Elanco Knowledge Systems · Child 01 of 3 Child 01 / Continuous Discovery · 2022–2023

Who actually uses company knowledge — and where it breaks.

Continuous Discovery produced the role taxonomy and service model the rest of the program built on. Double-diamond, stakeholder interviews across four functions, journey + service mapping, RACI. The deliverable wasn't a report — it was a structure other people could design against.

DiscoveryService designStakeholder interviewsRACIPersona dev
Company
Elanco Animal Health
Timeline
Sept 2022 — Mar 2023 · 6 mo
Role
Lead Product Designer · research lead
Participants
n = 22 · Regulatory, Vet Affairs, Commercial, Support
Sites
Indianapolis, IN · Hook, UK · remote
Methods
Interviews · journey mapping · service blueprint · RACI
01 / JourneyKnowledge as a baton

A regulatory writer's morning becomes a customer's afternoon — if the system works.

Every Elanco knowledge handoff happens between these four actors, in this order. The whole project's job was to make the spaces between the lanes legible.

Actor
Originate
Verify
Transform
Land
Act
Regulatory writer
Drafts approved language for a new SKU.
Confirms FDA / EMA wording with legal.
No structured handoff to vet affairs.
Veterinary affairs
Reviews source for clinical accuracy.
Re-frames into clinical context.
Loses provenance during transform.
Commercial
Writes customer-facing version.
Cites no source. Re-derives next time.
Sends to customer in email / call.
Customer (vet clinic)
Receives an answer, no traceability.
Trusts or escalates back up the chain.

The tinted cells are where information loss was measurable: missing structured handoff, lost provenance, missing citations. Those three cells became the brief for everything that followed.

Discovery synthesis01 / 05
22 interview transcripts. Loose quotes. No structure yet.
01 · Intake

22 interviews. Loose quotes. No structure yet.

First pass: read everything, tag by speaker function. Resist the urge to theme prematurely. Every interview gets at least one quote that does not fit anywhere yet. Those unsorted quotes become the heaviest signal at synthesis.

n = 22 · 6 weeks · 4 functions interviewed

02 · Cluster

Patterns emerge by function, not by topic.

Initial clustering attempted by topic (dosing, contraindications, market positioning). It didn't compress. Re-clustering by speaker function did — every function had its own version of the same complaint about handoff.

7 functional clusters · 142 coded statements

03 · Loss

Information fidelity drops at every handoff.

Measured handoff loss by tracing 12 customer-facing answers back to their regulatory source. 61% traced back to a regulatory or veterinary source. Only 8% cited it. The bottleneck wasn't access. It was structured transformation.

Trace n = 12 · fidelity loss avg 53%

04 · Model

The service model isn't a process diagram. It's a contract.

Service blueprint v1 looked like a process diagram and stakeholders ignored it. v2 reframed it as a contract between roles — what each function commits to producing, what shape it has, and what the next role can depend on. That's when execs paid attention.

v1 → v2 turnaround: 3 weeks

05 · Taxonomy

Output: a role taxonomy that named the design system.

Discovery's last deliverable wasn't the research write-up. It was a naming taxonomy — RegulatoryCard, VetBrief, CommercialThread, CustomerInquiry — that became the EDS information architecture. The design system was named by the research.

→ EDS · → ElancoGPT

02 / ArtifactsImage slots reserved

The artifacts that did the work.

03 / OutcomesWhat discovery delivered

The work that named everything downstream.

Interviews224 functions · 2 sites
Provenance loss53%Measured handoff fidelity drop
Role taxonomy4Roles · adopted as EDS naming
Cited byEDS + GPTBoth downstream projects built on this
04 / ReflectionDefend / redo
What I'd defend

Discovery as a contract, not a report.

The output had to be something downstream designers and engineers could build against. A research write-up couldn't do that. A role taxonomy could.

What I'd do differently

Pilot the journey-map v1 sooner.

The v1 process-diagram dead end cost three weeks. A 30-minute stakeholder review after the first cluster would have caught it.