ICP & the invisible buyer.
Persona-by-job-title doesn't survive contact with 13-person buying groups and an 80% self-directed journey.
Your ICP is probably a job title and a firmographic filter. Modern B2B buying isn't. It's a 13-person committee, 80% self-directed before you ever see them, and the people who actually shape the deal rarely fill in forms. We rebuild ICP around the buying group, not the buyer.
Four symptoms we see, almost every time.
- 01
Sales says leads are bad; marketing says quota is too high
- 02
High MQL volume, low SQL conversion, no obvious pattern
- 03
Deals appear late-stage from accounts you never marketed to
- 04
Personas haven't been revisited since the last rebrand
Average stakeholders in a B2B buying group (Gartner)
Of the buying journey is now self-directed (Forrester)
Of buying time spent talking to any vendor
Three moves — in this order, every time.
- 01
Buying group model
Replace single-persona ICP with a 5–9 role buying group map per segment.
- 02
Invisible-buyer research
Interview lost-deal committees, not just champions, to surface the people who actually killed it.
- 03
Account fit scoring
Score accounts on group completeness, not lead count.
What this gap actually costs.
In cyber and enterprise technology, the person who signs is rarely the person who cares. A CISO owns the risk, a head of infrastructure owns the disruption, a procurement lead owns the paper, and a security architect quietly decides whether your product is credible. Target one of them and the other three can stop the deal without ever appearing in your CRM.
The buying group is also getting larger as budgets tighten. Every incremental reviewer adds a veto, and each veto has its own definition of risk. A single-persona ICP gives your team one message for a room that needs four, which is why campaigns that look well-targeted on paper produce pipeline that never progresses.
The consequence is expensive rather than obvious. Spend still lands on accounts that could buy, so cost-per-lead looks acceptable. What breaks is conversion downstream, where the missing stakeholder surfaces late, asks the question nobody prepared for, and turns a forecasted deal into a no-decision.
The version that works.
- 01
A written buying group map per segment: five to nine named roles, what each one fears, what each one needs to see, and who inside your team is responsible for reaching them.
- 02
An anti-ICP that is as specific as the ICP — the accounts and situations you will decline, and the reason, so sales can disqualify early without negotiating internally.
- 03
Account-level fit scoring that rewards group coverage. An account with four engaged roles is more valuable than an account with twelve contacts from the same team.
- 04
Evidence from lost deals, not just won ones. The committee that rejected you explains your ICP better than the champion who liked you.
Four ways good teams get this wrong.
Job-title targeting
Filtering by title and headcount produces a list, not an audience. It tells you nothing about who else has to agree, or what they need to believe first.
The champion echo chamber
Research done only with people who already like you confirms the story you already tell. The stakeholders who blocked you hold the information that would change it.
Personas as brand artefacts
Beautifully designed persona documents that no campaign, no sales call and no routing rule ever references. If it does not change a decision, it is decoration.
One ICP for every segment
Regulated financial services and high-growth software buy the same category for different reasons. One definition forces both into a story that fits neither.
- 01
Pull the last 12 months of closed-won and closed-lost, and count how many distinct roles appeared in each. The gap between the two numbers is usually the whole problem.
- 02
Interview three to five lost-deal committees, including the people who never spoke to sales.
- 03
Rebuild the ICP as a buying group model, then re-score the current pipeline against it to see how much of the forecast is single-threaded.
- 01Buying group blueprint per segment
- 02Updated ICP + anti-ICP definitions
- 03Account fit scorecard wired into CRM
Targeting that reflects who actually decides — and a sales motion that stops chasing the wrong human in the right account.
Book the diagnosticFrequently asked.
- How is a buying group model different from personas?
- Personas describe individuals in isolation. A buying group model describes the room: which roles must align, in what order, and what each one needs to see before the next can move. It changes targeting, content and routing rather than just tone of voice.
- We already have an ICP. Why revisit it?
- Most ICPs were written for a smaller, simpler buying committee and have not been tested against recent losses. The test is simple: can your ICP predict which deals will stall? If not, it is a description rather than a decision tool.
- How long does an ICP rebuild take?
- The diagnostic takes around two to three weeks, including lost-deal interviews. Wiring the resulting fit score into CRM and re-briefing campaigns typically runs alongside the first month of delivery.
- Does this mean we should stop measuring leads?
- No — it means leads stop being the primary unit. You keep lead-level data for operational reasons but make decisions on account and group coverage, which is the level at which the deal is actually won or lost.