The story How we approached it What it took The sales motion What I learned The workflow Do it yourself

BUILD LOG · THE ICP PLAYBOOK

A real ICP is not a headcount band and an industry. It is a person, their day, and the problem in their life.

Most teams call a firmographic filter an ICP: company size, a vertical, a tech tag. That is a list, not an ICP. The real one is who the buyer is, how their day works, the environment they operate in, and the problem you actually remove, and you derive it from evidence, not a workshop guess.

Steal this build ↓See the stack The people part →

~3x

paid conversion when the signup fits the ICP, 1 in 31 vs 1 in 95

2.6x

off-ICP accounts churn more, 47.8% vs 18.5%

6.9 days

median time to buy, no IT and no implementation in the loop

248

different titles bought it: the ICP is a job, not a title band

THE STORY, FROM THE SEAT

Here is one I actually shipped. What happened, what I did, and the stack if you want to run it yourself.

Ask most teams for their ICP and you get a filter: 50 to 500 employees, B2B SaaS, uses a CRM. That is a targeting list, and it is why the pipeline is full of accounts that look right and never buy.

A real ICP answers a different question: who is the person, what does their day look like, what environment are they selling in, and what problem in their actual life do you remove. You cannot guess that in a positioning workshop. You have to derive it from the people who already bought.

Keep Building,

Heath

FOUNDER, THE SALES OPERATOR

THE BUILD IN FIVE MOVES · THE GIANT METHOD

GMost teams call a firmographic filter an ICP, which is why the pipeline fills with accounts that look right and never buy.
IInterviews with real buyers, G2 and Gong for their words, twelve months of closed-won in Salesforce, product usage in Amplitude.
AThe operator talks to real buyers and writes the personas; AI mines the language, backtests the wins, and reconciles the profile.
NThe person and the problem in the ICP became the exact discovery questions reps asked, so every call started informed.
TSignups that fit the ICP converted to paid about three times as often, and off-ICP accounts churned 2.6 times more.

RUN THE PLAY YOURSELFThis exact build is a playbook. Run it on your own data.

→

01 / HOW WE APPROACHED THE PROBLEM

Ran the loop.

The starting point was a refusal: 50 to 500 employees, B2B SaaS, uses a CRM is a targeting list, not an ICP. A real ICP answers who the person is, what their day looks like, and what problem in their actual life you remove. You cannot guess that in a positioning workshop. You derive it from the people who already bought.

1
Interview the people who bought
Talk to real customers, not a panel. Why they came, why they bought, why they stayed, and the exact words they use for the problem. That language is the raw material for everything downstream.
2
Mine G2 and the calls for their words
Pull the voice of the customer from G2 reviews and call transcripts: what they praise, what they complain about, the job they hired you for, in their language, not your marketing.
3
Backtest closed-won for the pattern that repeats
Read twelve months of won deals for the shape that actually recurs, not the logos you wish you had. Who looks like a win across firmographic, technographic, and how they use the product.
4
Reconcile it into one profile
Merge the inputs and reconcile the divergences against a real data cohort, so the ICP survives scrutiny instead of being one loud opinion. One ICP can carry two motions when the underlying person and job are the same.
5
Write the personas: the person, not the size band
Turn the profile into personas that describe the human, their day, their environment, the problem you solve, and the hard disqualifiers. A persona a rep can picture, not a headcount band.

02 / WHAT IT TOOK CROSS-FUNCTIONALLY

Nobody ships this alone. Here is who had to move.

An ICP is a company-wide claim about who the product is for. Getting it out of a slide meant the functions that touch the customer built it together.

PRODUCT AND DATA

The won-cohort usage patterns.

The real ICP came from how winning customers actually behaved, and that behavioral read lived in product data, not in anyone's opinion.

MARKETING

The persona language and the channels.

The scored ICP set who to pursue; marketing carried the words and the motion mix that reached them, so the definition and the demand ran on one target.

03 / HOW WE TURNED IT INTO A SALES MOTION · THE PEOPLE PART

The ICP defined the person. Reps still had to run discovery like they knew them.

An ICP built as a person, a day, and a problem is wasted if the rep runs a generic discovery call. The leadership half was training the team to walk in already understanding the buyer the ICP describes.

THE FRAMEWORK BEHIND IT

SPRINT Discovery, grounded in the ICP

Situation, Problem, Risks, Impact, Next steps, Timeline, tuned to the person the ICP names. The rep opens knowing the day and the problem, so discovery confirms and quantifies instead of starting from zero.

01
Turned the ICP into discovery questions
The person and the problem in the ICP became the exact questions reps asked, so the call started informed.
02
Coached reps to confirm, not interrogate
Walking in informed changes the call from a checklist into a conversation the buyer wants to have.
03
Fed real discovery back into the ICP
What reps heard on calls sharpened the ICP, so it stayed a living document instead of a slide.

Run discovery on the ICP →

04 / WHAT I LEARNED

Not proof. Just what the build taught me.

01
A filter is not an ICP
A size band, a vertical, and a tech tag make a targeting list. That list is why the pipeline fills with accounts that look right and never buy.
02
You cannot guess an ICP in a workshop
The real one is derived from the people who already bought, in the exact words they use for the problem.
03
The ICP is a job, not a title band
248 converter titles pointed at one motion. Different titles, one job to be done underneath.
04
Reconcile the inputs or the ICP is opinion
Merging interviews, G2, closed-won, and product usage against a real data cohort is what lets the ICP survive scrutiny instead of being one loud opinion.

05 / THE WORKFLOW

The runnable version. Copy it into your stack.

