Strategy & Tracking
Reading attribution in GA4 without fooling yourself
GA4 now defaults to data-driven attribution and has retired four old models. Learn to read the Attribution reports correctly so you stop double-counting your remodel leads and stop trusting numbers that are quietly off by 5 to 20 percent.
Attribution in GA4 changed under most marketers without warning. Data-driven attribution is now the default reporting model, and Google retired first-click, linear, time-decay, and position-based attribution in late 2023 because fewer than 3% of conversions ever used them. That shift, plus consent mode, data thresholds, and the gap between GA4 and Google Ads reporting, means the same estimate request can read three different ways depending on where you look, which matters when each remodel lead is worth thousands. The levers that keep you honest are a short list: the right reporting model, the Attribution and conversion-paths reports, the lookback window, and knowing why GA4 and Google Ads will never fully agree.
Know which model GA4 is actually using
GA4 reports on one attribution model at a time, and the default is data-driven attribution. Instead of crediting the last click, DDA uses machine learning to split credit across the touchpoints that statistically moved a conversion. The catch: when a property lacks enough data, GA4 silently falls back to a rule-based model. Google recommends roughly 600 conversions and 3,000 path interactions a month before DDA produces stable credit, so smaller accounts may be reading a last-click model while believing they see DDA.
Your other choice is paid and organic last click, which hands 100% of the credit to the final non-direct channel. There is also a channel-credit setting that decides whether only Google paid channels or all paid and organic channels can earn credit; Google recommends paid and organic for a true cross-channel view. WellBuilt checks both settings on day one of any engagement, because a chart built on the wrong model sends budget to the wrong place no matter how clean the data underneath it is.
The reports that survived the 2023 purge
Google sunset four attribution models across Ads and Analytics by November 2023, citing adoption under 3% and journeys those rigid models could not adapt to. First-click, linear, time-decay, and position-based vanished from the model-comparison view, leaving data-driven and last click. The April 2026 GA4 restructure went further, merging the old attribution-paths and model-comparison reports into a single Attribution report with tabbed views. The tools are fewer now, but the two that matter most still answer the questions worth asking.
Live in the Advertising workspace, under Attribution. The conversion-paths report shows the real channel sequences that lead to conversions, splitting early, mid, and late touchpoints so you can see which channels open the door versus close it. Last-click attribution systematically under-credits those assists, and that blind spot matters when buyers now commonly cross several touchpoints, often cited as six or more, before they convert. Read paths before you cut a channel that looks weak on a last-click chart.
Where to look in GA4's Advertising workspace:
- Attribution report: compare how DDA and last click credit the same conversions
- Conversion paths: the actual channel sequences, split into early, mid, and late touchpoints
- Attribution settings: confirm your reporting model and channel-credit scope
- Key event lookback window: 90 days for most events, 30 for acquisition
- Reporting identity: blended or observed triggers thresholding; device-based does not
Why GA4 and Google Ads will never match
Stop trying to reconcile them to the row. GA4 and Google Ads answer different questions: Google Ads asks how many conversions its ads drove, GA4 asks how many conversions happened on your site and what contributed. Attribution logic alone opens a 5 to 20 percent gap, because Google Ads claims 100% of the credit for its own click while GA4 may spread that conversion across several channels. The two platforms are not broken when they disagree; they are measuring different things on purpose.
Timing widens the gap further. Google Ads books a conversion to the date of the click, while GA4 records it on the day it actually happened, so a single campaign can look different in each tool for days. Add view-through conversions, which Google Ads counts by default on a one-day window and GA4 does not, plus Google Ads layering its own conversion modeling on top. The fix is not to make them agree. It is to pick one platform as the source of truth per decision and stay consistent.
GA4 and Google Ads disagree on purpose. The goal is not to make them match, but to know which one to trust for each decision.
Consent mode is shaping your numbers
Since March 6, 2024, consent mode v2 has been mandatory for advertisers reaching the EEA and UK, driven by the EU's Digital Markets Act. When a visitor declines tracking, the tag still fires a cookieless signal, and Google reconstructs the missing behavior with modeling. Early Google data shows conversion modeling recovers more than 70% of ad-click-to-conversion journeys lost to consent choices, on average. That recovered slice is an estimate, not an observed count, so a meaningful share of your reported conversions may be modeled rather than measured.
This is not optional plumbing. Without consent mode v2 properly wired up, Google stops collecting new EEA and UK data for audiences and stops sharing it with Google Ads, so your modeling never activates and your numbers quietly shrink. Modeling also needs minimum volume to switch on, around 700 ad clicks over seven days per country. WellBuilt treats consent mode as a measurement decision, not a legal checkbox, because the configuration directly changes how complete every downstream attribution report is.
