A buyer sees your YouTube ad on Monday, clicks a Shopping ad on Wednesday, abandons cart, returns via a retargeting Display ad on Friday, and purchases through a brand Search ad on Saturday. Under last-click attribution, Brand Search gets 100% of the credit. YouTube, Shopping and Display get zero. That reporting says YouTube and Display are worthless — when they actually started the entire journey.
E-commerce buyers rarely purchase on their first interaction. They discover a product through one channel, research it through another and buy through a third. Attribution determines which channel gets credit for the sale — and that credit directly shapes where budget gets allocated. Get attribution wrong and the campaigns that initiate demand get starved while the ones that merely capture it are over-funded.
A typical e-commerce buyer journey
Discovers the product in a video ad. Watches 70%. Does not click.
Searches "buy cotton kurta online." Clicks the Shopping ad, browses three products, leaves.
Sees a retargeting ad while reading the news. Clicks, adds to cart, abandons.
Searches the brand name. Clicks the brand ad. Completes the purchase. ₹2,800.
Under last-click: Brand Search gets the full ₹2,800 of credit. YouTube, Shopping and Display retargeting get ₹0 — despite touching every stage that led to the sale.
YouTube and Display show zero direct conversions under last-click, so they lose budget first. But they created the demand that Shopping and Brand Search captured. Remove them and the whole funnel shrinks — while the cause stays invisible in the report.
Brand Search claims most conversions at the lowest CPA and looks like the best-performing campaign. In reality, it is capturing demand created elsewhere. Doubling its budget would not double conversions, because the bottleneck sits upstream, not at the brand query.
PMax touches multiple surfaces. Under last-click, it absorbs conversions from Display views and YouTube impressions that other campaigns initiated. A "12x ROAS" on PMax might be 5x once credit is distributed fairly across the journey.
Google deprecated most rule-based models — first-click, linear, time-decay, position-based — in 2023 and made data-driven attribution (DDA) the default. That was the right call for most accounts, but understanding how the models differ still helps in reading reports and making sharper budget calls.
The same journey, credited two different ways (illustrative split)
DDA uses machine learning to analyse every conversion path in the account and distributes credit based on which touchpoints actually influenced the purchase decision. It assigns fractional credit — a Shopping click that appears in 80% of conversion paths gets more credit than a Display impression that appears in 20%.
100% of credit goes to the final click before purchase. Simple to understand but systematically biased toward bottom-of-funnel campaigns (Brand Search, retargeting) and against top-of-funnel ones (YouTube, Display, non-brand Shopping). Use it only as a secondary comparison view, never as the primary model for budget decisions.
Still available as a reporting comparison. Not recommended for bid optimisation.
Google deprecated these rule-based models in mid-2023. Position-based (40% first, 40% last, 20% middle) was useful for manual analysis, but all of them applied arbitrary fixed weights that did not reflect actual buyer behaviour. Conversion actions using these models were automatically migrated to DDA. References to them in older reports or agency proposals no longer apply.
Selecting DDA in conversion action settings is the first step, not the last. Real attribution work means using multiple data views to understand what each campaign contributes to the full purchase journey — and making budget decisions that account for assisted value, not just last-touch credit.
The Assisted Conversions report in Google Ads shows how many conversions each campaign assisted (appeared in the path) versus how many it directly converted (was the last click). A campaign with high assist volume and low direct conversions is a demand initiator — cutting its budget reduces conversions across other campaigns downstream.
Example from a real e-commerce account:
| Campaign | Direct conv. | Assisted conv. | Assist ratio |
|---|---|---|---|
| Brand Search | 142 | 18 | 0.13 |
| Shopping — Tier 1 | 89 | 67 | 0.75 |
| PMax | 74 | 112 | 1.51 |
| Display retargeting | 12 | 95 | 7.92 |
| Non-brand Search | 31 | 54 | 1.74 |
Display retargeting has an assist ratio of 7.92 — it participates in eight times more conversion paths than it directly closes. Judging it by last-click alone would make it look like a failure.
GA4’s Advertising → Conversion Paths report shows the actual sequence of channels a buyer touched before purchasing. This is the only place to see the full cross-channel journey — Meta Ads, organic search, direct, email and Google Ads together. We review this monthly to identify which channel combinations produce the highest conversion rates and AOV. If "Meta Ads → Google Shopping → Brand Search" produces double the AOV of "Google Shopping → Brand Search" alone, that shows Meta is driving higher-intent buyers into the Google funnel.
The ultimate attribution question is: if this campaign turns off, how many of its conversions disappear entirely versus shift to another campaign? That is incrementality testing. We run it on campaigns with questionable attribution — particularly Brand Search (would those buyers have come through organic anyway?) and PMax (how much of its ROAS is genuinely incremental?). The method: pause the campaign for two to four weeks in a controlled test, measure the impact on total account conversions, and compare against the credit the campaign was claiming. The gap between claimed credit and actual incremental impact reveals the campaign’s true value.
