Picking an attribution model used to mean choosing from a long menu, and most of that menu is now gone. GA4 quietly deleted the classic rule-based models, so the real decision in 2026 is smaller, sharper, and easier once you know what is actually on offer. Here is how to choose.

The menu shrank, on purpose

In November 2023 GA4 removed first-click, linear, time-decay, and position-based attribution from its reports, and they have not come back. What remains is three models: data-driven attribution, paid and organic last click, and Google paid channels last click. Google Ads moved the same way, leaning on data-driven as the default. So the choice is no longer which of six rules, but data-driven versus a last-click variant.

Start with data-driven if you qualify

Data-driven attribution is the recommended default, because instead of a fixed rule it uses your own data to distribute credit across touchpoints. The catch is a volume threshold: it needs a few hundred conversions for the key event and a larger total across all key events within the lookback window. Clear that bar and DDA is almost always the better in-platform choice.

Know the fallback

If your property does not meet the DDA threshold, GA4 silently falls back to last-click, which matters because you may think you are on data-driven while seeing last-click numbers. Low-volume accounts should therefore not agonize over DDA, since they will effectively be on last-click regardless until they grow into the threshold.

Match the model to the decision

The right model depends on what the number will be used for:

  • For in-platform optimization and everyday reporting, data-driven where you qualify, last-click where you do not.
  • For understanding how channels assist each other, data-driven beats any last-click view, which ignores everything but the final touch.
  • For budget decisions across channels, no click-based model is enough on its own, and you should pair it with incrementality or marketing mix modeling.

Respect the window and keep it consistent

Your model choice interacts with the lookback window, 30 days for acquisition events and 90 for others by default. Changing either reshapes the numbers, so set them deliberately and leave them stable, because switching models or windows mid-stream makes period-over-period comparisons meaningless. Document what you chose and when, so a later shift in the data traces to a setting change rather than a mystery.

Do not expect any model to be truth

The most important move is to stop treating the chosen model as the answer. Every attribution model is a simplifying story about credit, not a measurement of cause, which is why the serious work pairs attribution with experiments. Pick a model for consistency and in-platform guidance, then validate the big calls with incrementality rather than trusting the model to tell you what truly drove sales.

Revisit the choice as you grow

The right model is not a permanent decision, because the thing that gates it, conversion volume, changes over time. An account too small for data-driven today may clear the threshold in a quarter, at which point it should move off the last-click fallback it was silently using. Put a reminder to re-check the active model after any big change in volume, seasonality, or tracking, so you are not still on a crude model long after you earned a better one, or assuming data-driven while quietly running on last-click.

Choosing an attribution model is a smaller decision than it used to be, because GA4 pruned the menu to data-driven and two last-click variants. Default to data-driven if you clear the threshold, know you fall back to last-click if you do not, match the model to how the number is used, and validate the real money decisions with experiments. Do that and you choose deliberately from what actually exists, instead of chasing rule-based models that no longer do.

Want a stronger data analyst role or a raise? Grab the FREE Product Analyst Playbook and get the exact roadmap to your next offer.