What will my developer ask about an attribution tool?

What it puts on the site, what it collects, whether it touches checkout, and what happens when it fails. An export-based read answers nothing, nothing directly, nothing, and a late report, which usually ends the review quickly.

How does attribution work if the tool never sees a session?

It estimates an aggregate effect from variation already present across time and channel in your analytics data, then reports how sure it is. It knows nothing about any individual session.

Should I describe a GA4-export-based tool as cookieless?

Only carefully. The accurate claim is that the tool adds no cookie, tag or identifier of its own. Your analytics continues collecting however it already collects, and overstating that distinction tends to backfire in a technical review.

Explaining pixel-free attribution to your developer

The developer review is where measurement tools die. Four questions engineering always asks, the honest answers, and the one objection worth taking seriously.

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 2026-10-063 min read

Engineering does not object to measurement, it objects to unbounded collection on a critical path.

Here is what gets asked, in the order it gets asked.

The four questions

Question	The answer for an export-based read
What does it put on the site?	Nothing. No tag, no script, no identifier
What does it collect?	Nothing directly. It reads a file you export
What does it touch in checkout?	Nothing
What happens if it breaks?	A report is late. No visitor-facing impact

The fourth row is the one engineering cares most about and marketing never anticipates. A script in the critical path is a thing that can fail in front of customers. A batch process that reads a file cannot.

The objection that deserves a real answer

The serious version of the pushback is not about collection, it is about validity: "if it never sees a session, how does it know anything?"

That is a fair question and the honest answer is that it does not know anything about a session. It estimates an aggregate effect from variation that already exists in your data across time and channel, and it reports how sure it is.

An engineer will usually accept that framing quickly, because it is the framing they already use for anything inferred rather than measured. What they react badly to is a vendor claiming certainty the method cannot support, which is a reasonable reaction. The method is written out in plain language on how it works, and the wider argument for publishing it is in attribution vendors that explain their methodology openly.

What to bring to the review

Bring three things. The description of the input, which is a Google Analytics CSV export you produce yourself. The description of the output, which is a per-channel estimate. And the privacy policy, because the data handling question will come up and pointing at a document is faster than relaying an answer.

What not to claim

Do not tell your engineer it is "cookieless" if what you mean is that it adds no cookie of its own. Analytics still collects the way it collects; we simply do not extend it. Overstating that is the fastest way to lose credibility in a technical review, and the accurate version is a strong enough position without the embellishment.

The reason this matters organisationally

Measurement tools that need engineering time get scheduled, and scheduled means quarterly at best. Removing the install removes the dependency on someone else's roadmap, which is the difference between a marketing team that can read its channels monthly and one that reads them when a ticket clears. That is usually the real unlock, and it is covered from the buying side in attribution without a pixel or engineering.

Frequently asked questions

What will my developer ask about an attribution tool?
What it puts on the site, what it collects, whether it touches checkout, and what happens when it fails. An export-based read answers nothing, nothing directly, nothing, and a late report, which usually ends the review quickly.
How does attribution work if the tool never sees a session?
It estimates an aggregate effect from variation already present across time and channel in your analytics data, then reports how sure it is. It knows nothing about any individual session.
Should I describe a GA4-export-based tool as cookieless?
Only carefully. The accurate claim is that the tool adds no cookie, tag or identifier of its own. Your analytics continues collecting however it already collects, and overstating that distinction tends to backfire in a technical review.

Go deeper: Incrementality testing, 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

Browse the full glossary