How to Figure Out Who Actually Buys From You
Who is your actual customer - and how do you find out?
Most companies answer this question with a persona doc that hasn’t been updated since the seed round. In the Growth Recon framework, the Research stage exists to replace that fiction with evidence. ICP mapping is the first move: pull revenue data, segment it by who actually pays and stays, and build your targeting around what the numbers say - not what the pitch deck hopes.
This is the tactical playbook for doing that work. Not theory. Not a template you fill in during an offsite. A step-by-step process for mapping who converts, why they convert, and where to find more of them.
Step 1: Export every closed deal from the last 18 months
Open your CRM. Pull every closed-won deal from the last 18 months. Not a summary. Not a dashboard. The raw deal data - company name, deal size, close date, industry, company size, source channel, sales cycle length, and current status (active, churned, expanded).
Eighteen months gives you enough volume to spot patterns and enough recency to reflect your current product. If you only look at the last quarter, sample size kills you. If you go back three years, you’re studying a different product.
What you’re looking for: the full population of people who said yes. Not the logos on your website. Not the deals your VP of Sales talks about at all-hands. Every deal, including the $3K annual contract from a 12-person company that nobody brags about but that renewed twice without a single support ticket.
If your CRM data is incomplete - and it probably is - supplement with payment records. Stripe, QuickBooks, whatever processes invoices. Revenue data doesn’t lie the way CRM fields do, because someone actually had to pay.
Step 2: Segment by revenue behavior, not demographics
The instinct is to segment by industry or company size. Resist it. Those are descriptors, not predictors. Instead, segment by the metrics that determine whether a customer is actually good for your business:
Lifetime value at 12 months. Not projected LTV - actual realized revenue from customers who’ve been around that long. A customer who pays $50K/year but churns at month six is worth less than a customer who pays $18K/year and is still active at month 30.
Customer acquisition cost. What did it cost to close each deal? Include sales time, not just marketing spend. A segment with a $40K LTV:CAC ratio of 5:1 is a fundamentally different business than a segment with a ratio of 1.8:1 - even if they’re in the same industry.
Time to close. How long from first touch to signed contract? Segments that close in 14 days require a different funnel than segments that take 90 days. You can’t run the same nurture sequence for both and expect either to work.
Churn rate by cohort. Group customers by the month they signed and track how many are still active. Some segments look fantastic at close but bleed out by month eight. Others start slow and compound. If you’re not tracking this per segment, you’re averaging away the signal.
Expansion revenue. Which customers buy more over time? Upsells, cross-sells, seat expansion. A segment that grows 30% year-over-year after initial purchase is worth more acquisition investment than one that flatlines.
Now map the demographics on top: industry, company size, buyer title, geography. The difference is sequence. Demographics describe the segments you found through revenue analysis. They don’t define them.
Step 3: Identify your top three segments
After you segment by revenue behavior, three to five clusters will emerge. Rank them by a combined score of LTV, LTV:CAC ratio, and expansion potential. Your top three are your ICP tiers.
Here’s what this looks like in practice:
Segment A - Bootstrapped B2B SaaS, 15–80 employees. Average deal: $14K/year. LTV at 12 months: $16K (low churn, occasional upsell). CAC: $2,800. LTV:CAC: 5.7:1. Sales cycle: 11 days. Source: 78% inbound (blog, organic search). They find you, they evaluate fast, they stay.
Segment B - VC-backed SaaS, 100–300 employees. Average deal: $42K/year. LTV at 12 months: $38K (higher churn, some downgrades). CAC: $14,000. LTV:CAC: 2.7:1. Sales cycle: 47 days. Source: 55% outbound (SDR). Bigger checks but harder to land and harder to keep.
Segment C - Mid-market non-tech, 200–1,000 employees. Average deal: $28K/year. LTV at 12 months: $31K. CAC: $9,500. LTV:CAC: 3.3:1. Sales cycle: 62 days. Source: 40% partner referral, 35% events. Decent economics but operationally heavy - long procurement cycles, more stakeholders, custom security reviews.
Your pitch deck probably leads with Segment B because those logos look good on the website. Your revenue data says Segment A is the engine. That gap between narrative and reality is exactly what ICP mapping exists to expose.
Step 4: Map the six dimensions for each segment
For each of your top segments, fill in six dimensions. No empty cells. An empty cell means you haven’t done enough work.
1. Demographics and firmographics
Industry, company size (revenue and headcount), geography, growth stage, tech stack. This is the targeting layer - what you use to build audiences in LinkedIn, filter in your CRM, and brief your SDR team. Be specific. “Mid-market SaaS” is not a firmographic. “B2B SaaS, $2M–$15M ARR, 20–120 employees, Series A or bootstrapped-profitable, using HubSpot or Salesforce” is a firmographic you can actually target.
2. The problem they’re solving (in their words)
Not the use case on your product page. The specific, messy, real problem that triggered the buying process. You get this from sales call recordings, specifically the first three minutes of discovery calls before the rep starts steering.
