How to Build a B2B ICP for Outbound (Without Spending $50k on Research)

What is a B2B ICP for outbound?

A B2B ICP (Ideal Customer Profile) is a written, account-level definition of the companies most likely to buy your product, stay, and expand. For outbound specifically, it functions as a scoring rubric: every account in your TAM gets ranked against the ICP, and the top-ranked accounts become the target list SDRs actually work.

An ICP is not a buyer persona, a market size, or a wish list. It is a finite, prioritised set of companies, usually expressed as:

• 4 firmographic filters (industry, size, geography, business model)
• 3 technographic filters (tools they use that signal fit or budget)
• 2 behavioural / intent filters (events that signal active buying)

According to Abmatic's 2026 ICP fit benchmarks, deals sourced from ICP-fit accounts close at 68% vs 22% for non-fit accounts, with sales cycles 20-30% shorter. That gap is the single biggest lever in outbound. Most teams skip it because the work is unglamorous.

How is an ICP different from a buyer persona and TAM?

An ICP describes the company you sell to. A buyer persona describes the person inside that company you sell through. TAM describes every company that could theoretically buy.

Most teams collapse these into one document and end up with something useless for everything. They are three different artefacts with three different jobs:

Concept	Unit	What it answers	Used for
TAM	The entire market	How big is the prize?	Fundraising, market sizing
ICP	Company / account	Which accounts should outbound target?	List building, account scoring
Buyer Persona	Individual person	What do we say inside the account?	Messaging, sequences

HubSpot's customer education team summarises it cleanly: ICP is company-focused and quantitative; buyer persona is individual-focused and qualitative; TAM is the universe both live inside. You need all three, but the ICP is the one that ships an outbound list this week.

How do you build a B2B ICP in 7 steps?

The 7-step build runs on data you already own (your CRM) plus three external tools (LinkedIn Sales Navigator, BuiltWith, Crunchbase). Budget 8-12 hours. The output is a tiered 5,000-account starter list ready for an SDR team.

Step 1: Pull your top 25 closed-won accounts

Open the CRM, filter to closed-won in the last 12 months, sort by ACV descending, take the top 25. Exclude any deal sourced from a personal network -- those skew the pattern.

For each account, capture: industry, employee count, revenue band, HQ country, funding stage, tech stack, source channel, deal cycle in days, ACV, and current MRR. Drop it into a Google Sheet, one row per logo. This is your raw material.

Step 2: Run a fit vs intent matrix on your 25

Plot each account on a 2x2: Fit (how well their firmographics match your product) on the X axis, Intent (how clearly they had an active need when they bought) on the Y axis.

The top-right quadrant -- high fit, high intent -- is your ICP. Usually 8-12 of your 25 land there. The bottom-left is who you should not target. The other two quadrants are the noise you need to filter out of outbound.

Step 3: Identify 4 firmographic signals

Look across your top-right quadrant accounts for shared firmographic traits. Pick the 4 most predictive:

• Industry / vertical (be specific: 'B2B SaaS' not 'Technology')
• Employee count band (e.g. 100-2,000)
• Revenue or funding stage (e.g. Series B-D, $10M-$100M ARR)
• Geography (country or region, not 'global')

Per Lenny Rachitsky's ICP framework, you want at least three narrowing characteristics that make the ICP feel 'almost comically narrow.' If yours describes 50%+ of your market, tighten it.

Step 4: Layer 3 technographic signals

Technographics reveal budget, sophistication, and competitive context. Pick 3 that consistently appear across your top accounts:

• Competitor tools (they already pay for a rival = budget is allocated)
• Adjacent / complementary tools (e.g. Salesforce + Outreach = mature outbound team)
• Infrastructure tools (e.g. Segment, Snowflake = data-driven org)

Pull this from BuiltWith, Wappalyzer, or HG Insights. A 2026 HG Insights analysis found that adding deep technographics to ICP scoring lifts predictive accuracy by 30-40% over firmographics alone.

Step 5: Add 2 behavioural / intent signals

Behavioural signals are time-bound events that move an account from 'fit' to 'fit right now.' Pick 2 you can actually source weekly:

• Funding rounds (the 6-month window post-Series B/C is the highest-intent period)
• Role hires (open SDR, RevOps, VP Sales jobs signal GTM spend)
• Leadership changes (new CRO/CMO typically rip-and-replaces the stack within 90 days)
• Product launches, M&A, expansion announcements

Source these from Crunchbase for funding, LinkedIn Sales Navigator's 'Posted jobs' filter for hires, and Google Alerts for press.

Step 6: Score every account in your TAM 0-100

Build a weighted scoring sheet:

• Firmographic fit: 40 points (10 per signal)
• Technographic fit: 30 points (10 per signal)
• Behavioural / intent: 30 points (15 per signal)

Apply the rubric to your TAM. Most teams pull TAM from a 20,000-account export out of LinkedIn Sales Navigator (it has 50+ filters that map to firmographics) or Apollo/ZoomInfo. Score every account, then sort descending.

Step 7: Output a 5,000-account starter list with tiers

Slice the scored list into three tiers:

• Tier 1 (score 80+): top ~500 accounts. AE-owned, full personalisation, 12-18 touches over 8 weeks.
• Tier 2 (60-79): next ~1,500-2,000 accounts. SDR-owned, light personalisation, 10-12 touches.
• Tier 3 (40-59): remaining ~2,500-3,000. Automated sequences, minimal personalisation, used to feed top of funnel.

