Ecommerce advertisers paid roughly a third more for impressions in the week around Black Friday 2025, yet the median expense of winning a single sale fell to its lowest weekly level of the season, according to Billy Grace, the Amsterdam-based measurement company that published its 2026 Black Friday Playbook on September 22, 2026.

In Short

A company that measures advertising results looked at what online shops spent and sold between September and December 2025, and found that Black Friday week produced cheaper sales, not dearer ones, even though the ads themselves cost more. That matters for anyone who trims ad budgets at the peak on the belief that returns collapse, although the finding leans on a model that credits ads people saw but never clicked. What changes is the argument over which measurement system gets to steer peak-season budgets, not the price of advertising in November.

Cheaper orders in a dearer auction

The playbook draws on aggregated first-party performance data from ecommerce and omnichannel advertisers on the Billy Grace platform between September 1 and December 31, 2025. Released from London ahead of the 2026 shopping season, it takes aim at what the company calls "one of the most persistent assumptions in retail marketing": that Black Friday week compresses returns as advertisers pile into the same auctions for the same customers. The full report breaks the data down across nine paid channels: Google, Meta, Bing, YouTube, TikTok, Pinterest, Snapchat, Reddit and Criteo.

Through September and October, median CPMs across the cohort sat in a tight band around EUR 6.50, according to Billy Grace. They began lifting in early November and peaked in Black Friday week, which the company defines as November 24 to December 1, 2025, at EUR 8.83. That is about 36% above the autumn baseline; the playbook rounds it to "roughly a third". Weekly tracked revenue climbed to nearly 80% above its September level in the same week.

Returns moved the other way. Median return on ad spend rose about 26% from the pre-peak period, which the company calls the runway, into Black Friday week. What it cost to secure each order came in roughly 18% under the runway level, the lowest weekly figure of the four-month window.

Billy Grace attributes the gap to conversion. "More expensive impressions still produced cheaper orders, because a larger share of those impressions turned into purchases," the playbook states. It publishes neither conversion rates nor order values, but its medians allow a rough check. If a thousand impressions cost about 36% more while each order cost about 18% less, orders per thousand impressions would have had to rise by something like 60% to 70%, and order values by around 3%. Medians calculated advertiser by advertiser do not combine exactly, so both derived figures are indicative.

One qualification sits underneath all of this. The 26% and 18% movements are reported "measured with UMM", the company's unified model. The playbook says Black Friday returns "look far weaker" under Last Click attribution but does not publish the cohort-wide Last Click equivalents, so the size of that difference cannot be checked.

Every channel got dearer, some far more than others

Search remained the most expensive inventory in absolute terms, with Google and Bing CPMs near EUR 13 in Black Friday week. In relative terms, social and video inflated hardest: Snapchat CPMs rose 72% against the runway, YouTube 51%, Meta and TikTok 45% each, Criteo 38%, Pinterest 26%, Bing 25% and Google 22%, the smallest rise of the eight channels charted. Reddit does not appear in that chart. The channels that "teams treat as cheap prospecting lose the most of their price advantage exactly when demand peaks," the playbook notes.

Other readings of the season point the same way. Integral Ad Science measured a 219% rise in impressions between November 2 and December 5, 2025. AdRoll recorded display retargeting CPMs up 11% year over year in the fourth quarter of 2025, and Meta's results for the quarter showed ad impressions up 18% and average price per ad up 6%. On the demand side, Adjust data put Black Friday 2025 shopping app installs 53% above the yearly average in Europe, with sessions 62% higher.

Black Friday itself carried the week

Daily orders did not hold a flat high. Indexed to Monday, November 24, at 100, the median advertiser's orders ran at 98, 105 and 98 over the next three days, jumped to 150 on Black Friday, November 28, dipped to 102 on Saturday, recovered to 132 on Sunday and closed at 114 on Cyber Monday, December 1. CPM and cost per click topped out on Black Friday and eased afterwards.

On Black Friday the median advertiser recorded 30 orders, at a median cost per order of EUR 27.10, against roughly EUR 38 on the quieter days of the week, according to the report. That is about 29% cheaper on the busiest day. The index also gives a sense of scale: 30 orders at 150 implies around 20 orders on the Monday that opened the week. The typical account in this cohort counts orders in tens per day, not thousands, which suits a method designed so that "no single large account dominates".

