Enriched records arrived as flat fields. When a rep found one wrong, they discarded the whole record - and eventually the whole product - because nothing distinguished a verified field from a guess.
STUDY 07 / 30 · ANONYMISED · NDA · SALES AUTOMATION · INDIA · SERIES B
SALES AUTOMATION · INDIA Data enrichment where the source is always one click away.
RoleLead designer + systems
Timeline10 weeks
Team2 designers, 1 PM, 1 data engineer
VerticalGTM data platform · 340 revenue teams
Attached provenance, confidence and age to every individual field, rendered ambiently, with a one-click correction that writes back to the source.
Reps stopped discarding records wholesale. Correction volume rose sharply, which improved the underlying data faster than any vendor change had.
8%RECORDS DISCARDED ON ONE BAD FIELDfrom 61%
2,400FIELD CORRECTIONS / WEEKfrom 40
4.2 / 5REP-REPORTED DATA TRUSTfrom 1.8
THE ARGUMENT
Why the obvious solution was wrong.
The study matters because the product problem was reframed before the interface was polished.
Enrichment products present a company record as though every field is equally true. In reality one field came from a verified filing, one from a scraped page, and one from a model inferring headcount from a hiring page. When the inferred one is wrong the user has no way to isolate the failure, so they generalise it. We watched reps delete entire enriched lists over a single wrong phone number.
Every field now carries three ambient properties: where it came from, how confident the source is, and how old it is. These render as a compact three-part marker that costs almost no space and requires no interaction to read. A wrong field is corrected inline in one action, and the correction writes back to the provider with attribution. Corrections rose sixtyfold, which turned the user base into the largest single quality mechanism the product had. Trust rose because failure became local instead of total.
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.
SALES AUTOMATION
BUYING GROUP
DESIGN DECISIONS
Positions we would defend.
Each decision names the principle and the product consequence, not a stylistic preference.
Make failure local
A product where one wrong field discredits the record is a product with no error tolerance. Provenance localises the damage.
The user is the QA layer
Design correction as a one-action path and the user base becomes the fastest quality mechanism available.
Age is not metadata
In GTM data, a field's age is nearly as important as its value. It belongs on the surface, not in a tooltip.
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 GTM data platform · 340 revenue teams, the transferable lesson is not a copied screen. It is the condition the interface had to make legible: Source, confidence and age as a compact ambient marker on every field. Legible at a glance, expandable on hover, costing four pixels of density. Rebuild that visibility for your own roles, risk, terminology, and operating cadence.
Protect the design rule
A product where one wrong field discredits the record is a product with no error tolerance. Provenance localises the damage. 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 8% for records discarded on one bad field, from 61%. 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.
Discard-behaviour study
Instrumented and replayed 200 sessions where a rep abandoned an enriched list; 61% followed a single incorrect field.
Provenance taxonomy
Classified all 34 enriched fields by source type and built a four-grade confidence model with the data team.
Marker density test
Tested six visual treatments for the field marker against table scanning speed; the winning one cost 4px.
Correction loop pilot
Ran the write-back loop with 20 teams for three weeks before general release to validate provider ingestion.
“One wrong number used to kill a whole list. Now they fix it in a second and move on.”
Head of RevOps, logistics SaaS · under NDA
NEXT STUDY · 08 / 30 · SALES AUTOMATION