MetricRaven

Attribution

What is data-driven attribution?

Data-driven attribution is a model that uses machine learning on converting and non-converting paths to estimate how much each touchpoint contributed to a conversion, and splits the credit accordingly. It is one of the attribution models in Google Analytics 4 and Google Ads.

Updated

What data-driven attribution is

Rule-based models give credit by a fixed formula: all to the first click, all to the last, equal shares (linear), or more to recent touches (time decay). Data-driven attribution (DDA) learns the split from your data instead. It compares the paths of people who converted with the paths of people who didn't and asks which touchpoints made a conversion more likely.

Google describes it as a counterfactual approach: the model contrasts what happened with what could have happened without a given touchpoint, then gives credit by that estimated contribution. Each model is specific to one advertiser and one key event, and takes into account factors such as time before the key event, device type, number of ad interactions, order of ad exposure and creative.

Data-driven attribution in GA4

GA4 offers three attribution models in its attribution reports:

  • Data-driven: credit split by the machine-learning model.
  • Paid and organic last click: all credit to the last channel clicked, ignoring direct traffic.
  • Google paid channels last click: all credit to the last Google Ads click, if there was one.

All three exclude direct visits from receiving credit unless the whole path is direct. The first click, linear, time decay and position-based models were removed in November 2023, so in GA4 data-driven is the only model that spreads credit across touchpoints. Conversions can be reattributed for up to 7 days after they happen.

When data-driven attribution works

DDA needs a lot of paths, both converting and not, to learn from. With a few dozen sales a month, the split it produces rests on little evidence. It also only sees the touchpoints Google can see: clicks and engaged views it measures, not a podcast, a Slack link or an ad on a network it doesn't track.

Use it to compare channels inside Google's view, not as the answer to what caused your revenue. For that, a holdout or geo test (incrementality) is stronger, and for budget across all channels, media mix modeling.

Common data-driven attribution mistakes

  • Treating it as causal proof. It estimates contribution from patterns in paths; it isn't a controlled experiment.
  • Comparing it with ad platform numbers. Meta and Google Ads each credit themselves under their own rules.
  • Expecting to audit it. You can't see the weights or reproduce a single conversion's split.
  • Forgetting fractions. A campaign can get 0.4 of a conversion, so counts stop matching your orders.

How MetricRaven compares

MetricRaven doesn't use data-driven attribution. Each payment goes to the customer's first tracked visit (first-touch attribution), a rule you can check by opening the customer's journey. Revenue is real payments from your payment provider, not modeled key events, so a channel's numbers add up to your actual revenue.

To see more than the first touch, the ad tables add Assisted revenue for campaigns customers clicked later, and Pages can credit revenue to pages buyers viewed before paying. If you are weighing the two tools, see the GA4 alternative page and multi-touch attribution for how split models differ.

Questions

Is data-driven attribution the default in GA4?
Data-driven is one of three models GA4 offers, alongside paid and organic last click and Google paid channels last click. You choose the reporting model in the property's attribution settings.
How is data-driven attribution different from last click?
Last click gives 100% of the credit to the final channel clicked. Data-driven splits credit across the touchpoints on the path based on how much each one appears to raise the chance of converting.
How much data does data-driven attribution need?
Google doesn't publish a fixed minimum for GA4, and the model may use aggregate data when yours is thin. As a rule, the fewer conversions you have, the less weight to put on the split it shows.
Does MetricRaven support data-driven attribution?
No. MetricRaven credits revenue to the customer's first tracked visit and shows assisted revenue for ad campaigns. Every number traces back to real payments and visits you can inspect.
Full glossary

Attribution you can check, from real payments

MetricRaven credits each payment to the visit that first brought the customer, so you can open any sale and see why its channel got the credit.

14-day free trial, no card required.