GongThe call transcripts, for the words buyers actually use.

G2The reviews mined for voice of the customer, praise and complaint.

SalesforceTwelve months of closed-won, backtested for the pattern that repeats.

AmplitudeHow the won accounts actually use the product.

ClaudeReconciles the inputs into one profile and the personas.

[ SALES OPERATOR ]

The ICP Playbook

Heath Barnett · Sales Operator

G / Groundthe problem that started all of it, with the receipt and the cost

ICP

A filter is not an ICP.

50-500 employees, B2B SaaS, uses a CRM is a targeting list. It is why the pipeline is full of accounts that look right and never buy.

Signal

You cannot guess it in a workshop.

A real ICP is derived from the people who already bought, in the exact words they use for the problem.

A / Assignthe build, block by block, and who owns each one. Open a step to see it run.

HUMAN + AI, IN THE LOOP

The operator talks to real buyers and writes the personas; AI mines the language, backtests the wins, and reconciles the profile.

Go to the source

The operator01

Interview the people who bought

Details

Real customers, not a panel. Why they came, why they bought, why they stayed, in their words.

Human

G202

Mine G2 and the calls for their words

Details

Voice of the customer: what they praise, what they complain about, the job they hired you for.

Also reads

Find the pattern

Salesforce03

Backtest 12 months of closed-won

Details

The shape that recurs, not the logos you wish you had.

Amplitude04

Read how won accounts use the product

Details

The technographic and usage shape of a real win.

Reconcile and write

Claude05

Reconcile the inputs into one profile

Details

Merge and reconcile the divergences against a real cohort, so the ICP survives scrutiny.

The operator06

Write the personas: the person, not the size band

Details

Their day, their environment, the problem you remove, and the hard disqualifiers.

Human

N / Normalizethe motion that made it stick

A workflow without a motion is dead. This is how it became the way the team works.

Turned the ICP into discovery questions

The person and the problem in the ICP became the exact questions reps asked, so the call started informed.

Coached reps to confirm, not interrogate

Walking in informed changes the call from a checklist into a conversation the buyer wants to have.

Fed real discovery back into the ICP

What reps heard on calls sharpened the ICP, so it stayed a living document instead of a slide.

T / Tie backthe result it drove, and how they know

RESULT · 01

n=394 converters

profiled by size, role, geo, and CRM.

RESULT · 02

6.9 vs 13.9 days

time to buy, core ICP vs the next tier.

RESULT · 03

One ICP, two motions

same person and job underneath.

OUTPUT · 01

Voice-of-customer language

The words buyers actually use.

OUTPUT · 02

The closed-won pattern

The shape that repeats.

OUTPUT · 03

Personas with disqualifiers

The person, not a headcount band.

The Sales Operator

06 / HOW YOU DO IT TOO

Here is what I built. Here is how you build it.

The whole thing installs as one plugin. Or grab the skills a la carte. Everything on this page, runnable, on what you paste today.

✓ EVAL PASS · 4/4

THE PLAYBOOK · ONE INSTALL

Nail your ICP, then act on it

5 skills chained into one runnable play. Installs as a single plugin, no copy-pasting each skill. It runs on what you paste; connect your stack to go live.

SEE THE FULL PLAYBOOK →

SEE IT RUN IN CLAUDE

EXAMPLE CHATMaya, VP of Sales, running the loop in one sitting

Our ICP slide says mid-market SaaS, 200 to 2000 employees. That is basically everyone. I want to know who actually buys and act on it.

Step 1· Customer Research

Let's start from real buyers, not the slide. I pulled your 12 most recent won accounts and set up interview questions to get the language they used and the trigger that made them buy now instead of later.

Receipt

Interview guide plus 12 target customers, and 3 recurring triggers already showing up in your notes.

Good. Now what do those won deals have in common that we keep missing?

Step 2· Pattern Analyst

I read your closed-won, closed-lost, and churned accounts together and looked for the shared signals. The wins cluster tighter than your slide does, and two loud traits are ones you are not filtering on today.

Receipt

Wins share 4 traits; the strongest is a recent RevOps hire, present in 9 of 12 wins.

So take a specific account. Are they actually a fit or do they just look like one?

Step 3· Icp Analyst

I scored that account on the traits that predict wins, not headcount band. Firmographics, hiring intent, and stack all rolled into one number with the reasoning shown, so you can see why it landed where it did.

Receipt

Account scored 78 of 100, dragged down by no RevOps hire and a competitor already in the stack.

Do that for the whole list so I know where to point the team.

Step 4· Icp Scoring

I stack-ranked all 640 accounts with a reason on every row and kept it fair across channels, so a blank field never scores as a zero. You get tiers, not a wall of numbers.

Receipt

640 accounts ranked; 84 land in Tier 1, and each row carries a one-line why.

Last thing. Where do these actually convert best so we do not waste the good accounts on the wrong channel?

Step 5· Strike Zone Analyst

I checked how your Tier 1 accounts convert by channel and found where the funnel leaks. Outbound is burning your best accounts at the demo stage, while warm intro closes them at more than double the rate.

Receipt

Tier 1 via warm intro closes at 31% vs 12% outbound, so route the top 84 to intro-first.

THE SKILLS IT USES

Get a ranked slate of customers to interview.

icp-interview-finder

Find what your wins, losses, and churn share.

pattern-analyst

Turn raw account notes into a one-page brief, every time.

account-brief-prompt

THE OTHER HALF · LEAD THE TEAM

Run discovery on the ICP→

I write up one of these a week. Free, receipts only.