Building ICP and Target Lists

• Source: Report research library
• Frameworks applied: ABM account tiering (1:1 / 1:few / 1:many), TAM/SAM/SOM bottom-up market sizing, closed-won ICP attribute analysis, Clay waterfall enrichment, suppression and dedupe validation gate

Forces capacity declaration before tiering and demands closed-lost evidence so ICP attributes have discriminating power rather than survivorship shape.

Known weakness, published as found: Named vendor lists (nine enrichment providers, Clay-specific mechanics) will go stale fast — abstract to provider roles and move the named chain into references/enrichment-providers.md. Resolve the scope overlap with designing-territories-and-quotas: 'territory carve-up' appears as a trigger keyword here but the sibling owns it.

1. 1.Copy the SKILL.md text below, or download the raw file.
2. 2.Create a folder named exactly building-icp-and-target-lists in your agent's skills directory.
3. 3.Save the file inside that folder as SKILL.md.
4. 4.Ask the agent one of the trigger requests below.
5. 5.Check the output against what you already know before it leaves your desk.

• Help me define our ICP from last year's closed-won deals and build a tier 1 target account list
• We need a bottom-up TAM SAM SOM for the mid-market segment before annual planning
• Set up a waterfall enrichment plan so our Clay list has verified work emails and mobiles

• Write me a 5-touch cold email sequence for VP Sales personas
• Our emails are landing in spam since we doubled volume - what should we check in DMARC?

---
name: building-icp-and-target-lists
description: >-
  Derives an ICP from closed-won evidence, sizes TAM/SAM/SOM bottom-up, assigns
  accounts to 1:1 / 1:few / 1:many tiers against real capacity, and builds an
  enriched, deduplicated target account and contact list using waterfall
  enrichment. Use when the user says define our ICP, build a target account
  list, ICP refresh, TAM SAM SOM, market sizing, account tiering, tier 1
  accounts, named accounts, build a prospect list, waterfall enrichment, Clay
  table, list build, territory carve-up, or asks who should we be prospecting
  into. Use this skill whenever the task is deciding which accounts and
  contacts to work, even if the user never says "ICP". Do NOT use for writing
  the outreach messages or cadence design (see designing-outbound-sequences),
  for triggering plays off live buying signals (see running-signal-based-plays),
  or for sending-domain and inbox setup (see protecting-email-deliverability).
metadata:
  version: "1.0"
---

# Building ICP and target lists

Turn closed-won evidence into a defended ICP, a bottom-up TAM/SAM/SOM, a tiered
account list, and an enriched contact list ready for sequencing. One job:
deciding **who** to work. Message copy, cadence, signal plays, and sending
infrastructure are out of scope.

## Workflow

Copy this checklist into your reply and tick items as you complete them:

```
- [ ] 1. Pull closed-won and closed-lost evidence
- [ ] 2. Derive ICP attributes and write disqualifiers
- [ ] 3. Size TAM -> SAM -> SOM bottom-up
- [ ] 4. Declare capacity, then tier accounts
- [ ] 5. Build the account list and run waterfall enrichment
- [ ] 6. Validate the list, fix, re-validate until clean
- [ ] 7. Emit the ICP and target list brief
```

**1. Pull closed-won and closed-lost evidence.** Require a minimum of the last
4 quarters of closed-won deals with ACV, cycle length, employee count,
industry, region, buying-committee titles, source, and the technologies they
ran at purchase. Pull closed-lost with the loss reason in the same shape.
Never build an ICP from won deals alone, because attributes shared by wins and
losses have no discriminating power and inflate the list with lookalikes that
never convert.

**2. Derive ICP attributes and disqualifiers.** For each candidate attribute,
compute win rate, median ACV, and median cycle length for accounts with the
attribute versus without. Keep an attribute only if it moves win rate or ACV
materially in the won set *and* is queryable in your data provider — an
attribute you cannot filter on is a narrative, not a segment. This step is
judgment: you decide the cut points, but state for every kept attribute the
number that justified it, and write an explicit disqualifier list (segments,
sizes, geos, or tech stacks you will refuse) because outbound lists degrade
from the bottom, not the top.

**3. Size TAM -> SAM -> SOM bottom-up.** Use the nesting rule: TAM contains
SAM, SAM contains SOM, and each layer needs different data
([HG Insights](https://hginsights.com/blog/tam-sam-som-the-complete-guide-to-market-sizing/)).

