Attribution Evidence Triangulation
Combine imperfect signals into an ongoing verdict on campaign impact
- Difficulty
- Expert
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 97%
Attribution Evidence Triangulation treats measurement as an ongoing investigation rather than a solved dashboard. First clarify whether the business decision favors growth or profitability, because that objective affects which attribution lens is useful. Map the relevant touchpoints and inspect available cookie, platform, and first-party data, then choose a transparent provisional model such as first-touch, last-touch, or weighted multi-touch. Do not stop there. Cross-check its conclusion with other evidence, including customer or population surveys and statistical approaches such as media mix modeling. Investigate disagreements instead of averaging them away. The combined evidence may validate a campaign, narrow its plausible impact, or show that it does not work. That negative result is valuable because it directs capital away from an unsupported activity.
Origin
Becker defines attribution as the relationship between what a marketer did and what happened. Apple's IDFA changes weakened a formerly common data source, but he stresses that attribution was always subjective and that privacy shifts made the need for multiple validation methods more obvious.
Core principles
- 01No attribution method is a single source of truth
- 02The chosen lens must reflect whether the business prioritizes growth or profitability
- 03First-touch, last-touch, and multi-touch models embed subjective choices
- 04Cookie data, surveys, and statistical models answer different parts of the question
- 05A valid investigation must be allowed to conclude that a campaign does not work
How to run it
- 1
Define the decision objective
State whether the company is primarily optimizing profitable return, faster growth, or another explicit outcome. Use that objective to judge which attribution perspective is decision-relevant.
Pro tip Write the objective before choosing the model so the method is not selected to defend a preferred campaign.
- 2
Map the observable journey
List digital and offline touchpoints, including paid social, search, television, magazines, and billboards where relevant. Mark which interactions can be observed reliably and which have been weakened by privacy or platform limits.
Pro tip Treat missing data as an explicit property of the model, not as zero impact.
Watch out Platform-reported attribution may omit or over-credit interactions outside its own view.
- 3
Choose a provisional lens
Apply a clear attribution rule, such as first-touch, last-touch, or a stated multi-touch weighting, to create an initial view. Document the subjective assumptions built into the choice.
Pro tip Run more than one lens when the decision changes materially under different reasonable weightings.
Watch out A familiar model is not automatically an accurate model.
- 4
Add independent evidence
Compare the provisional result with customer surveys, population surveys, and statistical modeling such as media mix modeling. Use independent methods because they have different blind spots.
Pro tip A tool such as Recast can support media mix modeling, while larger organizations may build bespoke models.
- 5
Resolve the verdict
Examine where methods agree, disagree, or remain inconclusive. Decide whether the evidence supports continued investment, a smaller test, movement to another channel, or stopping the campaign.
Pro tip Preserve 'does not work' as a legitimate and useful conclusion.
Watch out Sophistication does not guarantee a positive answer; forcing one destroys the purpose of validation.
- 6
Reopen the investigation
Update assumptions and rerun the analysis as privacy rules, channel mix, customer behavior, and business goals change. Attribution remains a maintained decision system rather than a finished calculation.
In the wild
Becker says Thrive has helped sophisticated organizations build campaigns and detailed attribution systems, then cross-validated the results with media mix modeling or other tools. In some cases, the broader evidence indicated that the campaigns did not work despite the sophistication of the setup.
→ The company gained evidence to remove budget from those channels or invest it elsewhere rather than preserving an unsupported program.
Common mistakes
Claiming a single source of truth
Every tool and model observes only part of the customer journey and embeds assumptions about credit.
Hiding the business objective
A model chosen without a growth or profitability objective can produce technically coherent but strategically irrelevant credit assignments.
Demanding a positive campaign verdict
An attribution process that cannot conclude a campaign failed is advocacy, not validation.
Is it for you?
Best for
Organizations running multiple digital and offline channels that need defensible budget decisions despite incomplete tracking.
Not ideal for
Teams seeking a one-time tool installation that will permanently and precisely assign every conversion.
From the episode
Mastering paid growth
Jonathan Becker (Thrive Digital)