Multi-touch attribution software: a buyer's guide
Judge a multi-touch tool on what it sees, how it treats visitors who decline consent, what its numbers were compared with, and whether you can export the rows. Put five questions to the vendor in writing before any demo.
€99 once, excl. VAT. Full refund within 30 days, no questions asked. You keep the read.
By Joris van Huët, Founder & CEOPublished 5 min read
Run the numbers for your store: the free customer journey credit calculator.
Buy on what you can check before you pay: what the tool sees, what it does with visitors who decline consent, what its numbers were compared against, and whether you can take the rows away. The third matters most. In 15 Facebook advertising experiments, observational methods "often fail to produce the same effects as the randomized experiments" (Gordon and colleagues, Marketing Science, 2019), so a multi-touch tool that cannot show a comparison with an experiment is asking for trust, not offering evidence.
What does the tool see, and what does it lose?
Multi-touch credit needs a path, and a path needs an identifier. Google's documentation says that when users don't grant consent, events "are not associated with a persistent user identifier", so Analytics "can't detect if the events are the action of the same user". For those visitors there is no path to credit. A tool with its own tag faces the same choice: drop them, model them, or count them separately.
Google can model part of the gap, on conditions. The property needs "at least 1,000 events per day with analytics_storage='denied' for at least 7 days" and "at least 1,000 daily users sending events with analytics_storage='granted' for at least 7 of the previous 28 days". If you block Google tags until consent is given, Google says you will not get modeled data. With a consent solution that uses the IAB Europe TCF, Google Analytics properties "aren't able to model data to fill in the missing information".
What a tool reads matters as well. Google lists "Data export, for example, BigQuery export" among the features that do not support modeled behavioral data, and labels its daily BigQuery export "Last click observed, no modeling". A tool built on that export sees observed events, including cookieless pings where consent mode sends them, but none of Google's modeled numbers.
Which questions should you put in writing?
Five questions to send before any demo:
- Where does the data come from? Good answer: names each source (a tag, a server feed, platform APIs, a GA4 export) and what it misses. Red flag: a claim to capture everything.
- What happens to visitors who decline consent? Good answer: dropped, modeled or counted separately, and labelled as such in the output. Red flag: full coverage claimed with no mention of consent.
- How is Direct treated? Google's own models give it no credit: "All attribution models exclude direct visits from receiving attribution credit, unless the path to key event consists entirely of direct visits." Good answer: says whether Direct is excluded, credited or split, and shows what that does to each channel. Red flag: no answer.
- What was the output compared with? Good answer: a named comparison with a randomized holdout or lift test, with sample, dates and who ran it, labelled as the vendor's own if it is. Red flag: an accuracy percentage with no test design behind it.
- Can you leave with your rows? Good answer: an order-level export with the credit per channel, so you can recompute the totals. Red flag: dashboards and PDFs only. Google's own position on its export is that "you own that data".
What does the price model reward?
A tool priced on your revenue earns more when you believe it drove more revenue. Ask whether the price moves with your revenue or ad spend, how long the contract runs, and what happens to your data when it ends.
How do you judge the answers?
Measure your own consent gap first, so you can tell a fair answer from a flattering one. If GA4 is linked to BigQuery, this counts events and purchases by consent status. The privacy_info fields hold the consent status of a user when consent mode is enabled, with values Yes, No or Unset:
SELECT privacy_info.analytics_storage AS analytics_storage,
COUNT(*) AS events,
COUNTIF(event_name = 'purchase') AS purchases
FROM `your_project.analytics_PROPERTY_ID.events_*`
WHERE _TABLE_SUFFIX BETWEEN '20260901' AND '20260930'
GROUP BY analytics_storage
ORDER BY events DESC;
Pass: the No and Unset rows are small next to Yes for purchases, and the vendor's answer to the second question matches what you see. Fail: No or Unset holds a large share of purchases, or the answer ignores them. If you run basic consent mode, denied visitors send nothing, so the query cannot show them; compare GA4 purchases with Shopify orders for one closed week instead.
Then mark each vendor answer pass, partial or fail. A fail on the fourth question is a reason to run your own holdout before you buy.
Sources, 30 September 2026: A Comparison of Approaches to Advertising Measurement (Gordon and colleagues, Marketing Science, 2019; abstract via IDEAS/RePEc); Behavioral modeling for consent mode (Google Analytics Help, 2026); Consent mode on websites and mobile apps (Google Analytics Help, 2026); Set up BigQuery Export (Google Analytics Help, 2026); BigQuery Export schema (Google Analytics Help, 2026); Get started with attribution (Google Analytics Help, 2026).
Related answers
Frequently asked questions
What should I ask a multi-touch attribution vendor?
Ask five things in writing: where the data comes from, what happens to visitors who decline consent, how Direct traffic is treated, what the output was compared with, and whether you can export the rows behind every number. A vague answer to the fourth means the numbers are unvalidated.Does multi-touch attribution work with a cookie consent banner?
Only partly. Google's documentation says events from users who decline consent carry no persistent identifier, so Analytics cannot tell whether they are the same user. Google can model some of the gap when conditions are met, but its BigQuery export does not include modeled data.How can I tell if an attribution tool is accurate?
Ask what it was compared with. In 15 Facebook experiments, observational methods often failed to match the randomized results, so a comparison with a holdout or lift test is the evidence that counts. A vendor-run comparison is worth having if it is labelled as the vendor's own and the method is shown.
Go deeper: Incrementality testing, explained.
Sixty-second versions of these ideas: Causality Engine on YouTube Shorts.
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.
- Attribution ModelAn Attribution Model defines how credit for conversions is assigned to marketing touchpoints. It dictates how marketing channels receive credit for sales.
- Attribution PlatformAttribution Platform is a software tool that connects marketing activities to customer actions. It tracks touchpoints across channels to measure campaign impact.
- Attribution SoftwareAttribution Software measures campaign impact by tracking customer interactions across touchpoints. It assigns value to each channel, showing what drives conversions.
- Causal AttributionCausal Attribution uses causal inference to determine which marketing touchpoints genuinely cause conversions, not just correlate with them.
- ExperimentsExperiments are scientific procedures that test hypotheses or demonstrate facts. In marketing, experiments like A/B tests determine the causal effect of campaign changes, enabling data-driven decisions.
- Google AnalyticsGoogle Analytics is a web analytics service that tracks and reports website traffic.
- Multi-Touch AttributionMulti-Touch Attribution assigns credit to multiple marketing touchpoints across the customer journey. It provides a comprehensive view of channel impact on conversions.