“We were spending 12 hours a week manually pulling data from three platforms to build a report nobody reads” is a problem statement. “We need better analytics” is a category. You need the former. The language matters because it feeds directly into your Language Audit - and from there into every piece of copy, every ad, every landing page.
3. Buying trigger
The event that made them start looking. Not the general pain - the specific moment. Common triggers: new executive joins and audits existing tools, a competitor launches a feature that embarrasses them, a manual process breaks during a high-stakes moment, budget opens at fiscal year start, or a board meeting where someone asked a question nobody could answer.
Buying triggers tell you when to show up, not just who to target. If your best customers consistently start evaluating tools in Q1 after board reviews, your outbound campaign should start in December.
4. Decision process
Who’s involved? Who has veto power? How long does it take? Map every stakeholder:
- Champion - the person who found you and is pushing internally. Usually a director or senior manager who owns the problem day-to-day.
- Economic buyer - the person who signs the check. Often a VP or C-level who won’t evaluate your product but will kill the deal if the business case doesn’t land.
- Technical evaluator - the person who runs a trial, checks integrations, and raises security concerns.
- Blocker - the person who benefits from the status quo or has budget competing against your deal.
If you don’t know the blocker, you don’t know the decision process. Most deals don’t die because the champion lost interest. They die because a blocker raised an objection the champion couldn’t answer.
5. Price sensitivity and budget authority
What are they comparing your price against? Other tools in your category, an internal build, a consultant, or doing nothing? “Doing nothing” is your most common competitor and the hardest to price against because the cost of inaction is invisible until you make it visible.
Document the approval threshold. Many companies have a spending limit below which a director can sign without VP approval - often $15K–$25K/year. If your deal lands above that line, you just added a stakeholder to the decision process and three weeks to the sales cycle. Pricing strategy and ICP mapping are not separate activities.
6. Objections and dealbreakers
What kills deals in this segment? What slows them down? Separate the two - they require different responses.
Dealbreakers are binary. “We need SOC 2 compliance” - you either have it or you don’t. “We need native Salesforce integration” - same. You can’t overcome these with better sales technique. They’re filter criteria that determine whether a segment is viable at all.
Objections are concerns that can be addressed. “We tried a similar tool and it didn’t stick” is an objection - you can counter with an implementation plan. “Your competitor is cheaper” is an objection - you can counter with total cost of ownership. Knowing the top three objections per segment lets your sales team prepare instead of improvise.
Step 5: Validate with primary research
Data tells you what happened. It doesn’t tell you why. For each of your top three segments, do primary research:
Win interviews (3–5 per segment). Talk to customers who closed in the last six months. Ask: what triggered the search, what alternatives they considered, what almost stopped them, and who else was involved in the decision. Don’t ask “Why did you choose us?” - that invites flattery. Ask “What would have happened if you hadn’t bought anything?” - that reveals the real stakes.
Lost-deal interviews (3–5 per segment). This is where most teams skip, and it’s the most valuable data. Talk to prospects who evaluated you and chose something else - or chose to do nothing. Ask: what were you trying to solve, where did we fall short, and what did you end up doing instead? The answers will either validate your ICP boundaries or redraw them.
Churned customer interviews (2–3 per segment). People who bought and left. Ask: what changed between when you signed and when you cancelled? Sometimes the answer is “the champion left” - which tells you your segment has a keyholder risk. Sometimes the answer is “we never fully implemented it” - which tells you your onboarding is a bottleneck that differentially affects certain segments.
These interviews aren’t optional. Every team that skips them ends up with an ICP that looks right on paper and fails in execution because it’s missing the qualitative layer that data alone can’t provide.
Step 6: Build the ICP one-pager
Take everything above and compress it into a single page per segment. Not a 20-page persona document. One page that any marketer, SDR, or product manager can read in two minutes and immediately use.
Format it as a table:
| Dimension | Segment A | Segment B | Segment C |
|---|---|---|---|
| Firmographics | [specific] | [specific] | [specific] |
| Problem statement | [their words] | [their words] | [their words] |
| Buying trigger | [event] | [event] | [event] |
| Decision process | [stakeholders + timeline] | [stakeholders + timeline] | [stakeholders + timeline] |
| Price sensitivity | [threshold + comparisons] | [threshold + comparisons] | [threshold + comparisons] |
| Top 3 objections | [specific] | [specific] | [specific] |
| LTV:CAC | [number] | [number] | [number] |
| Recommended channel | [where they come from] | [where they come from] | [where they come from] |
Example - filled in from real data:
| Dimension | Segment A: Bootstrapped SaaS | Segment B: VC-backed SaaS | Segment C: Mid-market Services |
|---|---|---|---|
| Firmographics | B2B SaaS, $1M-$8M ARR, 15-80 employees, bootstrapped | B2B SaaS, $5M-$30M ARR, 100-300 employees, Series A-C | Professional services, $10M-$50M revenue, 200-1,000 employees |
| Problem statement | ”We can’t tell which marketing channels actually bring paying customers" | "Board wants a growth plan and we don’t have the data to build one" | "We’re spending on digital but can’t prove ROI to the partners” |
| Buying trigger | New marketing hire audits existing tools, finds gaps | Board meeting where pipeline question went unanswered | Lost a major client and realized acquisition was on autopilot |
| Decision process | Founder + marketing lead, 2 weeks, no procurement | VP Marketing champions, CFO approves, security review, 6-8 weeks | Managing partner sponsors, operations lead evaluates, 8-12 weeks |
| Price sensitivity | Under $15K/yr decided by director-level. Compare against freelancer or DIY | $30K-$80K/yr range, compare against agency retainer or in-house hire | Budget exists but approval requires committee. Compare against consultancy |
| Top 3 objections | ”Can we do this ourselves?”, “What if we outgrow it?”, “Too small for this" | "How does this integrate with Salesforce?”, “Timeline is tight”, “Last tool failed" | "Our industry is different”, “Partners won’t adopt new processes”, “Prove ROI first” |
| LTV:CAC | 5.7:1 | 2.7:1 | 3.3:1 |
| Recommended channel | Inbound (blog, organic search) | Outbound (SDR) + events | Partner referral + industry events |
This table goes into The Source Doc - the single artifact that everything else in the Research stage feeds into. It’s not a static document. You revisit it every quarter with fresh deal data and update the segments that shifted.
