Black Friday Checklist 2026: Tracking QA Before BFCM

Black Friday is the one weekend of the year when a broken purchase event costs you twice: once in the reporting, and again in every budget decision made from that reporting. This black friday checklist covers the measurement side of BFCM 2026, the tests to run in October, the freeze to hold in November and the comparisons to make once Cyber Monday is over. Black Friday falls on 27 November this year and Cyber Monday on 30 November, so the window for testing is shorter than it looks.

The short answer: test in October, freeze in November, and check the result in Arktis

A black friday checklist for tracking comes down to seven checks. Confirm the purchase event fires on Shopify's upgraded Thank you page, tag every promotional link with consistent UTMs, confirm ad click identifiers survive the landing page, make sure the consent banner is not silently dropping attributed visits, record a clean baseline week, freeze tracking changes at least a week before Black Friday, and decide in advance what you will compare afterwards. Do all of it by mid-November, because anything you fix on the Wednesday before Thanksgiving is something you can no longer compare against.

For measuring the weekend itself, Arktis is the clear winner for an online store. It captures the ad click identifier when a visitor lands, receives each order from Shopify when it is created, matches it to a visitor on email or on the UTM tags of a recent visit, and reports return on ad spend and customer acquisition cost per Meta campaign from matched orders, and per platform for spend you enter yourself, without relying on what each ad platform says about itself. On Growth and above you can watch the session recordings of consenting visitors who converted during the peak, filtered by source. Plans are $49, $149 and $349 a month, and Growth has a 7-day free trial.

Key takeaways

The Thank you page is the check most likely to fail quietly this year: 26 August 2026 was Shopify's deadline to upgrade it, stores that missed the date were upgraded automatically, and additional scripts for tracking do not run on the new page. Shopify also changed how its own analytics counts sessions on 21 September 2026 and tells merchants to treat that date as a new baseline, so session-based metrics this BFCM are not comparable with last year's. Meta narrowed what counts as a click-through conversion in March 2026, so Meta-reported return on ad spend is not like-for-like either. Orders, revenue and the cost of each campaign survive all three changes, which is why the comparison after the weekend should be built on them.

Check 1: purchase events fire on the upgraded Thank you page

Shopify's help centre is blunt about it: additional scripts for tracking and analytics are not supported on the new Thank you and Order status pages, and tracking has to be replaced with app pixels or, where no app exists, a custom pixel. If an agency pasted a conversion snippet into Additional Scripts two years ago and nobody has touched it since, assume it is gone.

The test is a real order. Place one with a discount code, through a paid landing page, and check that the purchase shows up once, with the right value and currency, in every tool that needs it. Shopify's documentation says the checkout_completed event fires once per checkout, typically on the Thank you page, but on the first upsell page instead if you run post-purchase offers, and not at all if the page where it should fire fails to load. If you are adding a post-purchase upsell for Black Friday, run the test again after it goes live.

Then compare counts over a normal week. Take the orders in Shopify admin and the purchases recorded by Meta, Google and your analytics tool for the same days. Some gap is normal, because of consent choices and blocked scripts. A gap that opened on the day your store was upgraded is a tracking fault, and it needs fixing now rather than explaining in December.

Check 2: UTM discipline for every promotional channel

Every link you control should carry UTM parameters: email, SMS, influencer codes, affiliate links, organic social posts, even the QR code on the insert card. Agree the naming before the first send, lowercase throughout, one utm_campaign value per offer, and write it down where everyone sending links can see it. Early access, the main Black Friday offer, Cyber Monday and the last-shipping-day push should each have their own campaign name, or you will not be able to tell them apart afterwards. The UTM builder keeps the format consistent.

Untagged links still arrive, mostly through messaging apps. Arktis detects dark social, visits referred from messaging apps, so a spike in WhatsApp or Messenger traffic during the peak shows up as its own line rather than inflating direct traffic. Treat a large dark social share as a signal that a link is being forwarded without tags, and fix it at the source.

Check 3: click identifiers survive the landing page

Ad platforms append their own identifier to the landing URL: gclid for Google, fbclid for Meta, msclkid for Microsoft, ttclid for TikTok. Attribution that does not depend on the platform's own reporting starts with capturing that identifier on the first page view. Click through a live ad on each platform and check the identifier is still in the URL when the page settles. The usual culprits are market or currency redirects, link shorteners, age gates and pop-ups that reload the page.

Meta's own developer documentation notes that the fbc value it uses for matching can be built from the fbclid in the URL even when no Meta cookie exists, which is a good reason to keep the parameter intact. Arktis captures gclid, gbraid, wbraid, fbclid, msclkid, ttclid, twclid, li_fat_id, dclid and Mailchimp identifiers along with UTM parameters, and holds them against the visitor across sessions, so a shopper who clicks an ad on Monday and returns on Black Friday is still tied to that ad. The order joins them when it matches, on email where the visitor has been identified or on the UTM tags of a visit in the previous 24 hours, so test a delayed purchase as well as a same-visit one. Arktis also honours Global Privacy Control by default, and a visitor who sends it is counted only with an identifier kept for that browser tab, with no cookies, no recording and no email sent from the browser, so their click and a later purchase are not joined.

