Attribution in 2026: What You Can and Cannot Measure

A few years ago, many store owners believed their ad dashboards showed exactly which ad led to which sale. That belief was never fully accurate, and it is less accurate now. Browser restrictions on cookies, operating-system privacy prompts on mobile apps, consent requirements in many regions and ad blockers all reduce what can be tracked directly. Ad platforms have responded by filling gaps with modelled conversions. None of this means measurement is impossible. It means you need to know which numbers are observed, which are estimated and which questions can only be answered with a different method. This guide is a practical map of what you can and cannot measure at the time of writing.

What changed, in plain terms

• Third-party cookies are blocked or limited by several major browsers, which makes it harder to follow a person across websites.
• Mobile app tracking prompts let users decline tracking across apps, which reduces the data platforms receive from people who tap ads inside apps.
• Consent rules in places such as the UK and the European Union mean tracking often depends on what the visitor accepts in a cookie banner.
• Shorter attribution windows are now the default on some platforms compared with what was common in the past.
• Modelled conversions are used by platforms to estimate sales they cannot observe directly.

The details differ by platform, region and device and continue to change, so check each platform's current documentation and your legal obligations for your markets.

What you can still measure reliably

Your own store's orders and revenue

Your store platform records every order, its value, the products and, usually, discount codes used. This is your most reliable source of truth for what was actually sold. It does not tell you which ad caused each order, but it tells you what happened.

Spend

What you paid each platform is exact. Combined with total store revenue, it gives you a blended efficiency figure (total revenue divided by total ad spend) that does not depend on any attribution model. Many experienced advertisers watch this closely precisely because it cannot be double counted.

On-platform engagement

Impressions, clicks, video views and click-through rates are measured on the platform's own surfaces and remain relatively dependable for comparing creatives and audiences.

First-party conversion signals

Server-side connections, such as Meta's Conversions API or similar tools from other platforms, send purchase events from your store's server rather than relying only on the browser. These can recover some signal lost to browser restrictions. They still need to respect the consent choices your visitors make.

What you can only estimate

Platform-attributed revenue

When a platform reports that a campaign drove a certain amount of revenue, part of that figure may be modelled. It is an estimate based on the platform's rules, not a list of verified orders. Each platform also tends to credit itself, so adding up revenue reported by Meta, TikTok and Google can produce more than your store actually took.

Cross-device and view-through journeys

Someone who sees a video on their phone and buys days later on a laptop is hard to connect with certainty. View-through conversions, credited when someone saw but did not click an ad, are especially uncertain.

Multi-touch paths

Multi-touch attribution tries to divide credit across every interaction. With fewer touchpoints visible, these models rely more on assumptions. They can still be useful for comparisons, but the precise shares they assign should be read with caution.

What attribution cannot tell you on its own

The biggest question in advertising is incrementality: would this sale have happened without the ad? Attribution, however sophisticated, assigns credit to touchpoints. It does not prove cause. A retargeting campaign can look extremely efficient because it reaches people who were about to buy anyway.

To answer incrementality questions, advertisers use tests rather than models:

• Holdout or lift tests, where part of an audience is deliberately not shown ads and results are compared. Some platforms offer tools for this; availability and requirements vary.
• Geographic tests, where spend changes in some regions and not others.
• On-off tests, where a channel is paused for a period and the effect on total store revenue is observed. These are simple but affected by seasonality and other changes happening at the same time.

These tests take planning and budget, and results are rarely perfectly clean, but they answer a question no dashboard can.

A practical measurement setup for a store

1. Get the basics right. Install each platform's pixel correctly, add server-side events where available, and confirm that test purchases appear. Make sure your consent banner works as intended.
2. Use consistent UTM parameters on every ad link so your analytics tool can group traffic by source and campaign.
3. Track blended efficiency weekly. Compare total ad spend with total store revenue and new-customer revenue.
4. Use platform numbers for decisions within that platform, such as which creative or audience to favor, where their relative comparisons remain useful.
5. Add a simple post-purchase question such as "How did you hear about us?". Self-reported answers are imperfect, but they can reveal channels that tracking misses.
6. Run occasional incrementality tests on your largest spending areas before making major budget shifts.

Where Novaris fits

Novaris is an AI growth platform for e-commerce brands, built with our partner Atiko. Our attribution tracking brings platform data and store data together so you can see which channels and campaigns appear to drive revenue, alongside blended figures, and it is open about its limits: some data is modelled, and privacy choices mean some journeys cannot be seen. Reporting depth by plan is explained on our advanced analytics page. Results depend on your products, market, budget and ad accounts.