The mistakes that waste the most time
Mistake 1: Building the ICP from the sales team’s memory. Sales remembers the big logos and the painful losses. They don’t remember the 200 mid-market companies that closed in two weeks with zero drama - which might represent 60% of your revenue. Start with data, then validate with the team’s qualitative insights. Not the other way around.
Mistake 2: Treating all MQLs as equal. A form fill from Segment A and a form fill from Segment C are not the same lead. If your scoring model doesn’t weight by ICP tier, your SDRs are spending equal time on a 5.7:1 LTV:CAC prospect and a 1.8:1 LTV:CAC prospect. That’s not a lead quality problem. It’s an ICP implementation problem.
Mistake 3: Conflating “biggest deal size” with “best customer.” The $120K enterprise deal that took nine months to close, required three custom integrations, and churned after 14 months generated less profit than a $14K bootstrapped SaaS deal that closed in 11 days and renewed for four years running. Vanity metrics in ICP selection are just as dangerous as vanity metrics in reporting.
Mistake 4: Setting the ICP once and never updating it. Markets shift. Your product changes. New segments emerge that didn’t exist when you last did this exercise. Treat the ICP like an operating rhythm artifact - reviewed quarterly with fresh data, not an annual strategic exercise that gathers dust.
Mistake 5: Ignoring the sacred cow. Every company has a customer segment that is politically protected - the founder’s first market, the CEO’s favorite logo category, the segment the board presentation highlights. If the data says that segment underperforms, you need to say so. Document the math. A 1.2:1 LTV:CAC ratio is a 1.2:1 LTV:CAC ratio regardless of who likes the narrative around it.
How to tell when your ICP is actually working
A good ICP changes behavior. If your SDR team is still prospecting the same way they were before the ICP exercise, the exercise failed - not because the research was wrong, but because it didn’t translate into operational changes. Here’s what “working” looks like:
- Marketing channel allocation shifts. If Segment A comes through content and Segment B comes through outbound, your budget split should reflect that. If it doesn’t, the ICP is a shelf document.
- Lead scoring updates. Sales-qualified lead definitions now include ICP tier. Segment A leads get routed faster. Segment C leads get nurtured differently - or disqualified.
- Win rates by segment are tracked and improving. You should see win rates climb in your priority segments within two quarters. If they don’t, either the ICP is wrong or sales execution hasn’t adapted.
- Content maps to segment problems. Each blog post, case study, and landing page should be tagged to a segment. If your content team can’t answer “who is this for?”, the ICP hasn’t been operationalized.
- ROAS improves on paid channels. When you target tighter, waste drops. Your cost per click might go up because you’re bidding on more specific terms, but your cost per qualified opportunity should drop.
Where this fits in RECON
ICP mapping is the opening move of the Research stage - and it’s the foundation everything else depends on. Without a data-backed map of who actually buys, every stage downstream inherits a targeting error that compounds.
Your Language Audit can’t work if you don’t know whose language to audit. Your data audit measures nothing meaningful if you’re tracking metrics for the wrong segments. Your Expose strategy reaches the wrong people with the wrong message. Your Convert stage optimizes a funnel for visitors who were never going to buy. And your Navigate stage builds operating rhythms around numbers that don’t reflect the business reality.
The ICP one-pager feeds directly into The Source Doc alongside the Language Audit, the Data Audit, and the Adversarial Assessment. Together, they form the evidence base that makes the rest of the RECON loop function. Skip it or rush it, and you’ll feel the drag in every stage that follows. Do it right, and every dollar, every campaign, every piece of content has a clear target with proven economics behind it.
The work isn’t glamorous. Exporting CRM data, segmenting spreadsheets, reading churn reports, sitting through interviews with customers who left - none of it makes for a compelling conference talk. But this is the work that separates teams who grow from teams who spend. And the gap between the two is almost always traceable back to whether anyone bothered to figure out who actually buys.