Black Friday can mislead an analyst twice. On peak day the data shows too little, because conversions in the ad platforms are still trickling in. A week later it shows too much, because record revenue hides the margin given away in discounts and the sales borrowed from neighbouring weeks. This guide organises data work around the peak into three phases: preparing measurement, the sale day live, and a reckoning you can trust.
Phase 1 · before
Baseline: what to compare Black Friday with
Without a baseline, every peak result is a “record” and nothing follows from it. Before the season, set two comparison baselines, both from your own data:
- Last year’s peak, per category: not just the overall result, but the hourly distribution of the day, category shares and bestsellers. This curve tells you when to expect spikes and what must be in stock (how to calculate it: the guide to Q4 planning).
- Ordinary weeks just before the season: the baseline for measuring the borrowed sales effect. You will need it after the peak, so it has to be “frozen” in advance: average weekly sales and margin from a period without promotions.
Phase 1 · before
Measurement checklist before the peak
During the peak, a tracking failure costs the most and is the hardest to notice, because all the numbers are going wild anyway. That is why the measurement set-up is checked before, not during:
Measure normal GA4 coverage
The number of purchase transactions in GA4 divided by the number of paid orders in the shop for the same days, over recent weeks. This figure is your benchmark: if coverage drops clearly below normal during the peak, it is tracking that is broken, not sales. Step-by-step method: the guide to the GA4 audit.
Clean up product statuses in Merchant Center
Fix disapproved and expiring products while traffic is cheap. A bestseller disapproved on Thursday evening will not be back in the auctions for Friday’s peak.
Calculate bestseller stock cover at peak sales rate
Stock divided by daily sales from last year’s peak, not by the October average. Products without cover for the whole weekend drop out of campaigns earlier (the mechanics: the guide to stock levels).
Set alert thresholds and a monitoring schedule
What counts as an alert (a drop in GA4 coverage, a jump in unpaid orders, bestseller stock below the threshold) and who watches the dashboard at which hours. On peak day there is no time to invent procedures.
Phase 2 · peak day
Peak day: what to read live and what to leave
Sale day has its own rhythm and its own set of metrics. The hourly curve from the previous year suggests when to look (illustrative data, an electronics shop):
Orders every 3 hours · peak day vs an ordinary Fridayelectronics shop · illustrative data
Two peaks instead of one: a jump just after midnight (the promotion starts, customers arrive with ready baskets) and a long evening peak. Between them sits a morning dip that every year tempts people to “fix something in the campaigns”. The curve says: this is the normal shape of the day, wait.
The day dashboard splits metrics into two groups: those that are complete straight away and allow you to react, and those that are still misleading on peak day:
Read live
Leave for after the peak
An alert that really is an alert
A sudden drop in GA4 coverage at peak is a tracking failure, and every hour of delay damages the season’s data. That is why coverage sits on the dashboard next to orders: sales according to the shop are rising, yet analytics suddenly “loses” orders? You react immediately instead of explaining odd reports on Monday.
Phase 3 · after the peak
Post-peak data matures as the windows close
A Saturday morning report shows only part of Friday’s result. Ad platforms add conversions with a delay: click-through and view-through windows, modelling where consent is missing, customers returning after a weekend of thinking it over. The share of peak-day conversions visible in the platforms grows roughly like this (illustrative data):
Peak day
raw reading in the evening+1 day
window conversions arrive+3 days
post-weekend returns added+7 days
picture complete, time for verdictsThere is one practical conclusion: verdicts on campaigns and channels are made about a week after the peak. A campaign “cut” on Saturday for weak ROAS was usually fine; its result simply had not arrived yet. The mechanics of windows and modelling are covered in the guide to conversions and attribution; on peak day it is enough to remember that this effect is strongest precisely when volume is highest.
Phase 3 · after the peak
The verdict: margin after discounts and borrowed sales
The borrowed sales effect
A sale adds some orders and merely shifts others in time: customers hold off before the peak and do not come back after it. That is why the result is judged on the total of the weeks around the peak against the baseline, not on the sale day alone.
The weekly chart makes this crystal clear (sales index, 100 = an ordinary baseline week, illustrative data):
Weekly sales index · 100 = pre-season baselinedips before and after the peak are borrowed sales
Peak +115%, period +6%. The five weeks total 529 points against 500 at baseline: the real growth over the whole period was just under 6%, even though the peak week alone looked like the business had doubled. Both numbers are true; only the second tells you whether the promotion added sales.
