How I Check Product-Market Fit, Churn and ICP Gaps Before Scaling Paid Demand Generation

When a founder asks me how much to put into paid campaigns next quarter, I ask for two numbers first: how many customers left in the last twelve months, and which segment they came from. Paid spend multiplies whatever the funnel already does. If one segment churns twice as fast as the other, a bigger ad budget buys more of the customers you will lose.

A blended scorecard hides this. A company can show a healthy LTV to CAC ratio and a respectable churn rate while one segment carries most of the losses. The averages look like a green light, and the board approves a paid budget aimed at the whole market.

Paid is also not the only way demand grows. At FieldAssist, where I was CMO from April 2023 to September 2024, organic website traffic grew 30% with zero paid activity. I mention it only to make one point: paid budget is a choice you make once the base holds, not a default line in the plan.

So the order in this guide is fixed. Score the fit, find out why people leave, see who they are, then decide where paid money goes.

Each tool answers one question, and each answer sets the input for the next, as in the table alongside.

The scorecard comes first because it tells you which number is weak and which data is missing. It does not tell you why. The retention playbook takes the churn reasons you already know and gives each one a signal, a first step and a draft message, so the customer success team can act this month. The churn interview kit is for the reasons you only half know: it gives you the questions to ask the people who left.

The ICP gap analysis comes last because it needs what the first three produce: a churn rate you trust, and a view of which kind of customer leaves. It compares the base you have with the base you want and names the gaps. That comparison is where the paid demand decision gets made.

I keep one rule across the whole workflow, and it is my own working rule, not a tool's: no paid budget beyond a capped measurement test goes to a segment until its own churn and its own paid-only payback are known. Blended numbers are for the board slide, not for the media plan.

Step 1. Score product-market fit

Input to the CRAFT GTM scorecard: Shelfwalk, a field sales app for consumer goods brands and their distributors; target market enterprise SaaS; 1 to 2 years in market; "MRR: $60K, Churn: 1.5%, NPS: 44, CAC: $9000, LTV: $72000, Retention: 95%".

What came back: "Emerging PMF" at 7.5 out of 10, "Based on 4 of 5 dimensions", with the line "Good foundation but gaps to address before aggressive scaling." It read the sector as vertical SaaS, FMCG retail execution, and listed what that buyer measures: secondary sales, productive calls, lines per call, outlet coverage, strike rate, order fill rate and distributor claim settlement time. Its two priority actions were to read and group the NPS detractors' comments, and to define the first action that shows a customer got value.

How to read it: activation shows 0/10 and "MISSING" in its row, but it is left out of the average, which is why four scores of 7, 5, 9 and 9 give 7.5. Treat the missing row as the first task, not as a low score. Every benchmark is labelled as an example. When I chose vertical SaaS instead of enterprise SaaS, the scores stayed the same, and the tool said it has no example ranges of its own for vertical SaaS, so it uses its enterprise SaaS ranges. Put your own ranges next to them.

Then check the inputs against each other, because the tool scores each one separately. A churn of 1.5% a month keeps about 83% of customers over a year (0.985 to the twelfth power). That cannot be the same measure as the 95% retention scored "EXCELLENT", so the 95% must be revenue retention or a different period. Write the definition next to each number before a board sees two green rows that describe different things.

The LTV holds up on my own arithmetic: with an example ACV of $18,000 ($1,500 a month) and an example gross margin of 72%, $1,080 of margin a month divided by 1.5% churn is $72,000. CAC payback is $9,000 over $1,080, about 8.3 months. But the $9,000 CAC is blended. If paid-only CAC came in at $24,000, an illustration of mine, the ratio would fall to 3.0x, the tool's healthy mark, and payback would stretch to about 22 months. That is the number to find before scaling paid, and the scorecard cannot find it for you.

Step 2. Build a playbook for each churn reason

Input to the retention playbook: segment "Regional consumer goods brands", SaaS subscription, churn "1.5% monthly", churn reasons "low field rep adoption, distributor data not synced", a customer success team of 1 to 3.

What came back: it turned 1.5% a month into "about 18% of customers lost in a year (the monthly rate times twelve, not compounded)" and said the rate reads low against its own example thresholds, which it calls illustrations. Each reason got a signal to watch, a first step, an owner and a draft email. For low rep adoption the first step is "An onboarding reset, a use case discovery call and one quick win with a date." For the sync problem it is personal outreach and a root cause review within two days.

How to read it: the first run had no product name and no data signals, so the drafts said "our service" and every health score signal was marked "to add". The tool listed four inputs that would sharpen it, and I filled two of the four: the product, with a few words on what it sells, and the data signals. After that, only current interventions was left on its list. In the second run it named Shelfwalk in the title, mapped "rep logins by region" to login frequency and "support tickets" to tickets, and added a step to the sync reason: "Name the distributor systems involved and agree how stock and orders flow back before the rollout." It also said two of my signals were not matched: "orders captured per rep per day" and "distributor sync errors". Expect the draft emails to keep saying "our service" and "Your account team" even with the product given: the name reaches the title, not the drafts.

Those two unmatched signals are the two churn reasons, so they matter most. My own fix, not the tool's: add them as their own lines in the health score and take the weight from billing health, which tells a small team little in a monthly subscription. Rewrite the draft emails around the data too. "Your reps in two regions captured fewer orders in the app this month than last" opens a better call than a general check-in.

