Quick answer: Attribution is how marketers decide which ads, clicks and channels get credit for a conversion. The attribution model and attribution window you choose decide which campaigns look profitable, and therefore where your next rupee of ad budget goes.
A customer rarely buys on the first touch. Someone searches “running shoes Mumbai” and clicks a Google Search ad, leaves, sees an Instagram retargeting ad three days later, then types your brand name into Google and buys. That is one sale and three touchpoints. Attribution is the set of rules that splits credit for that one sale across those three touchpoints.
The rules are not neutral. Last-click attribution hands the entire sale to the branded search ad and makes your prospecting campaigns look like a waste of money. Data-driven attribution splits the credit across all three. Same customer, same ₹4,000 order, three completely different stories about which campaign to scale. Attribution does not measure reality; it applies a rule to incomplete data, and different rules produce different answers from the same sale.
Why attribution matters
Attribution is the layer between your ad platforms and your budget decisions. Every “pause this campaign” and “double the budget here” call traces back to which touchpoint your reporting gave credit to. Get the model wrong and you defund the campaigns that actually create demand.
The most expensive symptom is double counting. Google Ads and Meta both report on their own conversions using their own rules, and neither knows the other exists. Google Ads and Meta each claim credit for the same sale under their own attribution rules, so adding their reported revenue together will almost always exceed the revenue that actually landed in your bank account. Meta’s default setting for website conversion campaigns is 7-day click, 1-day engage-through and 1-day view-through, which is more generous than what GA4 counts by default (Jon Loomer, 2026).
The tracking underneath has also thinned out. Safari and Firefox have blocked third-party cookies by default for years, and while Google reversed its plan to remove them from Chrome in 2024 and confirmed in April 2025 that it would not add a choice prompt either, the wider signal loss did not reverse (Objective Platform, 2026). Practitioners now estimate that user-level multi-touch attribution sees roughly 30% to 60% of journeys, down from over 90% before the privacy changes (Leadgen Economy, 2026).
So the honest position in 2026: attribution is directional, not exact. It is still the best tool you have for day-to-day optimisation, and it is a bad tool for settling arguments about how much a whole channel is worth.
Attribution vs MMM vs incrementality
These three get lumped together as “measurement” and they answer different questions.
| Attribute | Last-click attribution | Data-driven attribution | Marketing mix modelling (MMM) | Incrementality testing |
|---|---|---|---|---|
| Question it answers | Which touchpoint closed the sale? | How much did each touchpoint contribute? | How much did each channel contribute to total revenue? | Would this sale have happened anyway? |
| How it works | Fixed rule: 100% credit to the final click | Machine learning on your converting and non-converting paths | Statistical model on aggregate weekly spend and sales | Holdout or geo test: some users see no ads |
| Data needed | Basic conversion tracking | High conversion volume, clean tagging | About 2 years of weekly spend and revenue data | Enough spend and volume to detect a lift |
| Handles offline and untracked? | No | No | Yes | Yes |
| Main blind spot | Ignores everything before the last click | Only sees inside one platform’s walls | Too slow and coarse for daily decisions | Costs real revenue to run; snapshot in time |
| Best used for | Short, direct purchase paths and audits | Daily campaign and bid optimisation | Setting budget splits across channels | Settling “is this channel real?” arguments |
Attribution guides daily optimisation, incrementality proves causality, and MMM sets the budget across channels; no single one of them replaces the other two. Google’s Meridian and Meta’s Robyn are both free and open source, so MMM is no longer restricted to brands with enterprise budgets (Digital Applied, 2026).
How to set up attribution properly
Six steps, in order. Do them once and your reporting stops arguing with itself.
- Pick one source of truth. Your backend, CRM or e-commerce platform holds the real revenue number. Platform dashboards are inputs, not verdicts. Decide this before anyone builds a report.
- Set the model in Google Ads. You have two options: data-driven attribution (the default for most new conversion actions) and last click. First-click, linear, time-decay and position-based were retired and are no longer supported (Google Ads Help). Data-driven is the right default because Smart Bidding uses that signal.
- Set the model in GA4. GA4 offers data-driven, paid and organic last click, and Google paid channels last click. Choose data-driven with paid and organic channel credit, then verify it is actually running: data-driven attribution needs roughly 400 conversions for that key event and 20,000 conversions across all events in the lookback window, and GA4 quietly falls back to last click when you are under the threshold.
- Match the windows to your sales cycle. Google Ads defaults new conversion actions to a 30-day click-through window and lets you set 1 to 90 days (Google Ads Help). Meta removed 7-day view and 28-day view in January 2026, so 7-day click plus 1-day view is now the widest window available there. If your buyers take three weeks to decide, a 7-day window is not measuring your business.
- Repair the signal. Server-side tagging, Meta’s Conversions API, Google’s enhanced conversions, consistent UTM tags on every paid and email link, and GCLID or fbclid stored against the lead in your CRM. Attribution models are only as good as the events reaching them.
