What a real Shopify channel lift answer requires
Lift means a comparison against a world where you did not run the channel. Three requirements before the word is honest, and where speed genuinely trades against strength.
Three platforms, three answers, one budget to defend. Your GA4 export already holds the answer, and the first finding is free and never leaves your browser. Get your first finding free
By Joris van Huët, Founder & CEOUpdated 3 min read
Run the numbers for your store: the free holdout test planner.
Lift is not a number your store reports, it is a comparison between what happened and what would have happened without the channel. The second half of that sentence is the whole problem, because it did not happen and cannot be observed.
Every method that claims to measure lift is a different strategy for constructing that missing comparison. Understanding which strategy you are buying is more useful than comparing feature lists.
Three things any honest lift answer carries
| Requirement | What it means | What it looks like when missing |
|---|---|---|
| A stated comparison | What the estimate is measured against | "Lift" with no baseline named |
| A stated uncertainty | How wide the plausible range is | A single confident number |
| A stated design | How the comparison was constructed | "Our model determines" |
The first is the one that separates real answers from marketing. Lift against last month is not lift against no-channel. Lift against a platform-reported baseline is not lift at all, it is a comparison between two claims. A counterfactual has to be constructed deliberately, and the construction has to be named.
The three ways the comparison gets built
Randomisation builds it by holding a group out, so the held-out group is the comparison. This is the strongest design and it costs the revenue you forgo in the holdout, plus weeks of waiting. The geo testing guide covers how.
Quasi-experimental designs exploit something that behaved like a randomiser without being one, such as a regional rollout or a platform outage. Strong when the accident is clean, and rare.
Observational designs use variation already present in the data across time and channel. Weakest of the three, available immediately, and honest when labelled as observational rather than dressed as a test. That distinction is worked through in incremental ROAS from GA4 without a geo test.
What "fastest" costs
Speed trades against strength, and the trade is real rather than a marketing framing. The fastest possible read is observational, takes minutes, and produces an estimate you should hold loosely. The strongest possible read is randomised, takes weeks, and produces an estimate you can defend to anyone.
The mistake is not choosing the fast one. It is choosing the fast one and then describing it as though you had run the slow one. A report that says observational is more useful than one that implies experimental, because it lets the reader weigh it correctly.
Where Shopify data fits
Your store holds the outcome side of the comparison: orders, revenue, timing. It does not hold the counterfactual, and no Shopify app can conjure one from order data alone. What order data does well is anchor the read, because it is the ground truth on what actually sold, against which channel-side claims can be checked. We covered that check in platform-reported ROAS against orders.
The question to ask any lift tool
Ask what the estimate is compared against. If the answer is a period, a platform number, or "our model", you have a comparison of claims rather than a measurement of lift.
Related answers
Frequently asked questions
What does marketing lift actually mean?
The difference between what happened with the channel running and what would have happened without it. The second half is unobservable, so every lift method is a strategy for constructing that missing comparison, and the strategy is what you are really choosing between.Can a Shopify app measure channel lift on its own?
Order data tells you what sold, not what would have sold without the channel. An app reading only Shopify can anchor the outcome side accurately, but it cannot construct the counterfactual, which has to come from a test design or from modelled variation.Is a fast lift estimate worth anything?
Yes, provided it is labelled honestly. An observational estimate produced in minutes is weaker evidence than a randomised holdout and stronger than a rules-based model. The problem is not the speed, it is describing a fast read as though it were a test.
Go deeper: Causal attribution, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
Have an audience of ecommerce operators? Earn 20% recurring commission by recommending the €99 read.
Keep reading
Terms in this article
- AttributionAttribution identifies user actions that contribute to a desired outcome and assigns value to each. It reveals which marketing touchpoints drive conversions.
- CounterfactualA counterfactual is what would have happened under conditions that did not occur, such as a customer's outcome had they not received the campaign they did receive. It is never observed, so every effect estimate has to stand in for it with a control group, comparable units or a forecast.
- Quasi-ExperimentA quasi-experiment estimates the causal impact of an intervention without random assignment. It applies when random assignment is not feasible or ethical.
- RevenueRevenue is the total income generated by the sale of goods or services related to a company's primary operations.
- ShopifyShopify is an ecommerce platform for creating online stores and selling products. Attribution modeling shows which marketing channels drive traffic and conversions within Shopify.