Voice of the customer

ICP evidence log

ICP evidence log is licensed CC BY 4.0. Attribution: Katafacts (katafacts.com).

1 · What it is

What it is

An ICP evidence log is the same clustering discipline as the VOC evidence log — a theme drafted per entry, freely reassignable, with the tally recomputing live — pointed at win/loss call notes, discovery-call excerpts, and competitor mentions instead of product-feedback quotes. The lean connection master-plan §4.2 names for ICP is voice of the customer applied to "specify value — lean's step one": an ICP canvas built from assumption instead of real deal evidence is exactly the impression-not-data mistake this catalogue's own discipline exists to catch everywhere else.

2 · When to use it

When to use it — and when not to

Use it when

You have real win/loss call notes or discovery-call excerpts and want to test whether the ICP canvas's assumptions actually hold up against what happened in real deals.
You suspect a pattern in why deals stall or close fast (a trigger, a buying-committee shape, a segment signal) but haven't clustered the evidence to see if it's real or anecdotal.
You want the vital-few fit signals named with a defensible tally, not just a sales leader's gut sense of 'our best deals all look like X.'

Not when

You don't have real call notes or deal records to draw from yet — this clusters evidence that already exists, it doesn't generate deal history from nothing.
You only have two or three deals total — not enough material for a real pattern to emerge from coincidence.
You already know the answer and just want it confirmed — genuine synthesis means letting the entries suggest the clusters, not the other way around.

3 · How to fill it in

How to fill it in

Scope: Which deals or accounts does this evidence cover, and what prompted pulling it together — sharpening the ICP canvas, testing anti-ICP disqualifiers, something else?
Evidence entries: Enter each win/loss note, discovery excerpt, or competitor mention you've already gathered, with its source, source type, and date.
Themes: Drafted from your notes, then genuinely yours to edit — reassign any entry's theme and the clustering updates immediately.
Narrative: Drafted from the top themes — what the clustering actually suggests about who the ICP is.
Next steps: What will the team do with this synthesis — update the ICP canvas, add a disqualifier, change a qualification step — ranked by impact and effort, owned and dated?

4 · What good looks like

What good looks like

The example below clusters nine win/loss and discovery-call entries into four themes: a positive trigger signal (already-outgrown spreadsheet workarounds), a strong predictor of a clean close (an exec sponsor engaged early) paired with its inverse failure pattern (a stalled single-threaded champion), and two smaller but real anti-ICP signals — a company-size floor and an unplanned enterprise security-review gap.

Same example, as a downloadable xlsx workbook.

Download .xlsx

VOC evidence log

Q1 win/loss patterns — who our real ICP is

Renee Okafor, VP Sales · 2026-03-12

Team: Jordan Ellis, Sales Manager, Priya Anand, RevOps · Sponsor: Renee Okafor, VP Sales

Scope

Covers closed-won and closed-lost mid-market and enterprise deals from Q1 2026 — win/loss call notes and discovery-call excerpts pulled together specifically to test and sharpen the ICP canvas's assumptions against what actually happened in real deals, not a running capture log kept for its own sake.

Evidence entries

Outgrowing spreadsheet workaroundsTop theme

We'd been tracking renewals in four separate spreadsheets before this — nobody could tell you the real number without a day of reconciling.
Palisade Ventures — win · win-loss
Our ops lead said they'd hit the point where the spreadsheet 'actively lies to us' at scale.
Thornbury Group — discovery call · interview
They'd already built two internal tools to patch the gap before they came to us.
Calder Industries — win · win-loss

Exec sponsor engaged earlyTop theme

The VP was in the second call and had already looped in finance before we asked.
Palisade Ventures — win · win-loss
Deal moved fast once the CFO's office asked for the business case directly — we didn't have to chase it.
Meadowlane Co — win · win-loss

Stalled on single-threaded championTop theme

We never got past the single champion — she kept saying she'd loop in her VP and never did.
Ashgrove Retail — loss · win-loss
Champion went quiet for three weeks mid-cycle, and there was no one else in our contact list to check in with.
Driftwood Logistics — loss · win-loss

