Every December, store owners open the sales report, see the Black Friday spike, and call the event a success. The spike is real. What it means is less clear, because it contains at least four different things: shoppers who would never have bought without the offer, shoppers who would have bought a few weeks later anyway, shoppers who held off in early November because they knew a sale was coming, and regular customers who got a discount on an order they were going to place at full price.
Only the first group is new money. A proper Black Friday sales analysis separates it from the other three. Circana, which measures promotions for large consumer brands, puts the risk plainly: a promotion that shows a 12 percent bump in its own window can net zero incremental annual sales once the post-period slump is counted. The same arithmetic applies to an online store, and you can do it with nothing more than your order table.
What the spike is made of
Think of your daily sales as a line with a baseline under it: what you would have sold without the event. The event changes the line in three places, not one.
The fourth component, subsidy, does not show up on the line at all. A loyal customer who places their usual order during the sale adds nothing to volume. They just pay less for it. You only see subsidy when you look at who bought, not how much.
| Component | What it is | Where it shows up | How to measure it |
|---|---|---|---|
| Gross lift | Event sales above the baseline | The spike | Event sales minus baseline for the same days |
| Pull-backward | Shoppers delaying purchases to wait for the sale | A dip in the 2 to 3 weeks before | Baseline minus actual in the pre-window |
| Pull-forward | Purchases moved earlier into the event | A dip in the weeks after | Baseline minus actual in the post-window |
| Subsidy | Discounts on orders that would have happened at full price | Margin, not volume | Event orders from customers who were due to reorder anyway |
| Net incremental sales | The new demand the event created | Nowhere on a standard report | Gross lift minus pull-backward minus pull-forward |
Step 1: Build a baseline you can defend
Everything depends on the baseline, and the common shortcuts are weak. Comparing Black Friday to the week before ignores that the week before is already depressed by waiting shoppers. Comparing it to last year’s Black Friday mixes in a year of growth, price changes and a different calendar: Black Friday fell on November 28 in 2025 and falls on November 27 in 2026, so a same-date comparison lines up the wrong weekdays. Day of week matters more than most people expect. Metorik’s 2026 WooCommerce benchmark found that average order value peaks at $116 on Monday and Tuesday and falls to $86 at weekends, so any baseline has to compare like days with like days.
Three approaches hold up, roughly in order of effort:
- Seasonal forecast. Fit a forecast on daily sales from September through early November, with weekly seasonality, and project it through the event and post-window. This is the most general option. QuantumLayers describes how Holt-Winters and SARIMA forecasts handle the weekly pattern.
- Prior-year shape, current-year level. Take last year’s daily pattern around the event, aligned by weekday, and scale it to this year’s September and October level. Works well for stores with a stable mix.
- Untouched comparison group. If some products or regions were excluded from the sale, their sales trend over the same weeks is a natural control. Circana recommends this kind of comparison over before-and-after views. Beware that excluded products can lose sales to promoted ones, which would make the event look better than it was.
Step 2: Measure the windows on both sides
Pull-backward shows up as a dip in the two to three weeks before the event, and it has grown as shoppers have learned the calendar. Pull-forward is the larger of the two for most stores and lasts as long as your typical repurchase cycle. For consumables (coffee, supplements, pet food) the post-window should cover at least one full replenishment cycle, often four to six weeks. For durable goods, the dip is spread over months and is hard to separate from normal seasonality; there, focus on subsidy and on the quality of new customers instead.
The daily series is easy to pull. On WooCommerce, the analytics table wc_order_stats holds one row per order with net sales and a returning-customer flag:
-- Daily orders, net sales and coupon discounts, Oct 1 to Jan 15
SELECT DATE(s.date_created) AS day,
COUNT(*) AS orders,
SUM(s.net_total) AS net_sales,
SUM(CASE WHEN s.returning_customer = 0
THEN 1 ELSE 0 END) AS new_customer_orders,
COALESCE(SUM(c.discount), 0) AS coupon_discount
FROM wp_wc_order_stats s
LEFT JOIN (
SELECT order_id, SUM(discount_amount) AS discount
FROM wp_wc_order_coupon_lookup
GROUP BY order_id
) c ON c.order_id = s.order_id
WHERE s.parent_id = 0
AND s.status IN ('wc-processing', 'wc-completed')
AND s.date_created >= '2026-10-01'
AND s.date_created < '2027-01-16'
GROUP BY day
ORDER BY day;
One gap to know about: if your sale used product sale prices rather than a coupon, the discount never reaches wc_order_coupon_lookup, and WooCommerce does not store the regular price on the order. Record the regular price yourself (an order item meta field or a nightly price snapshot) or you will not be able to compute what the discount cost. Magento is better here: each row in sales_order_item carries both original_price and the price actually charged.