Steven Marsé, Insights Manager at Billy Grace, framed the result in the company's release: "We built this report because every peak season, marketers are told the same story: hold your nerve on Black Friday and returns collapse. Our data says the opposite. The busiest day of the year was also the most efficient, and the channels that looked weakest under last-click were often the ones bringing in the most new customers. If you're planning 2026 budgets on last-click numbers, you're planning against the wrong picture."

The playbook's own wording is narrower. It says Black Friday delivered "both the most orders and the cheapest orders of the week", and the dataset covers four months, not a year.

Where the credit goes under each model

The report runs the same peak-week spend through three attribution views. Last Click assigns the sale to the final touch before purchase. Multi-touch attribution (MTA) shares credit across tracked clicks. Unified marketing measurement (UMM), Billy Grace's product, combines multi-touch attribution with marketing mix modelling to credit impressions and view-through conversions that click-based models cannot see. MTA lands between the other two, though for view-heavy channels such as YouTube, Meta and Pinterest it "still tracks closer to Last Click than to UMM", according to the report.

The differences are large. Under Last Click, Meta, TikTok, YouTube, Pinterest, Snapchat and Reddit all "read as loss-making in Black Friday week". Under UMM every one of them clears break-even. Measured by peak-week revenue credited, UMM assigns Pinterest 9.2 times what Last Click does, Reddit 7.8 times, YouTube 7.4 times, TikTok 4.5 times, Meta 2.2 times and Snapchat 2.0 times. Google's credited revenue "barely changes" and Bing's is "effectively flat", because both were already collecting the final click. Under UMM, Google's return improved slightly into the peak and Bing's rose by more than a fifth.

The labelling is loose. The chart presents each multiple as "more" revenue, Criteo included at 1.0 times, while the channel section describes Criteo as crediting "roughly the same revenue as Last Click". Read with the text, which refers to "more than nine times the peak-week revenue that Last Click does" on Pinterest, the figures are ratios rather than increments.

A model built to find what clicks miss

PPC Land noted in August, covering Billy Grace's first-half 2026 benchmark, that a finding in favour of impression-led channels rested on a model designed to credit impression-led channels. The same dependency runs through the playbook more sharply, since its central claims are defined by the gap between two models.

Peak season is also where aggregate modelling faces its hardest test. Budgets rise before Black Friday because sales are expected to rise, a form of endogeneity that pushes observational estimates of advertising's effect upward. A preprint by a Zalando researcher found that a standard mix model using the default seasonal controls of popular open-source frameworks reported 10.61 times return on ad spend for a simulated paid search channel whose true return was 4.20. The playbook does not describe its seasonal controls, and it reports no holdout studies or lift tests used to calibrate the UMM figures.

The direction of the finding is familiar. Kochava research in September 2025 found marketing mix modelling credited TikTok campaigns with 35% higher incremental impact than last-touch attribution, and Meta's 307-study white paperargued that last-click attribution misrepresents where conversions come from. Both came from companies selling the measurement or the media that the argument favours.

Most peak orders came from first-time buyers

New customers accounted for about 66% of orders across the Black Friday and Cyber Monday window, according to Billy Grace, and that share "barely moved" into the pre-Christmas period.

By channel, under UMM, the share of orders from first-time customers in Black Friday week was 88% on Reddit, 77% on Pinterest, 67% on YouTube, 66% on TikTok, 56% on Meta and 46% on Google. Every channel peaked in that week. The same figure then shows each channel falling in December below its own runway level: Reddit to 71% from 82%, Pinterest to 63% from 72%, YouTube to 56% from 63%, TikTok to 54% from 61%, Meta to 42% from 49% and Google to 37% from 40%.

Two tensions sit in those numbers, and the report reconciles neither. Each of the six channels fell between 9 and 17 percentage points from Black Friday week into December, while the cohort-wide share is said to have held. And a cohort-wide 66% sits well above the Black Friday week shares for Google and Meta, the two channels that take the majority of spend. Medians computed separately can diverge in this way, and orders outside paid channels may carry a different mix, but the playbook does not show how the figures fit together.