Sub-20-employee self-serve rarely converts

Signed up for the trial, poked around for a day, never came back — under 20 employees, no real evaluation process.
Small-team trial cohort · other

Enterprise security review, unplanned

Security review request came in at week six, after we'd already built the whole proposal around a week-two close.
Calder Industries — win (delayed) · win-loss

Themes

Top themes: Outgrowing spreadsheet workarounds, Exec sponsor engaged early, Stalled on single-threaded champion

Outgrowing spreadsheet workarounds3 · 33.3%
Exec sponsor engaged early2 · 22.2%
Stalled on single-threaded champion2 · 22.2%
Sub-20-employee self-serve rarely converts1 · 11.1%
Enterprise security review, unplanned1 · 11.1%

Narrative

Outgrowing spreadsheet workarounds is the largest cluster and the clearest positive trigger signal — every account in it was already running a manual patch before engaging, which the ICP canvas's trigger section should name explicitly rather than leave generic. Exec sponsor engaged early is the strongest predictor of a fast, clean close, while its inverse — stalled on single-threaded champion — is exactly the failure pattern the qualification standard work's budget-authority step exists to catch before a deal advances. Sub-20-employee self-serve rarely converts and the enterprise security-review timing gap are both real anti-ICP signal: one a size floor, one a process gap worth fixing in the standard work rather than treating as a one-off surprise.

Next steps

Add 'already running a manual spreadsheet/tool workaround for this problem' as a named trigger question in qualification standard work, not just a firmographic check.
Linked finding: Outgrowing spreadsheet workarounds — largest cluster, three separate accounts
Impact: high · Effort: low · Owner: Priya Anand · Due: 2026-04-01
Add a sub-20-employee size floor and a 'no second stakeholder identified by week three' flag to the anti-ICP disqualifier list.
Linked finding: Sub-20-employee self-serve rarely converts / Stalled on single-threaded champion
Impact: medium · Effort: low · Owner: Renee Okafor · Due: 2026-04-01

5 · Common mistakes

Common mistakes

Clustering win notes only, leaving losses out of the evidence set.
A pattern that only explains why deals win says nothing about why similar-looking deals lose — the anti-ICP disqualifiers this evidence should feed depend on the loss entries just as much as the wins.
Treating a single memorable deal as a pattern.
Same vital-few discipline as every other artifact in this catalogue — one vivid story isn't a cluster; it needs enough independent entries to mean something beyond one account's specifics.
Pulling quotes from memory instead of the actual call notes or customer relationship management (CRM) system's history.
A pattern built from what a sales leader remembers inherits every bias in that memory — the same reason every other evidence-gathering artifact in this catalogue insists on real notes, not recall.

6 · What it connects to

What it connects to

upstream

downstream

ICP canvas
A theme with real weight behind it — a trigger, a buying-committee pattern — is exactly what should update the ICP canvas's definition, not sit unused in a log.
Anti-ICP
A loss-side theme (a size floor, a process gap) is a disqualifier candidate — the anti-ICP list should trace back to evidence like this, not a sales leader's unstated assumptions.
Deal A3 — win-rate root cause analysis
A theme that concentrates in lost pipeline value, not just deal count, is exactly the kind of finding a deal A3's Pareto would chase to root cause.

7 · Where AI helps

Where AI helps

Judgement — stays yours

Deciding whether a drafted cluster genuinely reflects a real deal pattern, or moving an entry to a different theme
Deciding which theme is real fit signal versus coincidence worth watching, not yet acting on

Analysis — AI helps

Drafting a first-pass theme label per entry from its actual content
Drafting the narrative from the top themes
Flagging when a theme is thin (one or two entries) rather than a defensible pattern

Drudgery — automated

Computing theme counts and percentages, recomputed live on every reassignment
Identifying the vital-few top themes
Exporting to xlsx in the house format

9 · Rate this kata