Black Friday hasn’t been a weekend for a long time, but this year made that impossible to ignore. What looked predictable from the outside was, inside accounts, a rolling cycle of volatility that refused to behave neatly. Most of the public commentary settled for the usual story: early demand, cautious consumers, headline ROAS increases. All of the usual comfortable stories, but none of them close to what performance teams were actually dealing with.
Inside the accounts, the real picture was shaped by how products moved through campaigns (or not). None of the commentary centred around SKU exposure or range behaviour, even though these were the levers that actually defined success this year for retailers. Agencies and platforms stayed with their familiar surface metrics, they didn’t touch the parts of the media world that reveal real performance. But those running a live account saw the truth immediately. A handful of SKUs took most of the budget. Mid range products made some contribution and were in and out of advertising auctions. New or newly discounted products often struggled to get meaningful traction in time and entire sections of the retail range remained undiscovered.
It’s not surprising this layer gets ignored. SKU level analysis is time consuming and it demands clean data, honest attribution, and a willingness to admit where account structures are no longer fit for purpose (performance structures rather than category). But if you were actually running peak this year, you knew the story was written at SKU level, not in topline ROAS charts.
Consumer behaviour followed the same pattern. Shoppers didn’t browse, they arrived ready to buy, or ready to leave. Discovery occurred up the funnel and decision making on who to buy from was compressed. A product page was no longer part of a journey; it was the moment everything resolved. Retailers prepared for that moment won.
In Upp.ai’s client base the year on year picture in November told the same story. Retailers who kept their accounts warm and let AI scale to market demand, those who stayed visible across their full range, who didn’t pull target ROAS levers constantly and throttle budgets at the wrong moments, started the month in a much stronger position than they did the year before.
In our retail customer base, the year-on-year picture across November reinforced this shift. On average, retailers increased spend by around 30%, with total spend managed up 67% year on year across the peak period. Despite that scale, ROAS held flat.
Cost per click fell by around 6% year on year, which is exceptional in a market where many retailers cite annual CPC inflation of 15% or more. At the same time, click-through rates increased by roughly 11%, signalling stronger auction positioning and more prominent product visibility in the auctions. This wasn’t about pulling back from competition, it was a reflection that products are being placed into the right auctions, achieving greater share of presence, and being pushed harder as demand emerged.
The consumer market didn’t behave differently. The retailers [working with Upp.ai] did, and so did their range, because they weren’t spending half of November trying to derail (optimise) their PMax campaigns, their momentum carried forward as PMax was able to align with more customer journeys and not be constrained by red flags like ‘Limited by Budget’. We see profound performance changes when you don’t spend the first two weeks undoing your own structure or changing the targets for PMax to go after.
Then came the part of peak that almost nobody talked about: the days after Cyber Monday. Demand didn’t fall away, yet advertising did. Teams shifted out of peak mode because the calendar said it was time to, but the market was still there in force with high intent traffic continuing. CPCs softened, conversion stayed strong and therefore overlooked SKUs finally became economically viable to retail through advertising. The retailers who are staying, not making claw backs in their budgets are present in the market and capturing one of the most profitable windows of the year.
Manual optimisation didn’t stand a chance in this environment. Not because teams lack experience, but because retailers are stuck in cycles of reporting, adjusting, and internal firefighting. For Upp.ai our clients experienced momentum that stayed, as learning was constant, visibility of SKU’s was maintained and performance achieved.
We saw this most clearly with retailers we managed across both years. Last year their December dropped the moment budgets were reset. This year, the same retailers carried momentum straight through and made some strategic decisions. Nothing dramatic changed, these were the same retailers, with pretty much the same product ranges, with the same demand and pretty much the same competition, the only difference was that they didn’t break their stride, and that approach is altering their December performance.
If this year showed anything clearly, it’s that Black Friday didn’t expose media problems. It exposed operating model problems, how retailers optimise and make decisions with their performance media. The market behaved rationally, but retailers often didn’t. Everything that drove growth sat in the gap between a new breed of behaviours that are needed for AI led media and the old school way that they’ve always optimised digital media.
And if you think that was the lesson of peak, it wasn’t. The real truth reveals itself in the weeks that follow. Read more here.