The middle school version: Imagine you're selling lemonade, but you don't know who loves lemonade the most — is it kids at soccer games? Office workers at lunch? People at the beach? Surface helps us test different groups of people to figure out who our BEST customers are. We call these groups "ICPs" (Ideal Customer Profiles). Instead of guessing, we let the data tell us who loves our lemonade the most.

The ICP problem

Every startup has an ICP document. It's usually a one-page Google Doc written during a strategy session, based on a mix of founder intuition, early customer conversations, and competitive analysis. It says something like "B2B SaaS companies, 50–500 employees, Series A–C, US-based."

That document doesn't get updated. It sits in a shared drive while the sales team sells to whoever shows up. Six months later, someone asks "are we still targeting the right people?" and nobody has a data-driven answer.

We were that startup. Our initial ICP was based on pattern matching from our first 20 customers. It felt right. But feeling right and being right are different things — and we didn't know the difference until we started testing.

How Surface's ICP definition feature works

Surface lets us define multiple ICP hypotheses and measure each one against actual conversion data — not just lead volume, but the metrics that matter: conversion rate, deal size, time to close, and retention.

We set up ICP definitions as combinations of attributes:

Every lead that enters Surface gets automatically tagged with which ICP hypothesis they match (they can match multiple or none). This tagging happens at the moment of capture — based on form data and enrichment — and follows the lead all the way through to closed/won or closed/lost.

What we measured

After 90 days of tracking, we pulled the data:

The surprises

Agencies were our best ICP — and we almost ignored them. ICP C had the highest lead-to-meeting rate, fast close times, strong retention, and decent deal sizes. We'd been treating agency leads as secondary. The data showed they were our most efficient customer segment.

Why? Agencies use Surface across multiple client accounts. They understand the product faster because they live in marketing ops tools daily. They have shorter procurement cycles. And they expand — when one client campaign works, they roll Surface out to three more clients.

Growth startups had volume but not value. ICP B generated the most leads. The close times were fast. But deal sizes were small and retention was weak — these companies churned at nearly 3x the rate of mid-market SaaS. They were easy to close but expensive to serve and hard to keep.

Mid-market services was our weakest segment. Low volume, low conversion, long sales cycles. The product fit was there, but the buying process in professional services is slow and consensus-driven. We didn't kill this segment, but we stopped investing in it as a primary target.

What we changed

Based on the data, we restructured our GTM around two primary ICPs: mid-market SaaS (ICP A) and agencies (ICP C). Growth startups (ICP B) became a self-serve segment — we let them convert but don't invest sales time in them.

The practical changes: ad targeting shifted to company size 50–500 in SaaS and agencies. Form qualification branching gives these segments the premium path (more qualifying questions, direct AE routing, instant scheduling). Content strategy pivoted to address agency use cases — a segment we'd barely mentioned in previous content.

How to test your own ICPs

Step 1: Define 3–4 hypotheses. Each hypothesis should be a specific combination of company attributes — not just "B2B companies." Be specific enough that you could look at a lead and say "this matches ICP A but not ICP B."

Step 2: Tag every lead at capture. Use form data and enrichment to automatically classify which ICP each lead matches. This should happen at the moment of capture, not retroactively.

Step 3: Measure the full funnel, not just volume. Track each ICP through to revenue — conversion rate, deal size, close time, and retention. A high-volume ICP with low retention costs more than it generates.

Step 4: Reallocate based on data. Shift ad spend, sales time, and content toward the ICPs that perform best on the metrics that matter. Review quarterly — ICPs can shift as your product and market evolve.

Stop guessing who your best customers are. Define and test ICPs with Surface.