- **TAM** — count companies fitting the ICP and multiply by average contract
  value; treat that as a TAM *floor*. Prefer this bottom-up count over
  top-down analyst figures, because analyst reports reflect "what survey-based
  analyst reports suggest the market might support," not what companies
  actually buy ([HG Insights](https://hginsights.com/blog/tam-sam-som-the-complete-guide-to-market-sizing/)).
  TAM is a ceiling, not a target — use it for board and fundraising sizing only.
- **SAM** — apply ICP filters to TAM: geography, employee band, revenue range,
  and specific technology installs. A well-formed SAM reads like "Mid-market
  North American companies with 200 to 2,000 employees that run Salesforce and
  have an active sales intelligence stack"
  ([HG Insights](https://hginsights.com/blog/tam-sam-som-the-complete-guide-to-market-sizing/)).
  These filters also define territory structure and hiring plan, so change them
  deliberately.
- **SOM** — layer intent signals onto SAM to isolate the in-market population,
  then apply honest competitive share and win-rate inputs. SOM is the number
  that drives the annual plan and quota model, and it is the hardest of the
  three because it needs those honest inputs
  ([HG Insights](https://hginsights.com/blog/tam-sam-som-the-complete-guide-to-market-sizing/)).

Show all three as a count of accounts *and* a dollar figure, plus the exact
filter expression used at each step, so the numbers are reproducible.

**4. Declare capacity, then tier.** Write down staffing, content bandwidth, and
sales engagement capacity **before** assigning a single account, because "the
tier an account deserves may differ from the tier the organization can actually
sustain" ([DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/)).
Then fill tiers from the top against those constraints:

| Tier | Account volume | Model | Operating requirement |
|---|---|---|---|
| 1:1 | 5–30 accounts | Bespoke content, personalized sequences, account-specific events, custom research | Dedicated marketing resource paired to an AE; visibility into buying context, org dynamics, evaluation criteria |
| 1:few | 5–15 accounts per program | Cluster-specific content, coordinated but not individualized outreach | Content turnaround of 2–4 weeks from briefing to deployment; account-level engagement tracking |
| 1:many | 100–500+ named accounts | Persona- and vertical-driven content, ads, triggered outreach | Clean target list, intent data integration, account-matched advertising |

(all rows: [DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/))

Assign on three inputs: (a) deal size and strategic value — mid-six or
seven-figure expected ACV justifies 1:1, and low-ACV accounts can still earn
top tier as anchor customers in a new vertical; (b) buying complexity — larger
committees and longer approval chains justify sustained 1:1 multithreading;
(c) organizational capacity
([DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/)).
Use exactly three tiers; a fourth "typically produces diminishing clarity
rather than additional strategic precision." Assign separate owners per tier —
the person owning the top 10 strategic accounts must not also own the
300-account 1:many infrastructure, because the 1:1 tier is what suffers
([DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/)).

**5. Build the list and run waterfall enrichment.** Waterfall enrichment
searches providers sequentially until a valid match is found and stops there,
so cost is incurred only on providers actually queried
([Clay](https://www.clay.com/waterfall-enrichment)). Configure it as an ordered
fallback chain, cheapest-and-highest-hit-rate first:

- Work email — inputs are name + company domain; chain across providers such as
  Prospeo, DropContact, Datagma, Hunter, PeopleDataLabs, Nimbler, Apollo,
  Lusha, and Snov ([Clay](https://www.clay.com/waterfall-enrichment)).
- Mobile number — chain across providers such as People Data Labs, ContactOut,
  and Selligence ([Clay](https://www.clay.com/waterfall-enrichment)).
- Any other field can be waterfalled the same way, from technology stacks to
  job openings ([Clay](https://www.clay.com/waterfall-enrichment)).

Waterfalls raise match *rate*, not accuracy, so pair every chain with a
syntax + MX + catch-all check and route unverifiable addresses to a phone- or
social-only track rather than the email sequence
([Clay](https://www.clay.com/waterfall-enrichment)).