The second pillar of the assessment is margin after discounts. A discount comes out of profit, not revenue: at typical margins, giving away a dozen or so per cent of the price takes a much larger slice of margin (the calculation: guide to discount code profitability). That is why record peak revenue is set against margin calculated on discounted prices, per category: it usually turns out that some “sale hits” delivered turnover at a token profit, while the season was saved by products bought alongside, at full price.
Methodological note: all figures in the examples are illustrative. The shape of the daily curve, the pace at which data matures and the depth of the dips around the peak vary between categories; your own values come from previous seasons, and the comparison baseline is set before the promotion starts, not after the fact.
DataOrganizer
How DataOrganizer supports data work at peak
The whole routine described here rests on data that DataOrganizer keeps in one model: paid orders, spend across all platforms, margin, stock levels and GA4 events, with automatic anomaly detection. The day dashboard requires no stitching of exports, and the post-season assessment fits into a single question:
Sale weekend compared with an ordinary weekend:
| Metric | Peak weekend | Ordinary weekend | Ratio |
|---|---|---|---|
| Revenue (paid) | €103,000 | €24,000 | ×4.3 |
| Effective discount | 12% | 2% | – |
| Margin after discounts | €18,750 | €6,725 | ×2.8 |
| Return on sales | 18.2% | 28.0% | 1/3 lower |
Scale delivered the result: margin grew 2.8 times despite profitability falling by a third. The total for the five weeks around the peak came in 6% above the baseline, so most of the pre-peak dip was recovered with a surplus. A note for next season: two categories sold at the peak with profitability below 8%; they are candidates for shallower discounts.
Key takeaways
Black Friday on data means three different disciplines in three phases. Before the peak, you check tracking and freeze the baseline. During the peak, you read the shop and watch coverage. After the peak, you wait for the data to mature and settle the margin over a wide window. Mixing up the phases costs the most.
FAQ
Frequently asked questions about Black Friday data
How do you prepare shop data for Black Friday?
One to two weeks before the peak: measure normal GA4 coverage (a reference for detecting tracking failures), clean up product statuses in Merchant Center, calculate days of stock for bestsellers at last year’s peak pace and freeze the comparison baseline (normal weeks and last year’s peak, per category) together with alert thresholds.
What should you track live on sale day?
Whatever is complete straight away: paid orders and shop revenue, total spend and the day’s MER, bestseller stock levels, GA4 coverage and the share of unpaid orders. Leave individual campaign ROAS and attribution alone: on peak day these figures are incomplete by definition.
When should campaigns be assessed after the peak?
Once the attribution windows have closed, in practice about a week after the peak. Ad platforms keep adding conversions for several days, so a Saturday report systematically understates Friday’s results. Cutting campaigns on raw day-zero data is the most expensive reflex of the season.
How do you work out whether Black Friday paid off?
With two calculations at once: margin after discounts (the discount comes out of profit, so record revenue can be worth little) and the total of the weeks around the peak compared with the baseline, which shows how much the promotion added in sales and how much it merely shifted in time.
How does DataOrganizer help?
The daily dashboard (orders, MER, stock, GA4 coverage) works without stitching exports together, data anomalies are flagged automatically during the peak, and after the season the AI assistant settles the result on margin after discounts and compares the weeks around the peak with the baseline, per category and product.
Cheat sheet for the whole guide
Black Friday on data: three phases, three rules
1. Before: check tracking and freeze the baseline. GA4 coverage, Merchant Center statuses, days of stock at peak pace and a reference point set before the promotion starts.
2. During: read the shop, leave the ad dashboards. Orders, the day’s MER, stock and tracking coverage live; verdicts on campaigns and channels only once the data has matured.
3. After: settle the margin over a wide window. Margin after discounts instead of revenue, and the total of the weeks around the peak instead of the record alone. DataOrganizer keeps all this data in one model, so each phase comes down to a glance at the dashboard or a single question in the chat.
DataOrganizer · a dashboard for peak season
Connect your shop, ad platforms and GA4: orders, MER, stock, tracking coverage and margin after discounts in one place, with anomaly alerts. Demo account available instantly; you pay only for connected sources.