Accounts scoring below 40 are suppressed. That is the entire output: one tiered Google Sheet your SDR team can work tomorrow morning.

What signals should a B2B ICP include?

A useful ICP includes 9 signals across 3 categories: 4 firmographic, 3 technographic, 2 behavioural. Fewer signals and the ICP is too broad to score with. More signals and you cannot consistently source data for every account.

Firmographic (the structural fit):

1. Industry / vertical (specific, not 'Technology')
2. Employee count band
3. Revenue or funding stage
4. Geography

Technographic (the budget and sophistication tell):

5. Competitive tools (rival product = allocated budget)
6. Complementary tools (signals mature stack)
7. Infrastructure / data tools

Behavioural (the timing signal):

8. Recent funding round (6-month window)
9. Active role hiring or leadership change

The Cleanlist 2026 ICP scoring guide recommends roughly 40/40/20 weighting (firmographics / intent / engagement) to avoid over-indexing on any one dimension. Pure firmographic ICPs qualify accounts but cannot prioritise them. Adding behavioural signals is what turns a list into a queue.

How granular should an ICP be?

Granular enough that it excludes 70-80% of your TAM. If your ICP describes more than half your addressable market, it is a market description, not an ICP. Outbound needs constraints to be useful.

The two patterns that work:

• Narrow + specific: 'Series B-D B2B SaaS, 100-2,000 employees, headquartered in US/UK/Germany, using Salesforce + Outreach, raised in the last 6 months OR hiring 3+ SDRs.' That returns roughly 3,000-5,000 accounts globally.
• Comically narrow (per Lenny Rachitsky): 'Companies with 20+ open remote roles on LinkedIn' for an employee onboarding platform. That is one criterion that does the work of five.

First Round's Emery Rosansky tells founders the ICP is 'never really done' -- it gets tightened or loosened every quarter based on closed-won data. Start narrow. You can always expand. You cannot un-spam the inbox of an account that was never going to buy.

How do you score accounts against your ICP?

Score every account on a 0-100 scale using a weighted rubric, then route by tier. The weighting that works for most outbound motions:

• 40 points -- firmographic fit (industry, size, geo, business model)
• 40 points -- intent / behavioural signals (funding, hiring, leadership change)
• 20 points -- engagement (site visits, content downloads, email opens)

For pure outbound (no engagement data yet), reweight to 50/50 firmographic and intent.

Routing rules:

Score	Tier	Routing
80-100	Tier 1	AE-owned, full personalisation
60-79	Tier 2	SDR sequences, light personalisation
40-59	Tier 3	Automated sequences
<40	Suppressed	Excluded from outbound

The impact is measurable. GrowthSpree's 2026 ICP scoring benchmarks show MQL-to-SQL conversion of 25-35% with ICP scoring vs 13% without, and cost per SQL dropping from $800-3,000 to $350-750.

What tools produce the cleanest ICP lists?

Three tools cover 80-90% of the work at roughly $150-300/month combined. You do not need ZoomInfo enterprise.

• LinkedIn Sales Navigator ($99/month): firmographic and people filtering. The 'Posted jobs,' 'Recent leadership changes,' and 'Company headcount growth' filters double as behavioural signals.
• BuiltWith (free for spot checks, $295/month for bulk): technographic enrichment. Best for tool stack and competitor displacement signals.
• Crunchbase Pro ($49/month): funding rounds, leadership changes, hiring intent. The cleanest source for the 6-month post-funding window.

For enrichment and orchestration, Clay pipes all three into one workflow with AI prompts per row. For data-source comparisons across the broader category, see our breakdown of Apollo vs ZoomInfo.

What you do not need on day one: 6sense, Demandbase, Bombora intent data. Those are layer-2 tools once your ICP is locked. Building the ICP itself is a Google Sheet problem, not a software problem.

What are the most common ICP mistakes?

Five mistakes kill more outbound programs than bad copy or bad cadence ever do:

1. Defining the ICP from the product, not the data. 'Anyone with a sales team' is not an ICP. Your closed-won list is.
2. Casting too wide. Lenny Rachitsky's #1 ICP mistake is the 'wide net.' If your ICP returns 100,000 accounts, you have no ICP.
3. Collapsing ICP into persona. 'VP of Sales at a SaaS company' mixes the two. ICP first (which companies), persona second (which humans inside).
4. Never updating. First Round calls the ICP 'a living definition.' Quarterly refresh, full rebuild every 6 months.
5. No scoring rubric. A definition without a score is a slide. A score without tiered routing is a spreadsheet nobody opens. Ship the rubric, ship the tiers, ship the suppression list.

OpenView's Kyle Poyar has made the same point about PLG companies: the win is not just identifying ICP accounts, it is prioritising sales time on accounts the product alone cannot close. The ICP exists to allocate the scarcest resource: rep attention.

Concept	Unit of analysis	What it answers	Example	When to use
TAM (Total Addressable Market)	The entire market	How big is the prize if every possible buyer bought?	All 412,000 SaaS companies globally	Board decks, fundraising, market sizing
ICP (Ideal Customer Profile)	Company / account	Which accounts should outbound actually target?	Series B-D B2B SaaS, 100-2,000 employees, US/EU, using Salesforce + Outreach, hiring SDRs	Building outbound lists, account scoring, territory design
Buyer Persona	Individual person	Who do we talk to inside the account, and what do they care about?	Head of Sales, 5-15 reps, owns pipeline number, hates dirty CRM data	Messaging, sequences, content, demo scripts