To judge whether a multi-touch attribution (MTA) platform is credible and unbiased, evaluate it on three levels: methodology, measurement integrity, and commercial incentives.

1) Check the methodology

A credible platform should clearly explain how it assigns credit.

Ask:

• Is the model rule-based, algorithmic, or causal/incrementality-based?
• Does it distinguish between:
  • correlation and causation
  • assistive touchpoints and incremental impact
• Can it explain why a channel got credit?

Red flags:

• “Black box AI” with no transparent logic
• Overly precise claims without methodology
• A model that always favors the channels easiest to measure

2) Verify data quality and identity coverage

Attribution is only as good as the data feeding it.

Check:

• What percentage of journeys are actually tracked?
• Does it handle:
  • cross-device behavior
  • logged-out users
  • cookie loss / privacy restrictions
  • offline conversions
• How are missing touchpoints treated?

Ask:

• What is the match rate?
• What is the deduplication logic?
• How are conversions linked across devices and channels?

Red flags:

• Heavy reliance on deterministic matches only
• No visibility into unmatched traffic
• Claims of “full-funnel attribution” with limited identity resolution

3) Test for incrementality, not just attribution

A platform may assign credit but still overvalue channels that were not truly driving lift.

Ask whether the vendor:

• runs holdout tests
• supports geo experiments
• integrates with conversion lift studies
• compares modeled attribution against incrementality results

Strong sign:

• Their MTA output is validated against experiments, not presented as a replacement for them.

Red flags:

• Attribution reports are treated as proof of business impact
• No mention of causal validation

4) Look for bias in commercial incentives

Some vendors have incentives that can shape results.

Questions:

• Does the vendor sell media, or is it independent?
• Are they optimizing toward certain channels they benefit from?
• Do they have partnerships that could bias credit allocation?

Red flags:

• Platform is tied to a media seller and tends to favor that media
• Vendor is unwilling to discuss conflicts of interest
• “Our platform proves your best channel is the one we help you buy”

5) Evaluate model assumptions

Every attribution model embeds assumptions.

Ask:

• What lookback window is used?
• How are view-through vs click-through interactions handled?
• How are channel interactions and sequencing modeled?
• Are diminishing returns or saturation effects included?
• Can the model adapt by product line, geo, or audience?

Red flags:

• One-size-fits-all lookback windows
• No ability to inspect or tune assumptions
• Attribution results that are implausible relative to campaign timing

6) Demand transparency and reproducibility

A credible platform should let you audit its outputs.

Ask for:

• Model documentation
• Feature definitions
• Sample journey-level outputs
• Versioning of model changes
• Audit logs for adjustments and overrides

Red flags:

• “Trust us” dashboards
• No explanation for shifts in credit over time
• Inability to reproduce prior results

7) Compare against independent benchmarks

Don’t judge by the platform’s report alone.

Compare:

• MTA results vs. experiments
• MTA results vs. MMM (marketing mix modeling)
• MTA results vs. channel-reported conversions
• MTA results vs. business outcomes like revenue, margin, CAC, LTV

Strong sign:

• The platform identifies similar directional patterns to other independent measurement methods.

Red flags:

• Huge divergence from other methods with no explanation
• Only favorable results are highlighted

8) Pilot before you commit

Run a limited test.

Do:

• Choose a few campaigns with clear conversion paths
• Freeze assumptions for a fixed period
• Compare model outputs to controlled experiments or historical baselines
• Evaluate whether the platform changes decisions in a sensible way

Measure:

• stability
• explainability
• lift in decision quality
• consistency across channels and time

---

Practical checklist

A credible, unbiased MTA platform should be able to answer “yes” to most of these:

• Do you clearly explain your attribution logic?
• Can we inspect the underlying assumptions?
• Do you quantify identity coverage and missing data?
• Do you validate against incrementality tests?
• Are you independent from media sales incentives?
• Can outputs be audited and reproduced?
• Do results hold up against MMM and platform-reported data?

---

Bottom line

A credible MTA platform is transparent, experimentally validated, and commercially independent.
If it is black-box, unvalidated, and tied to media sales, treat its recommendations as hypotheses—not truth.

If you want, I can also give you:

1. a vendor evaluation scorecard, or
2. a list of interview questions to use in demos.