The ICP Sharpening: why do you need it?
- Bohdan Lytvyn
- April 20, 2026
RECOMMENDED READING FOR A BETTER CONTEXT:
KEY TERMS:
What is ICP before GTM? — unanswered question
Let’s ask ourselves a question: is ICP framing retroactive or proactive? Or both?
If we “start with data”, we must take the existing customer profiles by criteria of effectiveness, including:
- Sales cycle speed
- Revenue
- Churn and
- NPS
This may potentially produce several combinations that, nevertheless, revolve around the “Buyer champion” archetype. It means we retroactively sculpt the ideal customer profile from the actual customers that we like most.
Effectiveness threshold works like the first and principal filter — that works like a rod — that is applied to different patterns:
- company size,
- maturity,
- category (industry),
- behavioural patterns, etc.
This is the focus of the data analysis effort that tends to unfold new and promising combinations of patterns that meet the threshold requirements. Further refinement is possible through negative thresholds.
Good, what’s the alternative?
If we “start with a problem” — we will have no criteria of effectiveness as above. So, no verifiable scoring system can be produced — at least not until the first batch of paying customers.
The question of where to start building the ICP is a false dilemma. The proposition “Build the ICP from your best customers, not all customers” neglects the GTM reality where ICP must already be in place before active market moves. Starting with a problem is the only option for the Go-To-Market stage.
“There is no such thing as an ICP before GTM — only a hypothesis of accessibility.”
This brings us to the idea of a need for continuous ICP refinement.
If you look at the GTM roadmap, as I envisage it, The ICP appears twice: on the 3rd stage of “ICP definition” and on the 13th, the final stage, of the ICP refinement. The difference is immense: at the pre-GTM stage ICP is hypothesised and is close to the buyer Persona voice (in fact, it only starts its decomposition therefrom), while in the ICP sharpening stage the ICP is too real and is subjected to noise reduction practice.
ICP Entropy: Why your Ideal Customer Profile decays over time
The necessity of ICP refinement or sharpening seems intuitively comprehensible: it is a loop of ICP Building and Validation.
We win clients and simultaneously engage in ICP building using:
- Win / Loss internal analysis
- Customer interviews
- Patterns and
- Usage signals.
ICP Validation comes from:
- Comparing win rates
- Testing with Sales
- Refinement of criteria applicable to patterns definition and usage signals
It is the natural extension of ICP framing — come from problem-centered hypotheses around who the ideal clients are, accumulate the data and try to identify the underlying patterns that will help to reformulate / reposition the hypotheses themselves.
In reality, the ICP sharpening is often caused by the entropy informed by the growth motion itself: PLG, SLG or CLG. The mechanisms of ICP entropy by the growth motion looks as follows.
PLG entropy — silent drift (unseen)
Mechanism:
- ICP inferred from who signs up
- Not from who understands value
Result:
- False positives accumulate
- ICP shifts toward curiosity-driven users, not buyers
Detection signal
The earliest observable signal of silent drift is a widening gap between activation metrics and conversion metrics that opens gradually and is initially masked by top-of-funnel growth. Signups are increasing, activations look healthy, but the ratio of activated users who convert to paid is declining quarter on quarter. The key diagnostic is cohort conversion rate by acquisition source:
“if users acquired through content or SEO are converting at half the rate of users acquired through direct referral or word of mouth, the product is attracting curiosity-driven users who match the surface profile of the ICP but not its underlying motivation.”
These are the false positives accumulating silently in the pipeline.
A second detection signal is support ticket taxonomy. As silent drift progresses, support requests start clustering around basic comprehension questions — "what does this feature do," "how does this connect to my workflow" — rather than advanced usage questions. This signals that the user base has drifted toward people who are still trying to understand the product rather than people who have already validated its relevance to their problem.