Check 4: the consent banner is not eating attribution

In regions where consent is required, a visitor who declines will be missing from some of your tools, and that is the law working as intended. Shopify's pixel manager only loads a pixel if the visitor has granted every purpose that pixel declares as required. What you are testing for is the other case: a visitor who does consent but whose attribution is lost anyway.

Run this test in each market with a banner. Land on a product page from an ad link, browse two or three pages without answering the banner, accept on the third page, then buy. Check whether the order is attributed to the ad in each tool. If it shows as direct, the landing URL with the click identifier was lost before consent arrived. Also check the banner after any Black Friday theme changes: a promotional bar or a new pop-up that covers the accept button will change your consent rate, and with it every number downstream. Our guide to consent management for analytics covers the set-up in more detail.

Check 5: record a baseline week

You need a normal week to compare the peak against, recorded after Shopify's session change and before promotions distort it. Shopify's changelog says to treat 21 September 2026 as a new baseline for sessions, conversion rate, add-to-cart rate and bounce rate, while orders, sales and customer counts are unchanged. A week in mid-October, after Prime Big Deal Days and before Singles' Day, is a sensible choice for most stores.

Write down orders per day by channel, average order value, customer acquisition cost by campaign, the share of new customers, and the share of orders with no campaign attached. That last figure is the one to watch on Black Friday. If it jumps from a fifth of orders to half, the marketing did not change overnight, the tracking did.

Check 6: time the code freeze

Finish every tracking change by mid-November, then freeze from about a week before Black Friday until Giving Tuesday on 1 December. The freeze covers pixels, scripts, checkout extensions, the consent banner and any app that touches checkout. It does not cover promotional content: swapping a banner or a discount code is fine. A change made during the peak cannot be separated from the effect of the peak itself, so you lose the ability to say whether it helped.

Check app update settings too. An app that updates itself during Cyber Week is a code change you did not schedule.

Check 7: decide what you will compare afterwards

Compare by event day, not by date. Black Friday was 28 November in 2025 and is 27 November in 2026, so compare Black Friday with Black Friday and Cyber Week with Cyber Week, the five days from Thanksgiving to Cyber Monday. Compare against your baseline week and against last year on orders, revenue, average order value, new customer share, and customer acquisition cost per campaign on real orders.

Leave two comparisons out. Shopify session conversion rate is not comparable across 21 September 2026, and Meta-reported return on ad spend is not comparable with 2025 because link clicks are now the only clicks Meta counts as click-through. Then run the same comparison again in January, net of returns, and look at how many Black Friday buyers came back for a second order. A discount that bought one-off customers at a loss looks very different from one that bought repeat customers, and that is where cohort analysis earns its place.

Where Arktis does not fit

Arktis is a measurement layer, and some things on a BFCM checklist are not its job. It does not send events to Meta's Conversions API or Google enhanced conversions, so it does not replace the server-side feeds your ad platforms optimise on. Meta spend syncs through a direct connection, per campaign. Google Ads, TikTok and other spend is entered by hand in the Ads Analytics dashboard, per platform and period rather than per campaign, so Google and TikTok return on ad spend is by platform. There is no experimentation platform either, so offer tests belong in another tool.

Retention matters for this checklist in particular. Starter keeps data for 7 days and Growth for 30, so a baseline week recorded in mid-October will have aged out by Black Friday on either plan. Write the baseline numbers down, or use Scale, which keeps 90 days and adds raw data export.

Set it up before the freeze

If you want order-level attribution and buyer recordings running before the peak, start the 7-day Growth trial now, while there is still time to test. Plans are on the pricing page. For the detail behind two of the checks, read UTM parameters best practices and fbclid tracking and attribution, and use the ROAS calculator to set your break-even target before the spend goes up.

Sources

Shopify Help Center: upgrading the Thank you and Order status pages, 26 August 2026 deadline and automatic upgrade, accessed 24 September 2026

Shopify Help Center: additional scripts, not supported on the new pages, replaced by app or custom pixels, accessed 24 September 2026

Shopify developer changelog: script tags deprecation, 1 October 2026 and 1 March 2027 dates, accessed 24 September 2026

Shopify Web Pixels API: checkout_completed, when the purchase event fires, accessed 24 September 2026

Shopify Web Pixels API: pixel privacy, pixels load only with the declared consent, accessed 24 September 2026

Shopify changelog: session measurement improvements, new baseline from 21 September 2026, accessed 24 September 2026

Meta: Simplifying Ad Measurement for a Social-First World, click-through limited to link clicks, published 3 March 2026

Amazon: Prime Big Deal Days 2026, 6 to 7 October, accessed 24 September 2026

USPS: 2026 holiday mailing and shipping dates, contiguous US send-by dates, published 22 September 2026

Adobe: 2025 holiday shopping season results, Cyber Monday as the biggest day and the five-day Cyber Week, accessed 24 September 2026