**6. Validate, fix, re-validate.** Run every check below. If any fails, fix the
list and run the whole block again. Only proceed to hand-off when all checks
pass — a list published with unresolved failures poisons deliverability and
tier integrity at the same time.

```
- [ ] Every account matches at least one kept ICP attribute and zero disqualifiers
- [ ] No account appears in two tiers; tier counts are inside the volume bands
- [ ] 1:1 tier count <= declared capacity; no 1:1 account lacks a named owner
- [ ] Duplicate accounts merged on canonical domain, not company name
- [ ] Existing customers, open opportunities, active sequences, partners and
      competitors suppressed
- [ ] Every contact has a verified email OR a phone/social route flag
- [ ] Unsubscribes and prior spam complaints suppressed permanently
- [ ] Email verification failure rate per provider recorded
```

**7. Emit the brief.** Use the output template below.

## Output format

Use this exact section order and headings, because downstream sequence design
and signal routing read them positionally. Wording inside sections is yours.

```markdown
# ICP and target list — <segment> <quarter>

## ICP definition
| Attribute | Include | Evidence (win rate / ACV / cycle) | Queryable in |
## Disqualifiers
- <segment or attribute> — because <reason>

## Market sizing
| Layer | Filter expression | Accounts | Dollars | Use |
| TAM | ... | ... | ... | board sizing |
| SAM | ... | ... | ... | territory + ICP |
| SOM | ... | ... | ... | annual plan + quota |

## Capacity declaration
- Marketing resource: <n> | Content turnaround: <n> weeks | Sales capacity: <n> accounts/rep

## Tiers
| Tier | Accounts | Owner | Content model | Review cadence |

## Enrichment result
| Field | Providers in chain | Match rate | Verified rate | Cost per match |

## Suppression log
<counts by reason>

## Open risks
<attributes with thin evidence; layers with estimated inputs>
```

## Gotchas

- A nominal 1:1 tier holding 50 accounts that receive 1:many-quality engagement
  "is not a tier system. It is diluted effort with a misleading label"
  ([DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/)).
  Cut the tier to capacity instead of renaming the problem.
- Set a cull cadence at list creation: accounts sitting in 1:1 for 12 months
  without meaningful progress get a fundamentally different approach or a
  demotion, and a tier structure unrevised for 18 months is "a static list with
  tiered labels" ([DemandScience](https://demandscience.com/resources/blog/abm-account-tiers/)).
- Deduplicate on canonical root domain, not company name. Subsidiary and
  DBA-name records survive name-based dedupe and produce two reps emailing the
  same buying committee, which reads as spam from the recipient side.
- Never buy contact lists to fill a thin SAM. Google's sender guidelines state
  plainly not to purchase email addresses, and purchased data is the fastest
  route to the spam-rate ceiling that governs whether any of the list is
  reachable ([Google](https://support.google.com/a/answer/81126)).
- List size works against reply rate: campaigns under 50 recipients reply at
  5.8% versus 2.1% for campaigns over 500 recipients, a 2.8x gap
  ([11x](https://www.11x.ai/guides/josh-braun-cold-outreach-method)). Size
  1:many segments as many small clusters, not one bulk load.
- Do not treat SOM as a discounted SAM percentage. Build it from the in-market
  population that intent data surfaces, because a share-of-SAM guess carries no
  information about who is buying this quarter
  ([HG Insights](https://hginsights.com/blog/tam-sam-som-the-complete-guide-to-market-sizing/)).
- Enrichment match rate is a per-provider fact, not a list fact. Log it per
  provider per field; when one provider silently degrades, the waterfall hides
  it as a modest aggregate drop while costs shift down the chain
  ([Clay](https://www.clay.com/waterfall-enrichment)).

What does the Building ICP and Target Lists skill do?

Derives ICP from closed-won data, sizes TAM/SAM/SOM, tiers accounts, and builds an enriched target list

Where does the Building ICP and Target Lists skill come from?

Report research library. It was written by The Revenue AI Report against a 12 criterion quality rubric and graded in an independent scoring pass.

Why was the Building ICP and Target Lists skill chosen for this library?

Forces capacity declaration before tiering and demands closed-lost evidence so ICP attributes have discriminating power rather than survivorship shape.

When should the Building ICP and Target Lists skill not be used?

Do not use it for: Write me a 5-touch cold email sequence for VP Sales personas Or: Our emails are landing in spam since we doubled volume - what should we check in DMARC?

How do I install the Building ICP and Target Lists SKILL.md file?

Download the file, create a folder named exactly building-icp-and-target-lists inside your agent's skills directory, and save the file inside it as SKILL.md. The agent loads it when a request matches the description.