A third signal, detectable only in products with usage data, is feature adoption breadth versus depth divergence. True ICP customers adopt a narrow set of features deeply and repeatedly — they have found the product's core value and are extracting it. Drift-contaminated users adopt many features shallowly — they are exploring without anchoring, which is the behavioral signature of someone who signed up out of curiosity rather than recognized need.
SLG entropy — deliberate widening (pipeline pressure)
Mechanism:
- Sales expands definition of “qualified”
- Near-ICP becomes acceptable
Result:
- ICP boundary dissolves
- Organization loses discrimination capability
We may simply say:
“In SLG, ICP gets negotiated away.”
Detection signal
The earliest observable signal of deliberate widening is, statistically speaking, a divergence of distribution of sales cycle length — not average cycle length increasing, but the variance widening. A healthy SLG pipeline shows relatively consistent cycle lengths within deal size bands. A widening-contaminated pipeline shows a bimodal distribution:
- a cluster of fast-closing deals (true ICP, where the sales motion matches the buyer's decision process) and
- a growing tail of slow-moving deals (near-ICP, where the sales team is investing additional effort to compensate for poor initial fit).
The tail is where the waste is accumulating, but it's invisible in average metrics — the fast deals keep the average looking acceptable while the tail consumes disproportionate sales capacity.
A second signal is qualification criteria language in CRM records. The language sales teams use to justify moving deals through pipeline stages becomes increasingly hedged: "they're not exactly our typical buyer but," "the use case is slightly different however," "the budget is below our usual range though."
These qualifications are the individual-level record of ICP boundary negotiation. Aggregated across the pipeline, they reveal the degree to which the ICP has been quietly redefined to accommodate pipeline pressure.
A third signal is win rate by deal age: the probability of closing a deal that has been in the pipeline longer than the median cycle length drops sharply in a healthy SLG pipeline and drops only slightly in a widening-contaminated pipeline. The near-ICP deals that should be lost early are instead being kept alive by sales investment, consuming capacity that would otherwise be directed toward new true-ICP prospects.
CLG enthropy — elite distortion (top customers dominate)
Mechanism:
- Feedback weighted by engagement, not representativeness
Result:
- ICP becomes:
- narrower
- more sophisticated
- less scalable
So, ICP entropy that’s embedded in your growth strategy calls for continuous ICP sharpening. This brings us to the following idea:
“ICP is unstable equilibrium constantly destabilized by your GTM motion.”I can put it in a different way:
I can put it in a different way:
Detection signal
The earliest observable signal of elite distortion is roadmap complexity increasing while new customer onboarding time also increases — a combination that should not co-occur in a healthy product. Roadmap complexity should increase as the product matures, but onboarding time should decrease as the product becomes more refined and its value more immediately legible. When both increase simultaneously, it signals that the product is being optimized for sophisticated existing users at the expense of new user accessibility — the hallmark of elite distortion.
A second detection signal is support burden distribution. In an undistorted CLG environment, support requests are distributed across the customer base roughly proportionally to account size or usage intensity. In an elite-distorted environment, the advisory board and high-engagement customers generate a disproportionately low share of support requests — because the product has been built around their needs — while median customers generate a disproportionately high share. This distribution inversion is the observable signature of ICP drift toward the sophisticated minority.
A third signal is churn source analysis. Elite distortion produces a specific churn pattern: low churn among advisory board and high-engagement customers (the product fits them precisely) and elevated churn among customers acquired in the six to twelve months after the CLG motion became dominant (the product has drifted away from what they were sold). The churn appears in cohorts rather than uniformly, which makes it easy to misattribute to onboarding quality variations rather than recognizing it as the delayed consequence of ICP drift.
Why growth systems break and how to fix them?
Growth rarely fails because of lack of effort.
It fails when value becomes distorted across the system and waste accumulates unnoticed.
If you want to discuss your GTM or growth strategy, let's chat.
About the author
GTM strategy consultant, author of the Go-To-Market FOMO newsletter with 17 years experience in Growth Systems Design.
Bohdan Lytvyn
"WASTELESS GROWTH" BOOK AUTHOR