- Sanity-check with blended numbers. Divide total revenue by total marketing spend (MER) every week. If MER is flat while platform-reported ROAS is climbing, the models are inflating, not the business.
Here is what the double-counting problem looks like in rupees, for one month:
- Google Ads spend: ₹1,50,000, reported revenue ₹6,00,000 (4x ROAS)
- Meta Ads spend: ₹1,00,000, reported revenue ₹4,00,000 (4x ROAS)
- Platforms combined claim: ₹10,00,000
- Actual revenue in Shopify: ₹7,00,000
The platforms have overclaimed by ₹3,00,000, roughly 43% of your real revenue, because the same customers were credited twice. The blended reality:
MER = total revenue / total marketing spend = 7,00,000 / 2,50,000 = 2.8x
Neither channel is running at 4x. The account is running at 2.8x, and the split between Google and Meta is a question for an incrementality test, not a dashboard. Run the same calculation every month and the gap between claimed and actual becomes your working overlap factor.
Common mistakes
- Adding platform-reported conversions together. Google and Meta both count the same sale, so the total overstates reality. Always reconcile against your own backend revenue before reporting a number to anyone.
- Comparing platforms on different windows. Meta’s 7-day click plus 1-day view and GA4’s last non-direct click are not measuring the same thing, so a “discrepancy” between them is usually a definition difference, not a bug. Compare each platform to itself over time instead.
- Assuming data-driven attribution is switched on. GA4 falls back to last click when volume is too low and does not announce it. Open the model comparison report: if both models show identical numbers, data-driven is not active.
- Leaving traffic untagged. Email, WhatsApp, influencer and affiliate links without UTM parameters collapse into Direct, and a model cannot give credit to a touchpoint it never saw. Tag every non-organic link the same way, every time.
- Judging upper-funnel campaigns on last click. Prospecting, YouTube and awareness campaigns rarely close the sale, so a last-click view will always show them failing. Judge them on assisted conversions, new-customer volume and holdout tests, and change model or window settings only at a documented breakpoint so your trendline stays readable.
Frequently asked questions
What is attribution in digital marketing?
Attribution in digital marketing is the process of assigning credit for a conversion to the ads, clicks and channels that came before it. It uses an attribution model (the rule for splitting credit) and an attribution window (how far back the rule looks). Attribution determines which campaigns appear profitable and where budget gets allocated next.
What attribution models can I use in Google Ads in 2026?
Google Ads supports two attribution models: data-driven attribution and last click. First-click, linear, time-decay and position-based were removed, and conversion actions that used them were migrated to data-driven attribution. Data-driven is the sensible default because Smart Bidding optimises against that signal; last click is worth keeping only for very short, direct purchase paths.
Why don’t Google Ads, Meta and GA4 numbers match?
Because each one uses different rules. Meta counts conversions on a 7-day click and 1-day view basis by default, Google Ads uses data-driven attribution within its own network, and GA4 uses session-based cross-channel attribution. All three are counting the same sales in different ways, so the numbers will never tie out exactly. Pick one as your reporting standard and use the others for optimisation only.
What is an attribution window and what should mine be?
An attribution window is the period after a click or view during which a conversion still gets credited to that ad. Match it to your actual sales cycle: a 7-day window on a business with a 21-day consideration period will systematically hide real conversions. Check the time-lag report in Google Ads to see how long your buyers actually take before you pick a number.
Is multi-touch attribution still worth doing in 2026?
Yes, but as one input rather than the final word. Privacy changes and cross-device journeys mean user-level tracking now misses a large share of touchpoints, so multi-touch attribution is best used for in-platform optimisation. Pair it with blended MER for a weekly reality check and incrementality tests or MMM for budget decisions across channels.
How Uginix helps
Uginix builds the measurement layer before it builds campaigns: server-side tagging, GA4 and CRM alignment, conversion windows matched to your real sales cycle, and one agreed source of truth so nobody reports two different ROAS numbers for the same month. Our analytics and attribution work reconciles what Google and Meta claim against the revenue that actually landed, and if a channel is not contributing incrementally we say so and reallocate. Not sure where your ad spend is leaking? That is exactly what our free growth audit finds.
Related terms
- ROAS (Return on Ad Spend): revenue divided by ad spend; the number attribution rules are quietly deciding for you.
- Attribution window: the period after a click or view in which a conversion still gets credited to that ad.
- MER (Marketing Efficiency Ratio): total revenue divided by total marketing spend; the blended check that catches attribution inflation.
- Incrementality: the share of conversions that happened because of the ads, measured with holdout or geo tests.
- Assisted conversion: a touchpoint that appeared in the path to a sale without being the final click.
Related reading: how Google Ads automation uses your conversion signal in 2026 and what to hand over to Meta Advantage+ and what to keep.
Sources: Google Ads Help, About attribution models, Google Ads Help, About conversion windows, Jon Loomer, Meta ads attribution in 2026, Leadgen Economy, cookieless attribution stack, Objective Platform, cookieless measurement.