The report adds that "new-customer CAC" inflates "faster than blended CPA" under auction pressure, without figures for either. Platforms have been building for the same objective: Google added new customer acquisition modes to its customer lifecycle goals in April 2025, and its retail holiday guide of August 21, 2026 listed the New Customer Acquisition goal among recommended products.

Orders against spend

Peak week is also a question of scale. Google recorded the clearest result, according to the report: daily orders grew about 62% from the runway into Black Friday week while spend grew about 30%. Combined with its 22% CPM rise, those medians imply roughly 7% more impressions and a cost per order about 20% lower.

Bing, Meta and YouTube also sat on the side of the chart where orders outpaced spend. Criteo and Reddit "absorbed large budget increases into relatively small order gains", while TikTok stayed nearly flat on both, which the company reads as a priming role rather than last-week volume. The chart carries no axis values, so Google's are the only figures published for the comparison.

Reddit's placement contrasts with the platform's own account of that quarter. Reddit reported fourth-quarter 2025 advertising revenue of $690 million, up 75%, with revenue from lower-funnel objectives such as purchase conversions doubling, and cited a 91% improvement in return on ad spend for Dynamic Product Ads. Billy Grace describes Reddit in its cohort as "a pure prospecting channel".

Share of wallet, and captions the charts do not support

Google and Meta took the majority of spend, with Google leading, according to the playbook, which does not publish either share. The other seven channels' combined share of wallet rose from 9.6% in September to 11.8% in October and 12.9% in November, before easing to 12.4% in December. Billy Grace's first-half benchmark recorded Google's spend-weighted share falling from 62.1% to 57.3% in a differently defined cohort, so both datasets point towards smaller platforms.

Three statements in the text are not borne out by the charts printed beside them. The report says Reddit, "new this year", enters the mix "small but visible in November and December", yet the share-of-wallet chart shows a Reddit segment in the September and October bars too, at roughly 0.8% and 1.4% by eye.

YouTube and TikTok are described as "roughly doubling their combined slice between September and December". Read from the monthly chart, their combined share moves from about 3.9% to 5.0%; the weekly chart shows about 3.2% to 4.9%. Neither is a doubling.

The weekly chart's caption says both video channels "hold it right through December, and do not retreat after Cyber Monday". On the chart, YouTube's share peaks near 3.6% in the week of November 17, before Black Friday week, and eases to around 3.0% by the week of December 29. None of these readings alters the headline medians, but all three support the budget recommendations.

After Cyber Monday

From December 2, orders eased off the Cyber Monday high but stayed well above the autumn baseline, peaking each Sunday before tapering towards the Christmas Eve shipping cut-off, according to Billy Grace. As auction pressure came off, median cost per order fell back and returns held "close to Black Friday week levels on the strongest days". The playbook calls the period "a second window of profitable demand, not a comedown".

The season-long weekly chart goes further. Its final point, the week beginning December 29, shows median return on ad spend above the Black Friday week reading while revenue sits well below the peak. That week falls outside the report's three phases, which end on December 24, holds only three days of the measurement window, and goes unremarked in the text.

The 2026 plan carries 2025 dates

The playbook closes with a phase plan. For the runway, it recommends having first-party tracking and UMM live before November, funding upper-funnel and video activity, and locking creative and budgets in the first half of November. For Black Friday week, it recommends steering on cost per order or return on ad spend rather than CPM, adapting return floors so bidding is not starved on the most efficient days, and protecting social and video budgets. For pre-Christmas, it recommends a second push, keeping video on and triggering retention flows for new buyers.

One Black Friday week instruction stands out: "switch to MTA as the Attribution Model for real-time budget/campaign optimisation." That is the click-based model the same document says tracks closer to Last Click for view-heavy channels. The model behind the headline conclusions, in other words, is not the one proposed for in-flight decisions on the highest-spending days, and the playbook does not address the point.

The channel sheet assigns roles: Google as conversion engine, with temporary return targets below benchmark; Bing as low-risk scale; Meta as acquisition amplifier, with spend surged in the two weeks before Black Friday to fill retargeting pools; YouTube as upper-funnel scale; Pinterest as defensive prospecting; TikTok as a priming channel whose budget shifts away during the peak; Reddit and Snapchat as emerging prospecting and reach. Criteo, a retargeting layer that "saturates fast", is to be capped because extra peak budget "mostly raises its own auction price rather than adding orders".

