solve it first win.

Book a demo with Fospha to see where your measurement foundation stands today.

FAQ

What is AI budget optimization for Retail Commerce brands?

AI budget optimization for retail commerce is the use of machine learning and automated decision-making to allocate and adjust paid media budgets in real time — shifting spend across channels, campaigns, and ad sets based on daily measurement signals rather than manual review. When built on trusted, total commerce measurement, it allows brands to make hundreds of small, intelligent optimizations every week that compound into significant performance gains. When built on fragmented or biased data, it amplifies those errors at scale.

What is the difference between Daily MMM and a traditional Media Mix Model?

A traditional Media Mix Model (MMM) is typically run quarterly or annually. It provides strategic direction on channel contribution but arrives too late to inform daily decisions, and often lacks the ad-level granularity that AI automation platforms require. A Daily MMM, like Fospha's Core, delivers the same strategic rigor of a full MMM, updated every 24 hours at the ad level. This means brands get the scientific credibility of an MMM with the speed and granularity needed for AI-powered budget optimization.

Why can't click-based measurement or last-click attribution support AI automation?

Click-based measurement or last-click attribution capture only a fraction of the channel path, specifically the final click before conversion. They systematically ignore impressions, views, and upper-funnel activity, which means channels like paid social, video, and display are routinely under-credited. When AI automation acts on this biased signal, it over-invests in demand capture and under-invests in demand generation compressing the funnel and limiting long-term growth. Total commerce measurement corrects this structural bias.

How do marketplaces like Amazon and TikTok Shop affect marketing measurement?

Amazon and TikTok Shop cannot be tracked with pixels or cookies, which means they are invisible to most click-based measurement tools. Yet Amazon now accounts for 40% of US eCommerce, approximately $1.75 billion in daily revenue. Any measurement system that does not account for marketplace sales is missing a substantial share of the commercial impact of paid media. Total commerce measurement, which unifies DTC, Amazon, TikTok Shop, and all paid channels into a single view, is the only way to see the full picture.

How long does it take to see results from connecting measurement to AI automation?

Gymshark started in smaller markets to validate the signal before scaling. Once the measurement proved itself, they committed larger budgets with confidence. Brands typically see meaningful ROAS improvements within the first months of connecting trusted daily measurement to execution platforms. The key is building the foundation correctly before scaling — total commerce measurement first, trusted validation second, connected execution third.