Attribution & Conversion Counting

Data-Driven Attribution vs Linear Attribution

In short: Both models try to answer how much credit each touchpoint in a conversion path deserves, but they get there in opposite ways. Data-driven attribution (DDA) lets the platform learn contribution from comparing converting and non-converting paths in your account; linear just splits credit evenly with no learning involved. DDA is the current default in Google Ads and GA4 and needs enough conversion volume to model; linear was retired from both in 2023 and now lives only in third-party or custom tooling. If you are inside Google's ecosystem with reasonable conversion volume, DDA is basically the only option left; reach for linear only when you are working outside Google's reporting or need a transparent, non-modeled baseline.

By the AdFlint research team · Fact-checked against current Google and Meta platform behavior · Last reviewed July 2026

Data-Driven Attribution

Uses a platform's own modeling of converting and non-converting paths to distribute fractional credit based on observed contribution rather than fixed rules.

The platform compares paths that converted with paths that did not, then assigns each touchpoint a fractional share reflecting its measured contribution. It is the default in Google Ads and GA4 and needs enough conversion volume before the model has data to learn from. Two things get misread: the credit is modeled rather than observed, and it remains in-platform, so it cannot see channels outside that platform's data.

Full definition

Linear Attribution

Splits conversion credit evenly across every touchpoint in the path, so a five-touch journey gives each interaction twenty percent.

Every recorded interaction receives an identical share of the conversion, which makes long paths visible without asserting that any one step mattered more. Analysts use it as a neutral starting point when comparing channel involvement rather than channel impact. Google Ads and GA4 removed it from their reporting in 2023, though it remains standard in other analytics tools. The trap is treating equal credit as evidence of equal influence.

Full definition

Side by side.

The differences that actually change what happens in your account.

 Data-Driven AttributionLinear Attribution
Where the weights come fromModeled from your account's own converting and non-converting paths, a machine-learned estimate of contribution.A fixed rule: every touch gets one divided by the number of touches, no data involved.
Minimum data needed to runEnough conversion volume for the model to detect patterns; too little volume and it cannot build a reliable model.None, it works identically on one conversion or ten thousand.
Where you find it todayDefault in Google Ads and GA4 as of 2023.Removed from Google Ads and GA4 in 2023; available in third-party tools or custom modeling only.
Transparency of the creditA black box relative to linear, you see the resulting split, not the reasoning the model used to get there.Fully transparent, anyone can recompute the split by counting touches.
Behavior on a channel that never appears in converting pathsGets little to no credit, because the model has no evidence it contributes.Gets full proportional credit anytime it appears, whether or not it is ever near a conversion.
Stability over timeCan shift as the model retrains on new data, so the same path might get a different split next month.Never changes, the formula is fixed, so historical comparisons stay apples to apples.
What it costs you when wrongLow-volume accounts get a model with too little signal, producing credit assignments that look precise but rest on thin data.Treats a channel that only ever tags along with a channel that actually drives conversions as equally responsible.

What actually separates them.

01

DDA recomputes credit shares from observed path outcomes and can change its own weighting as more data accumulates; linear's formula never changes regardless of outcomes.

02

Linear runs on any volume of data, including a single conversion, while DDA needs a meaningful sample of both converting and non-converting paths before its output is trustworthy.

03

DDA can and will assign a touchpoint near-zero credit if it never correlates with conversions; linear guarantees every touch some nonzero share just for appearing.

04

Because DDA is in-platform only, it cannot see touchpoints outside that platform's tracking, while linear can be computed on any exported path data regardless of source.

05

DDA is the only one of the two still selectable inside Google Ads and GA4 as of 2026; linear must be recreated in a third-party tool or spreadsheet if you want to use it today.

Which one should you use?

Use Data-Driven Attribution when

  • You are working inside Google Ads or GA4 and have enough monthly conversion volume for the platform to model contribution.
  • You want credit assignment that reflects what actually correlated with conversions in your account, not a rule imposed from outside.
  • You are setting Smart Bidding targets and want the conversion data feeding the bidder to reflect real contribution rather than an even split.
  • You have already tried last-click and suspect it is overcrediting your closing channel.

Use Linear Attribution when

  • Your conversion volume is too low for a modeled approach to be reliable, or you are working outside Google's reporting entirely.
  • You want a transparent, recomputable number you can explain to a client without saying the algorithm decided.
  • You are doing a first-pass involvement audit and do not want a model's assumptions baked in yet.
  • You need attribution logic that stays fixed over time for consistent period-over-period comparisons.

Common questions.

Is data-driven attribution more accurate than linear?

It is better evidenced, not automatically more accurate. DDA reflects patterns Google's model found in your specific account's converting and non-converting paths, while linear imposes no assumption at all. In a low-volume account, DDA's model may be working from too little data to beat a simple, transparent split.

Why can't I select linear attribution in GA4 anymore?

Google retired linear along with the other rule-based multi-touch models, time-decay and position-based, from GA4 and Google Ads reporting in 2023, leaving data-driven attribution and last-click as the built-in options. If you need a linear-style split today, you would have to export conversion path data and calculate it yourself or use a third-party attribution tool.

Does data-driven attribution use the same conversion data as linear would?

They can start from the same underlying event and path data, but DDA also needs a large enough set of non-converting paths to compare against, which linear does not require at all. That is the core reason DDA has a volume threshold to function well and linear does not.

Or stop choosing between them.

AdFlint picks the setting, writes the ads, and keeps optimizing inside the Google and Meta accounts you already own.

Try AdFlint freeSee how it works

Related comparisons

All comparisons