We report both numbers every month. Channel ROAS is the ROAS per campaign type (Shopping ROAS, PMax ROAS and so on) — useful for campaign-level optimisation. Blended ROAS is total revenue divided by total ad spend across all campaigns — the number that tells you whether the overall Google Ads investment is profitable. A campaign can carry a "bad" channel ROAS but still improve blended ROAS by feeding conversions into other campaigns. Display retargeting at 2x ROAS looks weak in isolation — but if removing it drops blended ROAS from 5.8x to 4.2x, it is clearly pulling its weight.
Display retargeting rarely gets the last click — buyers see the ad, remember the product and return through Search or direct. Its value sits almost entirely in assists. Cutting it removes a critical re-engagement touchpoint, and Shopping and PMax ROAS drop two to four weeks later — by which point the cause is no longer obvious.
Brand Search ROAS is high because it captures demand created elsewhere. Doubling its budget does not double demand — it just increases impression share on queries already being won. The incremental return on extra Brand Search spend is close to zero in most accounts.
Google Ads and Meta have separate attribution systems, different conversion windows and different counting methodologies. A 5x ROAS in Google Ads and a 3x ROAS in Meta do not mean Google is better — Meta may be initiating the journey that Google closes. The only fair comparison happens in GA4, where both channels appear in the same conversion path.
View-through conversions — someone saw but did not click the ad, then later purchased — are over-counted by default. But dismissing them entirely misses real impact too. The right approach: report them separately, look at the view-through-to-click ratio, and discount view-through value by 50-80% rather than counting it at full value or zero.
Google Ads uses DDA by default; GA4 also defaults to DDA, but with a different model and data set. Reporting from both platforms without noting the model difference produces conflicting numbers for the same campaigns. We standardise reporting to use Google Ads DDA for campaign-level decisions and GA4 for cross-channel analysis.
Attribution models estimate credit distribution. Incrementality tests measure actual impact. Without ever running a pause test or geo-lift experiment, six-figure budget decisions rest entirely on modelled estimates. We run at least one incrementality test per quarter on the campaign with the most questionable attribution — usually Brand Search or PMax.
Confirm every conversion action uses DDA. Set conversion windows based on the actual purchase cycle length.
Assisted conversions, conversion path analysis, blended vs channel ROAS, and assist ratio trends per campaign type.
Monthly review of top conversion paths including Meta, organic, direct and email alongside Google Ads campaigns.
Pause tests or geo-lift experiments on campaigns with the most questionable attribution — typically Brand Search or PMax.
A live dashboard with blended ROAS, channel ROAS, assist ratios and conversion lag data, updated daily.
Attribution data translated into specific budget recommendations — which campaigns to scale, reduce or test further.
DDA in the conversion settings is the starting point, not the finish line. It still takes a monthly review of the Assisted Conversions report (most accounts never open it), GA4 conversion path analysis for cross-channel insight, conversion windows set to match the buying cycle (the default 30-day window may not fit), and quarterly incrementality tests to validate what the model reports. The attribution model distributes credit — someone still needs to interpret what that distribution means for budget.
The only way to know for certain is an incrementality test. Pause the campaign for two to four weeks and measure the impact on total account conversions. If conversions drop by the number the campaign was claiming, it is genuinely incremental. If total conversions barely move, those sales were happening anyway through other channels. We run this quarterly — typically on Brand Search first, the most common source of over-attribution, then on PMax.
Both — they serve different purposes. Use Google Ads DDA for campaign-level bidding and budget decisions inside the platform. Use GA4 for cross-channel analysis — understanding how Google Ads, Meta Ads, organic search, email and direct traffic work together in conversion paths. The numbers will not match exactly because the models and data sources differ. That is normal; the insight comes from reading both views, not reconciling them into one number.
Not necessarily. Meta and Google use different attribution systems with different counting windows and methodologies. Meta defaults to a 7-day click / 1-day view window and counts view-through conversions aggressively. Google Ads DDA distributes credit across the click path. Comparing them directly is like comparing different rulers. The fair comparison happens in GA4, where both channels appear in the same conversion path report under the same attribution model — that is where the actual relative contribution of each channel shows up.
Significantly. PMax touches multiple surfaces — Shopping, Display, YouTube, Search, Gmail, Discover — and absorbs credit from all of them. Under DDA, PMax gets fractional credit from every path it participated in, which tends to be generous because PMax impressions show up in a large share of conversion paths. Add brand query cannibalisation (where brand exclusions are not active) and view-through conversions (included by default), and PMax ROAS can be inflated by 30-60%. The first move with PMax attribution is separating brand-excluded ROAS, click-only ROAS and assisted ROAS — three different views that together show the real picture.
We will review your attribution setup, assisted conversion patterns and PMax credit distribution — and tell you which campaigns are over-credited and which are under-valued.
Share your website, the work already underway and the problem you want help with. We can discuss whether management, an audit or consulting fits the situation.