Data thresholds hide rows you assume are zero
GA4 withholds data to stop individual users being re-identified. When a report segment holds too few users, typically under 50, and the property uses Google Signals, GA4 hides that row entirely. A small-traffic site can open a demographics or search-query report, see far less than it knows occurred, and mistake suppressed rows for real zeros. Look for the thresholding icon in the report header; if it is there, you are seeing a filtered slice, not the full picture.
Two moves reduce it. Widen the date range so more users clear the threshold, and switch the reporting identity from blended or observed to device-based, which relies on first-party client IDs that thresholds do not touch. Device-based reporting drops Google Signals demographics, so it is a trade, not a free win. The point is to know when a report is hiding data from you, so you never build a budget case on a number that GA4 quietly truncated.
Set the lookback window to match your sales cycle
The lookback window decides how far back GA4 reaches to credit touchpoints, and the default is 90 days for most key events and 30 days for acquisition events like first visit. A short window starves long, considered purchases of credit for the awareness touch that started them; a window longer than your real sales cycle credits channels that had nothing to do with the deal. A homeowner often researches a kitchen remodel for months before requesting an estimate, so match the window to how long your buyers actually take, not to the default Google shipped.
Two warnings keep this honest. Lookback changes are not retroactive, so a new window only shapes data going forward and your historical reports stay on the old setting. And the window interacts with the model: data-driven attribution can only distribute credit across touchpoints that fall inside the window, so a too-short window quietly pushes DDA back toward last click. Set it once, deliberately, against your sales cycle, then leave it alone so trends stay comparable over time.
GA4 vs. Google Ads: reading the same conversion
- Asks what contributed to conversions on your site
- Default model is data-driven, splitting credit across touchpoints
- Dates a conversion to the day it actually happened
- Does not count view-through conversions by default
- Subject to data thresholds and consent-mode modeling
- Asks how many conversions its ads drove
- Credits 100% to its own click in last-click setups
- Dates a conversion back to the click or impression
- Counts view-through conversions on a one-day window by default
- Layers its own conversion modeling on top of imported events
Key takeaways
- Confirm your reporting model and channel-credit scope in Attribution settings; small accounts often read last click while assuming data-driven.
- Use the conversion-paths report before cutting any channel, since last click under-credits the assists in a multi-touchpoint journey.
- Stop reconciling GA4 and Google Ads row by row; attribution and timing differences alone create a 5 to 20 percent gap by design.
- Verify consent mode v2 is live for EEA and UK traffic, and treat modeled conversions, up to 70% of lost journeys, as estimates not counts.
- Set the lookback window to your real sales cycle and remember the change is not retroactive, so historical reports keep the old setting.
SourcesSearch Engine Land, Google removes four attribution models in Ads and Analytics, 2023 · Search Engine Journal, Google is removing 4 attribution models for advertisers, 2023 · Google Analytics Help, Select attribution settings and key event lookback window, 2024 · Cardinal Path, An overview of data-driven attribution in GA4, 2024 · Dataslayer / Kissmetrics, Data discrepancies between Google Ads and GA4, 2024 · Google Marketing Platform Blog, Conversion modeling through consent mode, 2021 · iubenda / CookieYes, Google Consent Mode v2 and the Digital Markets Act, 2024 · Google Analytics Help and Analytics Mania, About data thresholds in GA4, 2024 · ALM Corp, GA4 attribution model restructure (April 2026), 2026
Questions, answered straight.
Why are first-click and position-based attribution gone from GA4?
Google retired first-click, linear, time-decay, and position-based attribution across Ads and Analytics by November 2023. The reason was adoption: fewer than 3% of conversions used them, and the rigid rules could not adapt to longer, multi-channel journeys. Today you choose between data-driven attribution and paid-and-organic last click, so build your reporting around those two and stop looking for the retired models.
Why don't my GA4 and Google Ads conversion numbers match?
They answer different questions and will never fully agree. Google Ads credits its own click 100% while GA4 may split that conversion across channels, and Google Ads dates conversions to the click while GA4 dates them to when they happened. Attribution logic alone opens a 5 to 20 percent gap. Pick one platform as the source of truth for each decision rather than trying to reconcile them row by row.
What is data thresholding and how do I avoid it?
GA4 hides report rows when a segment holds too few users, typically under 50, and Google Signals is on, so visitors cannot be re-identified. It often makes small-traffic reports look emptier than reality. Widen the date range or switch the reporting identity to device-based to reduce it, but know that device-based drops Google Signals demographics, so weigh the trade before you change it.
Does consent mode change my attribution numbers?
Yes. When EEA or UK visitors decline tracking, GA4 models the missing behavior instead of observing it, and Google reports modeling recovers more than 70% of lost ad-click journeys on average. That means part of your reported conversions are estimates. Make sure consent mode v2 is properly implemented, since without it Google stops collecting new EEA data and your modeling never activates.
Strategy & Tracking
Want this run for you, not just read about?
Clean tracking and honest attribution, so you know which dollars actually produce revenue.