Step 3: Convert sales into profit
Revenue numbers flatter promotions because every discounted sale counts at full weight. Here is a worked example for a store with $3,000 of baseline daily revenue, a 55 percent gross margin at full price, a 25 percent sitewide discount over four days (Friday to Cyber Monday), and $3,000 of extra ad spend.
| Revenue view | Amount |
|---|---|
| Actual event revenue (4 days at $9,000) | $36,000 |
| Baseline for the same 4 days | $12,000 |
| Gross lift | +$24,000 |
| Pre-window shortfall (14 days at $300 below baseline) | -$4,200 |
| Post-window shortfall (28 days at $200 below baseline) | -$5,600 |
| Net incremental revenue | +$14,200 |
Now the same event in profit terms. At a 25 percent discount, $36,000 of event revenue represents $48,000 of goods at full price, which cost $21,600 to buy. Event gross profit is $36,000 minus $21,600, or $14,400.
| Profit view | Amount |
|---|---|
| Event gross profit at discounted prices | $14,400 |
| Profit the baseline would have earned in those 4 days ($12,000 at 55%) | -$6,600 |
| Profit lost in the pre and post windows ($9,800 at 55%) | -$5,390 |
| Extra ad spend | -$3,000 |
| Net incremental profit | -$590 |
The example is built to break even on purpose, because many real events land close to that line. Shipping, payment fees and returns would push it further down. A shallower discount, a narrower product selection or a lower ad budget could push it well above zero. The point of the exercise is to know which side of the line you are on.
Step 4: Estimate subsidy from who bought
Discounting is common enough that subsidy is rarely small. In Metorik’s 2026 WooCommerce statistics, about 25 percent of orders include a discount, averaging 22 percent of order value, and that is across the whole year, not just November.
A practical estimate: for every returning customer who ordered during the event, compute days since their previous order and compare it with their usual interval between orders. Customers whose gap had already reached their usual interval were “due” and would likely have bought around then anyway. Their discount is mostly subsidy. Customers who ordered well before they were due are more likely to be pull-forward. The share of event revenue from due customers, multiplied by the average discount, is a reasonable first figure for what subsidy cost you.
Step 5: Judge the new customers by what they do next
Stores often justify a margin-negative event by the new customers it brings in. That case depends on those customers coming back, and the published data is not encouraging. Ometria’s analysis of its retail clients found that only 4 percent of Black Friday 2024 new customers made a repeat purchase within 12 months, and that Black Friday-acquired customers were six times less likely to return than typical new customers. Other datasets show higher repeat rates, so treat this as a warning to check, not a rule.
Check it on your own data: build a cohort of first-time buyers acquired during the event and another acquired in October or early November, then compare their repeat rates and margin at 30, 60, 90 and 180 days. QuantumLayers’ guide to cohort analysis covers building and reading that table. If the event cohort repeats at a fraction of the normal rate, the acquisition argument does not rescue a weak profit number.
What to set up before this year’s event
Black Friday 2026 is November 27, which leaves a little time to make this year’s analysis easier than last year’s:
- Freeze a baseline now. Save your forecast for late November through mid-January before the event starts, so nobody adjusts it after seeing the results.
- Record list prices. If you discount through sale prices, capture the regular price on each order line or in a daily snapshot.
- Hold some customers back. For email and SMS promotions, withhold the offer from a random slice of the list. Talon.One suggests withholding the promotion from 10 to 20 percent of the eligible audience and comparing post-period sales between the groups. That comparison measures subsidy and pull-forward among existing customers directly. Read the result with the same care as any A/B test, including sample size and the peeking problem.
- Log the extra spend. Ads, influencer fees and agency hours for the event belong in the profit calculation.
- Book the review for mid-January. A verdict on December 1 cannot see the post-window.
For paid media specifically, a geographic holdout gives a cleaner causal read than any of the above. Triple Whale’s GeoLift guide recommends a pre-period four to five times longer than the test window, which in practice means planning it well before the holiday rush.
Sources
- Circana, How to Tell If a Promotion Drove Real Growth (Not Just a Spike) (July 2026)
- Metorik, WooCommerce Statistics & Benchmarks (2026)
- Metorik, Best day to sell online: the weekend discount
- Ometria, Black Friday 2025 data
- Talon.One, How to measure promotion ROI
- Triple Whale, GeoLift 101: Your Guide to Geo-Based Incrementality Testing
- WooCommerce, Installed Database Tables


