Skip to main content
Marketing Agencies

The Marketing Agency's Guide to Multi-Touch Attribution Data

Last-click attribution is lying to your clients. Here's what multi-touch attribution actually requires in terms of data quality — and how to set realistic expectations.

Key Takeaways
  • Adding up each channel's native conversion reporting gives 2–5x the actual conversion count
  • Use one conversion source of truth (GA4 or backend), not each platform's native reporting
  • Consistent UTM parameters across all channels are non-negotiable for any attribution analysis
  • Choose one attribution model and apply it consistently — changing models to get a better story is data manipulation
  • Honest, well-reasoned attribution analysis builds more client trust than false certainty

The Attribution Problem Nobody Wants to Talk About

Every agency has had the conversation: the client wants to know which channel is "really" driving conversions. The honest answer — "it's complicated, and your current data doesn't support a definitive answer" — is not what they want to hear.

The more common (and problematic) answer: pull last-click conversion data from each channel's native platform and present it as truth. The result: every channel takes credit for the same conversion, total attributed conversions exceed actual conversions by 2–5x, and nobody is making good decisions.

Why Attribution Data Is Hard

Cross-device journeys: A customer sees a Facebook ad on their phone, searches on a laptop, reads a review on a tablet, and converts on their phone again. Most attribution models see only the last touch.

Walled garden reporting: Meta, Google, and TikTok each report conversions using their own measurement — and they each take credit. Adding up their reported conversions gives you more conversions than actually happened.

Cookie and tracking limitations: iOS privacy changes, cookie deprecation, and adblockers mean 20–40% of digital journeys are invisible to pixel-based tracking. Models built on incomplete data produce unreliable conclusions.

Sohovi automatically detects PII in your datasets — emails, phone numbers, SSNs — all processed client-side so your data never leaves the browser.

Lookback window inconsistency: A 30-day lookback on one channel and a 7-day lookback on another means you're not comparing the same thing when you compare performance across channels.

What Good Attribution Data Requires

For multi-touch attribution to produce trustworthy insights, you need:

  • A single conversion source of truth: Your actual conversion data should come from your analytics platform (GA4) or your backend, not from each ad platform's native reporting. Reconcile ad platform conversions to actual backend conversions.

  • Consistent tracking parameters: Every ad from every channel uses UTM parameters consistently. Inconsistent UTMs mean some touchpoints are invisible in your data.

  • A unified customer ID: The ability to connect touchpoints from the same user across sessions and devices. This typically requires a logged-in user journey and is unavailable for anonymous browsing.

  • A defined attribution model: Choose a model (linear, time-decay, data-driven) and apply it consistently. Changing models mid-analysis to get a better story is the agency equivalent of cherry-picking.

Setting Client Expectations Honestly

The most valuable thing an agency can do on attribution is be honest about its limits:

"Our best data shows Google Search drives the most last-click conversions. Meta and display appear earlier in the funnel based on view-through data. We can't tell you definitively how much each channel contributed to the final decision — no one can with certainty given current tracking limitations. Here's what we're confident about, and here's what we're measuring more closely."

Clients who are given honest, well-reasoned analysis trust their agencies more — not less.

Frequently Asked Questions

Should agencies use data-driven attribution or a rule-based model?

Data-driven attribution requires significant conversion volume (typically 3,000+ conversions/month) to be reliable. Below that threshold, a rule-based model (time-decay or linear) with consistent application is more defensible.

How do I reconcile Google Ads and Meta conversion reporting?

Export actual backend conversions by date. Compare to each platform's reported conversions for the same period. The ratio of platform-reported to actual conversions is your 'inflation factor' — useful context for interpreting platform metrics.

What's the best way to explain attribution to a non-technical client?

Use the supermarket analogy: 'Multiple things influenced the purchase decision, like a TV ad, an Instagram post, and a Google search. Attributing 100% of the sale to the Google search ignores everything that came before it.'

Selva Santosh

Data quality, for people who ship

Selva writes practical guides on data quality, profiling, and governance to help teams ship better data.

Start for free

Stop guessing. Start knowing your data quality.

Sohovi profiles your datasets in minutes — surfacing completeness gaps, type mismatches, and duplicate patterns before they reach production.

No credit card required · Free forever plan