Attribution During Holiday Promotions: Reading a Spike Without Fooling Yourself

Promotions bend every attribution input at once: shorter windows, deal hunters, stacked codes. How to read holiday performance without crediting noise.

Holiday promotions produce the most confident attribution reports of the year and some of the least reliable. Everything moves at once: conversion rates jump, windows compress from weeks to hours, shoppers arrive deal first rather than brand first, and codes stack on top of tracked links. Last click declares winners with total certainty about a period when its assumptions are weakest.

This is not an argument to fly blind through peak season. It is a guide to which holiday numbers deserve trust, which deserve an asterisk, and how to set the reading rules before the spike, when the team can still agree on them calmly.

Why promotion periods break normal reading

Baseline attribution assumes a reasonably steady shopper: they see, consider, return, buy. Promotion shoppers behave differently. Many were waiting, cart loaded, for the sale date, so the 'discovery' your report credits happened weeks earlier and the click that gets the credit was a formality. Others are pure deal traffic that will never return at full price, and crediting their acquisition channel as a growth engine builds next year's plan on shoppers who do not exist at normal prices.

Windows compress too. A thirty day lookback that made sense in September spans three promotions in December. Credit piles onto whichever touchpoint stood nearest the sale, and channels that create demand slowly, content, creators, brand search building, watch their numbers fall precisely because the season rewards closers.

Separate the deal cohort before judging channels

The single most useful holiday reporting move is cohort separation: shoppers acquired during the promotion, tracked separately for ninety days, judged on their own repeat rate and margin. A channel that looks miraculous in December and produces no second orders by March was renting revenue. A channel with modest December numbers whose cohort repeats at full price was doing quiet, compounding work.

Apply the same separation to existing customers. Promotions mostly harvest demand from people who already know you, at a discount. Reporting that lumps harvested demand with new demand will tell you the promotion 'grew the business' when it largely prepaid January's revenue at a lower margin. Both things can be worth doing. They should not be confused.

Codes, links, and the stack problem

Holiday stacks defeat single source logic mechanically. A shopper clicks a creator link, returns through a retargeting ad, and checks out with a code from a coupon site. Three channels hold evidence, one order exists, and each platform claims it. Platform reported numbers during peak season routinely sum to far more revenue than the store actually made, which is the tell that everyone is grading their own homework.

Anchor on the store's own numbers: total revenue, total new customers, total orders, and the post purchase survey's self reported discovery for a directional split. Use platform reports for relative movement within a channel, never for absolute credit across channels, and say so in the report before someone presents the summed version to leadership.

Set the reading rules before the sale

Decide in October how December will be judged: which windows, which cohort cuts, which source is the anchor, and what would count as the promotion working. Written down in calm weeks, those rules survive contact with the spike. Invented during the spike, the rules bend toward whichever story the loudest channel is telling.

Include the January reading in the plan: the post promotion hangover, returns, and the cohort's first repeat window are part of the promotion's results. A sale judged on its weekend is a trailer judged as the film. The full picture is slower and less flattering, and it is the one next year's budget should be built on.

Related reading

If this topic is useful, these earlier guides go deeper on connected questions: Attribution for Influencer and Creator Seeding Programs, How to Attribute Sales From In-Store and Pop-Up Events, and Attribution When the First Visit Is Research and the Purchase Comes Weeks Later.

Frequently asked questions

Why does last click attribution mislead during promotions?

Promotion shoppers often decided weeks earlier and were waiting for the sale, so the final click is a formality that collects credit for demand created elsewhere. Compressed windows and stacked codes pile more credit onto closers. Last click reads the season as a triumph of closing channels precisely when its assumptions are weakest.

How should holiday channel performance be judged?

By cohort, over time: separate promotion acquired customers, track their ninety day repeat rate and margin, and separate harvested existing demand from genuinely new demand. Anchor totals on the store's own revenue and order counts, and use platform reports only for movement within a channel.

Platform reports sum to more than our revenue. Which is right?

The store's numbers are right. Each platform credits itself for shared orders, so overlapping claims are expected and peak during promotions when journeys stack links, ads, and codes. Treat the sum as a known artifact, and use post purchase self reporting for a directional view of discovery.

When should we decide how to evaluate the holiday sale?

Before it starts, in writing: windows, cohort definitions, the anchor source, and what success means, including the January follow up reading. Rules set calmly survive the spike. Rules invented mid spike bend toward the loudest story in the room.