The calendar inside the plan is last year's. The checklist dates Black Friday week as "24 November to 1 December" and the pre-Christmas phase from "2 December". In 2026, Black Friday falls on November 27 and Cyber Monday on November 30, so the equivalent week runs from Monday, November 23, to Monday, November 30.

The phasing echoes other peak-season work. A programme run by Fospha with Google between October 29 and December 16, 2025 found that retail brands allocating 10% to 20% of Google budget to Demand Gen recorded double the return on ad spend of those allocating under 5%, with one brand, Derek Rose, building demand before the peak and moving that budget to Performance Max and brand search on the highest trading days.

Method, and what the figures leave out

Metrics are computed per advertiser and reported as the median across advertisers, with phase changes comparing each advertiser with itself, according to the methodology appendix. A channel is reported only where a minimum number of advertisers ran it, and offline channels are excluded. Attribution uses a 30-day lookback window. CPM, click-through rate and cost per click are identical across all three models by construction.

Several elements are missing. The playbook gives no advertiser count, no total spend, no category breakdown and no value for the minimum-advertiser threshold. The data "is not split by country", although the release is a UK release datelined London and every monetary figure is in euros. The PDF version of the release prints its dateline year as "20206"; the emailed text reads 2026.

Billy Grace describes itself as an "Amsterdam-based, Fortino Capital-backed attribution platform". It was founded in Amsterdam in 2022 and expanded into the United Kingdom in 2026. The playbook ends with an invitation to book a meeting, and the UMM model at the centre of its findings is what the company sells. The figures are vendor-supplied, and the report cites no independent audit.

Why this matters for the marketing community

Peak-season planning for 2026 arrives with a stack of vendor material, from Google's holiday guide to Teads' exclusive 10-day home screen takeover on V smart TVs and Integral Ad Science's warnings on inventory quality. Billy Grace's playbook differs in kind: it quantifies, channel by channel and for the peak week itself, how far two attribution models diverge.

That gap is the material point. A team reading Last Click would see six of nine channels losing money in Black Friday week; a team reading UMM would see all six clear break-even. Budgets set at the largest spend levels of the year follow whichever number sits on the dashboard. Other datasets tell other stories: Channable's analysis of EUR 1.38 billion in Google Ads spend found return on ad spend falling 46% on Performance Max between June 2025 and June 2026. Populations and models differ, which is why the pictures do.

Could higher conversion rates outrun higher media prices without any help from the model? In principle, yes: an advertiser's total orders and spend are the same whichever channel gets the credit. What depends entirely on UMM is the channel ranking, and with it the case for protecting social and video budgets at the peak. Without an advertiser count, disclosed seasonal controls or experimental calibration, that part of the argument remains the company's reading of its own model.

Timeline

Summary

Who: Billy Grace, an Amsterdam-based attribution and measurement company backed by Fortino Capital, with commentary from Steven Marsé, its Insights Manager. The data covers ecommerce and omnichannel advertisers on its platform, with results reported as medians across advertisers.

What: The 2026 Black Friday Playbook, which found median CPMs rising from about EUR 6.50 to EUR 8.83 into Black Friday week 2025 while median return on ad spend rose about 26% and the median outlay per order fell about 18% under the company's unified marketing measurement model. New customers made up about 66% of orders in the Black Friday and Cyber Monday window. Under Last Click, six of nine channels read as loss-making in peak week; under UMM each of those six cleared break-even, with Pinterest credited 9.2 times the Last Click revenue.

When: The data covers September 1 to December 31, 2025, with Black Friday week defined as November 24 to December 1, 2025. Billy Grace published the playbook and release on September 22, 2026, ahead of Black Friday on November 27, 2026.

Where: The release was issued from London, the figures are in euros, and the data is not split by country. It covers nine paid online channels: Google, Meta, Bing, YouTube, TikTok, Pinterest, Snapchat, Reddit and Criteo.

Why: The report challenges the assumption that Black Friday compresses returns and argues that last-click measurement makes social and video channels look loss-making at the moment they are acquiring most new customers. Its conclusions depend on a proprietary model combining multi-touch attribution with marketing mix modelling, for which cohort size, seasonal controls and experimental calibration are not disclosed, and several of its chart captions are not supported by the charts themselves.