How a UK cycling retailer grew revenue 23.6% year on year on just 5.6% more spend, with AI-assisted analysis and 304 product exclusions doing the heavy lifting.
What did AI performance infrastructure achieve for this retailer?
A UK cycling retailer grew revenue 23.6% year on year on only 5.6% more ad spend, with 17% higher ROAS, an 89% lower CPA and 857% conversion growth, driven by Social Surge's AI-assisted account analysis and the exclusion of 304 zero-conversion products.
This UK cycling retailer was already advertising and already profitable. The question was the one every maturing e-commerce account faces: how do you keep growing without simply buying growth? Increasing spend 20% to gain 20% more revenue isn't strategy, it's arithmetic, and it stops working the moment margins are squeezed.
Cycling retail adds its own complications: large SKU catalogues spanning £5 inner tubes to four-figure bikes, sharp seasonality, and price-comparison shoppers who make broad targeting expensive. In accounts like this, the difference between good and great performance lives at the product level, and the product level is exactly where human-only management runs out of hours. No account manager can review thousands of SKUs weekly. So most agencies don't. Spend drifts to products that will never pay for it, one click at a time.
This account runs on our AI performance infrastructure: a managed intelligence layer that continuously analyses account data, product performance, search terms, budget pacing, anomalies, at a depth and frequency no human team can sustain. The AI does the reading; our team makes the calls. Every recommendation is reviewed before it touches the account, which is precisely why it works: machine-scale analysis, human-grade judgement.
The flagship move was product exclusion at scale. The analysis identified 304 products that were absorbing ad spend and had produced zero conversions. Individually, each looked like noise, a few pounds here and there. Collectively, they were a steady tax on the whole account, dragging down ROAS and starving proven sellers of budget.
We excluded all 304 from Shopping. The effect was immediate and measurable: Shopping ROAS rose from 12.09x to 16.20x, a 34% improvement, because every reclaimed pound flowed to products with demonstrated demand. Average order value climbed too, from £33.41 to £51.72 (+54.8%), as budget shifted away from low-value catalogue clutter and toward the products that actually build a business.
Product exclusion isn't a single cleanup, catalogues change, seasons turn, yesterday's dead SKU becomes next spring's seller. The infrastructure re-evaluates continuously, flagging new zero-converters and candidates for re-entry, so the account gets cleaner over time instead of slowly silting back up.
Product exclusions were the headline, but the infrastructure earns its keep in the accumulation of smaller decisions it surfaces every week: search terms drifting off-intent, budget pacing running hot or cold against the seasonal curve, anomalies in conversion data that would otherwise sit unnoticed until month-end reporting. Each flag arrives with context, what changed, when, and what it's costing, so decisions that used to take an afternoon of spreadsheet work take minutes. Over a full year, in a seasonal category like cycling, that speed compounds into a genuine competitive edge: the account adapts in days while competitors adjust in quarters.
Four times more revenue growth than spend growth. That gap, growth outpacing budget, is the entire point of performance infrastructure.
Most ad accounts don't need more budget; they need more attention than humans alone can give. AI infrastructure makes product-level precision affordable at catalogue scale, and in niches like cycling, where we specialise, that precision is the margin between funding growth and funding waste. Book a free audit and we'll run the same product-level analysis on your account.
Book a free audit and we'll run product-level analysis on your catalogue, including a zero-conversion product check.
Book Your Free Audit