Audit Survivorship Bias in Your Ideal Customer Profile

If nine of twelve won customers operate several sites, multi-site businesses may look like the obvious ideal customer profile. The missing question is how many multi-site accounts were considered, pursued and unsuccessful. A winner-only review sees the survivors of your selection process, not the full evidence needed to assess the trait.

Reconstruct the account frame before rewriting targeting criteria. Include the organizations considered at the relevant decision point, their original traits, the route by which they entered the process and their eventual or unresolved outcomes. Keep exclusions visible without pretending excluded accounts received the same treatment as pursued opportunities.

Recover the original selection boundary

Choose a period, service and market with reasonably comparable definitions. Use the records available at entry, not today's enriched profiles. A company may add sites after becoming a customer; using its current footprint could make a later development look like a pre-purchase characteristic.

Record why accounts were excluded. A genuine capability mismatch differs from missing public information, unavailable contact routes or a seller's preference. Those decisions shape which companies could ever appear among the wins. If a segment was never considered, a winner-only profile cannot tell you whether it would have been suitable.

Statistics Canada's Statistics Canada: Non-probability sampling explains why selection mechanisms matter when generalizing from non-random samples. This audit is an original business review, not a probability survey or a method that automatically removes all selection bias.

Reconstruct a fictional forty-account frame

The following invented counts illustrate the problem. A consultancy originally considered forty accounts for one service. Twelve became wins, eight selected another solution, ten reached no purchase decision in the defined period and ten were excluded during research.

The multi-site trait was present at entry as follows:

• Wins: 9 multi-site and 3 single-site accounts.

• Other-solution outcomes: 6 multi-site and 2 single-site accounts.

• No-decision outcomes: 8 multi-site and 2 single-site accounts.

• Research exclusions: 2 multi-site and 8 single-site accounts.

Among the thirty pursued accounts, twenty-three were multi-site and seven were single-site. The nine multi-site wins therefore came from twenty-three pursued multi-site accounts, while the three single-site wins came from seven pursued single-site accounts. The descriptive proportions are approximately 39.1% and 42.9%, respectively.

Those small, non-random groups do not establish that single-site accounts are better. They do show why “nine of twelve wins were multi-site” is insufficient evidence that multi-site status improves the chance of winning. Multi-site accounts dominated the pursued group before the outcomes occurred.

Inspect the exclusions rather than adding them blindly

The ten research exclusions should remain outside a pursued-opportunity win-rate denominator if they never entered that process. But they matter to the targeting audit. In the fictional review, six single-site exclusions lacked a published role matching the seller's preferred title; two were genuine service mismatches.

That distinction suggests a question about the role-research method. Smaller organizations may assign the same responsibility under different titles. The review should test that possibility before treating absence of a preferred title as evidence that the business cannot use the service.

Do not retroactively count excluded accounts as lost deals. Preserve their actual state and inspect the rule that kept them out. The aim is to understand coverage and selection, not to manipulate one headline ratio.

Separate purchase success from delivery suitability

A won customer may still require unsupported customization or create a poor delivery fit. Include the relevant, authorized delivery evidence when judging whether a trait belongs in an ideal customer profile. A segment can be easy to sell to and difficult to serve responsibly.

Keep buyer-stated reasons distinct from seller explanations. Unknown loss reasons remain unknown. Do not fill them with a theory that makes the preferred segment appear stronger. Likewise, do not infer financial health or personal motives from a no-decision outcome.

Finish with one bounded change to test. In this example, review responsibility evidence for a small set of otherwise suitable single-site accounts rather than discarding the multi-site segment. Preserve the prior criteria and compare new cases prospectively. A stronger ICP emerges from examining who could have become a success, not merely describing those who already did.

Continue with Measure Sales Navigator Prospecting with a Small, Honest Funnel or Compare Two Target Segments with a Fair Sales Experiment.