The dashboard reported attribution to two decimal places. Marketing leaders had learned it was directionally wrong and used it for nothing, while continuing to pay for it.
STUDY 11 / 30 · ANONYMISED · NDA · MARTECH · NETHERLANDS · SERIES B
MARTECH · NETHERLANDS Attribution that states what it cannot know.
RoleLead designer
Timeline11 weeks
Team1 designer, 1 PM, 1 data scientist, 1 frontend
VerticalMulti-touch attribution platform · B2B marketing teams
Replaced false precision with confidence bands and an explicit, prominently sized unattributable bucket. The model states its uncertainty rather than laundering it.
The dashboard began to be used in budget decisions, because leaders could finally defend the numbers in a room where someone would challenge them.
QuarterlyUSED IN BUDGET DECISIONSfrom never
22%UNATTRIBUTABLE, STATEDpreviously distributed
4.1 / 5CMO-REPORTED CONFIDENCEfrom 1.4
THE ARGUMENT
Why the obvious solution was wrong.
The study matters because the product problem was reframed before the interface was polished.
Attribution dashboards fail on credibility, not accuracy. A number reported as 18.42% invites a question the model cannot survive: how do you know the .42? The false precision is a tell, and sophisticated buyers read it correctly as a system distributing its ignorance evenly across channels rather than admitting it. A CMO who cannot defend a number in front of a CFO will not use that number, however good the underlying model.
Every figure now renders with a confidence band sized to the actual uncertainty, and the unattributable share sits in the report at equal visual weight rather than being silently allocated. Where the model has too little data for a channel, it says so instead of producing a number. Counter-intuitively this made the product more useful, not less: leaders started bringing it into budget conversations because its stated limits made its stated findings credible. Honest uncertainty is a feature that has to be designed, not a caveat to be footnoted.
THE INTERFACE CRAFT
The interaction, rendered as a working product surface.
The specimen below is code-native and uses the study's own design logic. The client interface remains protected.
MARTECH
MODEL CONFIDENCE
Paid searchHIGH
EventsLOW
UnknownDECLARED
DESIGN DECISIONS
Positions we would defend.
Each decision names the principle and the product consequence, not a stylistic preference.
Precision is a credibility claim
Two decimal places assert a certainty the model does not have, and sophisticated readers price that in against everything else on the page.
Design the unknown
The unattributable share is the most honest thing the system knows. Hiding it is the decision that destroys trust in the rest.
Refuse to report thin data
A number produced from insufficient volume is worse than an absence, because it will be quoted.
PRODUCT LEADER READOUT
What transfers, and what should remain specific to this product.
A case study is useful when its operating principle travels without turning the original interface into a template.
Read the operating condition
For Multi-touch attribution platform · B2B marketing teams, the transferable lesson is not a copied screen. It is the condition the interface had to make legible: No point estimate without its band. The band is drawn at the same visual weight as the number, not as a footnote. Rebuild that visibility for your own roles, risk, terminology, and operating cadence.
Protect the design rule
Two decimal places assert a certainty the model does not have, and sophisticated readers price that in against everything else on the page. Keep that rule in the acceptance criteria, component states, and production QA record so later visual cleanup cannot erase why the interaction exists.
Measure behaviour after ship
The evidence record is Quarterly for used in budget decisions, from never. Recreate the baseline and outcome window before rollout, segment the result by role and context, and state clearly what the measure cannot prove.
RESEARCH RECORD
The work behind the interface.
These artefacts connect the final interaction back to the evidence and product model that produced it.
CMO credibility interviews
Nine interviews on why attribution dashboards go unused; every one named false precision or an unexplained total.
Uncertainty quantification
Worked with the data scientist to produce genuine confidence intervals per channel rather than a smoothed allocation.
Band rendering test
Tested five visual treatments for confidence bands against correct interpretation by non-technical readers.
Budget-room simulation
Ran the redesigned report through three mock CFO challenges to test which figures survived interrogation.
“It tells me the twenty-two percent it can't explain. That's why I believe the seventy-eight it can.”
CMO, B2B software · under NDA
NEXT STUDY · 12 / 30 · MARTECH