TL;DR Perfect attribution is dead. In 2026, the best approach combines platform-reported data, incrementality testing, and media mix modelling. Here’s how to make it work. Book a 20-min call to build an attribution framework that works.
The state of attribution in 2026
Let’s be honest: attribution has never been harder.
Third-party cookies are gone. iOS privacy changes gutted Meta’s tracking. GA4’s attribution is based on incomplete data. Every platform claims credit for conversions they may not have caused.
If you’re still relying on last-click attribution from a single source, you’re making decisions with bad data.
But attribution isn’t dead. It’s just different. Less precise. More probabilistic. And if you accept that, you can still make good decisions.
What’s broken (and what isn’t)
Broken: deterministic cross-device tracking
You can no longer follow a single user from ad impression to website visit to purchase across devices with certainty. That era is over.
Broken: platform-reported conversions as truth
Google claims 100 conversions. Meta claims 80. Your actual total is 120. The sum of platform claims is always higher than reality because they double-count overlapping touchpoints.
Broken: last-click as the only model
Last-click attribution wildly overvalues bottom-funnel channels (branded search, retargeting) and undervalues top-funnel (display, social, video).
Still works: directional signals from platforms
Platform-reported data is biased but not useless. Trends still hold. If Meta’s reported ROAS drops 30% week-over-week, something changed — even if the absolute number is inflated.
Still works: first-party data and server-side tracking
Conversions tracked server-side (Conversions API for Meta, enhanced conversions for Google) recover much of the lost signal. Not perfect. But much better than client-side only.
Still works: incrementality testing
The gold standard. Turn a channel off in a market. Measure what happens. The difference is your true incremental impact. Can’t argue with it.
The three pillars of modern attribution
In 2026, no single method gives you the truth. You need three approaches working together:
Pillar 1: Platform-reported data (tactical)
Use for: Day-to-day optimisation. Budget allocation within a platform. Campaign-level decisions.
Trust level: Directional. Good for trends, bad for absolutes.
How to use it: Track platform-reported conversions, but never sum them across platforms to get a “total.” Use them to optimise within each channel. Compare week-over-week within the same platform.
Pillar 2: Media Mix Modelling (strategic)
Use for: Budget allocation across channels. Understanding the contribution of each channel to total revenue.
Trust level: High for strategic decisions. Low for tactical, day-to-day changes.
How it works: Statistical model that correlates spend (by channel, by week) with outcomes (revenue, leads, sign-ups). Accounts for seasonality, promotions, and external factors.
Who it’s for: Brands spending £50K+/month across 3+ channels.
Pillar 3: Incrementality testing (validation)
Use for: Proving whether a channel actually drives growth. Validating platform claims.
Trust level: Highest. Real-world causal evidence.
How it works: Geo-based holdout tests. Turn off spend in a region. Compare performance against a matched control region. The difference is your true incremental impact.
Who it’s for: Anyone making big budget decisions (start/stop a channel, double a budget).
Practical attribution for mid-market teams
Not every brand can afford a data science team building MMMs. Here’s what works for brands spending £10K-100K/month:
| Approach | Effort | Accuracy | Best for |
|---|---|---|---|
| Platform-reported with blended view | Low | Medium | Daily optimisation |
| GA4 data-driven attribution | Low | Medium-Low | Understanding multi-touch paths |
| Simple MMM (regression in BigQuery) | Medium | Medium-High | Quarterly budget planning |
| Geo-holdout tests | Medium | High | Validating channel value |
| Post-purchase surveys (“how did you hear about us?”) | Low | Low-Medium | Discovering dark social and offline |
The best approach for most teams: use platform data daily, run a simple MMM quarterly, validate with incrementality tests annually.
Setting up a workable attribution system
Step 1: Get your data into one place
You can’t do cross-channel attribution if your data lives in five different platforms. Connect Google Ads, Meta, TikTok, GA4, and your ecommerce/CRM data into BigQuery.
💡 This is what we do. We build the data infrastructure for attribution — connecting all your platforms into BigQuery with automated pipelines and unified dashboards. Book a 20-minute discovery call — no pitch, just scoping.
Step 2: Build a blended performance view
Create a single dashboard that shows:
- Total spend (all channels)
- Total conversions (deduplicated from your own data, not platform claims)
- Blended CPA and ROAS
- Channel mix over time
This gives you the “single source of truth” that platform dashboards can’t provide.
Step 3: Accept ranges, not precise numbers
Attribution in 2026 gives you confidence intervals, not exact numbers. Meta might drive 20-35% of your revenue. Google might drive 30-45%. That overlap is real — many conversions were influenced by both.
Get comfortable with ranges. Make decisions at the bounds. If a channel is profitable even at the low end of your estimate, scale it.
Step 4: Run periodic validation
Every 6 months, test your assumptions:
- Pause a channel for 2-4 weeks in a test market
- Compare revenue in test vs control market
- Adjust your mental model of that channel’s contribution
Common attribution mistakes in 2026
Trusting any single source completely. No platform, tool, or model has the full picture. Triangulate.
Cutting upper-funnel because it doesn’t show direct conversions. Display and video often don’t convert directly. That doesn’t mean they’re not working. Test with incrementality before cutting.
Over-investing in attribution technology. A £50K/year attribution platform won’t give you perfect data. It’ll give you a different flavour of imperfect data. Start simple. Add complexity only when it changes decisions.
Ignoring dark social. People share links in WhatsApp, DMs, Slack. These show up as “direct” in GA4. They’re not direct. Ask customers how they found you.
The role of first-party data
First-party data is your competitive advantage in attribution. The more you can track directly — email engagement, on-site behaviour, purchase history, loyalty program data — the less you rely on crumbling third-party signals.
Build your email list. Implement server-side tracking. Use post-purchase surveys. Invest in customer data infrastructure. This compounds over time.
Know someone drowning in spreadsheets? Share this guide with them.
If this sounds like more work than you want to take on, that’s what we do at Chartica. Book a 20-minute discovery call — we’ll scope it out, no pitch.