Put the churn in counts as well. At $1,500 a month, $60,000 of MRR is about 40 brands, and 1.5% a month loses about seven of them a year. Seven brands is a small enough number to interview every one.

Step 3. Interview the brands that left

Input to the customer interview kit: churn interview, "Field sales app for consumer goods brands and their distributors", Head of Sales Operations, enterprise software, moderate complexity.

What came back: a guide in the sector's own language. The industry opener asks "How do you measure secondary sales today, and how much does it vary?" The technical questions include "What happens to an order captured in an outlet with no network?" and "How do distributors receive and confirm orders today?", and the sector list asks "Which distributor systems (DMS) and ERP must the sales app work with?" It also names who usually decides: the National Sales Head or Managing Director signs, the Head of Sales Operations champions, and distributors hold the stock and order data.

How to read it: the core questions (primary reason, secondary reasons, what they switched to) work for any product; the sector and technical questions are the ones that will separate a rep adoption problem from a distributor data problem. I did not give it hypotheses, and next time I would, because the kit keeps them whole. For Shelfwalk they are the two churn reasons, plus a third worth testing: brands whose distributors run a DMS stay longer than brands whose distributors keep orders in spreadsheets. Rewrite the generic opening lines in your own voice before the first call.

Use the synthesis table as the tool says, counting "interviews that said it, out of the interviews you ran". Then add one column of my own: the segment of each brand that left, by rep count and by distributor system. That column is the input for the next step.

Step 4. Compare the base you have with the base you want

Input to the ICP Intelligence gap analysis: current customers "Regional consumer goods brands, owner-run sales, distributors on spreadsheets"; ideal profile "Packaged food and personal care brands, a National Sales Head who signs and a Head of Sales Operations who champions, distributors on a DMS"; current and target metrics as in the table alongside, all example inputs.

Churn needs one decision before you enter it. The retention playbook gave 18% a year, "the monthly rate times twelve, not compounded". Compounded, 1.5% a month keeps about 83.4% of customers, so about 16.6% are lost. I entered 18%, so that the two tools carry the same number and so that the gap errs on the side of more loss, not less. Better still, replace both with your real twelve-month logo churn: customers lost in the last twelve months divided by customers at the start.

What came back: the largest gap is ACV, "+100% needed", with churn next at "-56% needed". It named packaged food and personal care brands as "Named in your ideal profile but not in your current base" and told me to "Check whether you have won, lost or never pursued deals there." It did not throw out the current segment: "only move it out once win rate, ACV and churn show it is weaker." Its week 1 task is to pull win rate, ACV and churn for each segment, and its week 3 task lists the objections to answer: "reps will not use another app; our DMS vendor bundles a sales app; distributor data is not in sync."

How to read it: write each profile as a short list of phrases separated by commas. I ran it three times. A long sentence and a version with labels such as "Industry:" both produced garbled segment names, for example "Size: reps". The third run was the most usable, and it still garbled three lines. It listed "distributors on spreadsheets" as named in both profiles, most likely because both use the word distributors. It described the ideal customers as "a National who signs who champions", and it named the buyer "Head of Sales Operations and Sales Head", a join of my two roles. Check each line of the output against your own two texts before you copy it into the ICP. The sales cycle row is right in direction: a 90-day target is longer than today's 45, so there is no gap to close, but a longer cycle stretches paid payback, so put it into the payback sum.

The decision on paid demand comes from the week 1 split, not from the averages. The second table alongside is my own decision rule. Packaged food and personal care brands are not in Shelfwalk's current base, so the last row applies: win a few of them first through outbound and distributor partners, whose distributors run a DMS, and prove the tool's sector proof point on them: secondary sales or productive calls in the pilot region against a comparable region without the product. Only then run a capped paid test on that segment, sized to measure paid-only CAC, with a stop rule on payback. Regional brands stay on referrals, partners and customer success until their churn falls.

1. Write the definition next to every fit metric (customers or revenue, monthly or annual), then run the scorecard and fill the data gaps it lists.
2. Run the retention playbook with your product name and the data signals you can track, and add any signal it could not match as its own line in the health score.
3. Book a churn interview with every customer who left in the last twelve months, using the kit with your hypotheses added, and tag each one by segment.
4. Run the ICP gap analysis with short comma-separated profiles, then do its week 1 task: win rate, ACV and churn for each segment.
5. Work out paid-only CAC and payback for the segment you want before any paid budget beyond a capped test is approved, and set the stop rule in the same meeting.

1. Approving paid budget on a blended LTV to CAC ratio. Work out the ratio with paid-only CAC for the segment the campaign will reach.
2. Scoring two retention numbers that measure different things. Monthly logo churn and annual revenue retention can both look green while telling different stories.
3. Reading a missing metric as a low score, or a high average as fit. The scorecard leaves missing data out of the average, so fill the gap before you quote the score.
4. Interviewing churned customers without tagging their segment. Without that tag you learn why people leave, but not who leaves.
5. Cutting the current segment because it is not the ideal one. The gap analysis says to move it out only once the split shows it is weaker; it may still fund the business while you test the new one.

• How I Stress-Test a B2B GTM Plan Before the Board Sees It, With a Pre-Mortem and a Bottom-Up Count
• How I Pick the Lead B2B Go-To-Market Motion With EPIC, Then Size and Anchor It
• When High Adoption Hides Churn Risk: A GUARD Health Check, Save Plan, Expansion Case and QBR on One Book of Accounts
• All 15 guides