---
name: abm_playbook
description: "Build and run an account-based marketing program end to end: decide if ABM fits, score and tier target accounts, map the buying committee, run full orchestrated plays across marketing and sales, and measure account coverage, penetration, and pipeline against a non-ABM comparison group. Use when someone asks to start or fix an ABM program, tier a target account list, build an account scoring model, plan a one-to-one or one-to-few campaign, coordinate sales and marketing on named accounts, or report ABM results. For a single atomic play note or one cold email, use an outreach skill instead; this skill ships the whole program."
---

# ABM Playbook

## Purpose

Account-based marketing treats a small set of high-value accounts as the unit of work. Marketing and sales pick the accounts together, learn how each account buys, run coordinated plays against the whole buying committee, and judge the program on pipeline and revenue from those accounts, not on lead volume.

This skill produces a working program, not a strategy memo: a scored and tiered account list, a buying committee map, at least one full play per tier with owners and timing, an orchestration calendar, and a measurement sheet with definitions written down.

The three program models, from the ITSMA typology that most ABM practice follows:

| Model | Also called | Shape | Personalisation |
|---|---|---|---|
| Strategic ABM | 1:1 | A handful of top accounts, each handled on its own | Custom research, custom content, executive contact |
| ABM Lite | 1:few | Clusters of similar accounts (shared industry, pain, or tech stack) | Cluster-level content adapted from existing assets |
| Programmatic ABM | 1:many | Hundreds of accounts | Rules-based personalisation, intent-triggered touches, ads and automated sequences |

Most real programs run all three at once, one per tier. Match the model to account value and to the hours the team actually has. A Tier 1 promise made to 200 accounts is a Tier 3 program wearing a costume.

## Workflow

### 1. Run the fit gate first

ABM costs more per account than broad demand generation. It pays back only when most of these hold:

- Average deal size is large enough that one win covers the cost of working its tier for a quarter.
- Sales cycles are long and involve a committee, so warming several people early changes the outcome.
- The addressable market is named and finite: the team can list the accounts that matter.
- Sales will co-own the list. If account executives will not work named accounts weekly, stop and fix that before building anything else.

If almost anyone can buy the product and deals close fast with one buyer, broad campaigns will beat ABM on cost. Say so and stop.

### 2. Write the ICP as testable criteria

Define the ideal customer profile in fields a CRM or data provider can actually filter on:

- Firmographic: industry, employee band, revenue band, geography, growth stage.
- Technographic: tools in use that imply the problem, tools that integrate with the product, a competing tool with a known renewal window.
- Situation: trigger events such as a new executive in the buying role, funding, expansion into a new region, a regulatory deadline, hiring for roles tied to the problem.
- Disqualifiers: industries never served, sizes the product cannot support, existing customers handled by an expansion motion, accounts in an active legal or support dispute.

Pull the criteria from the last 10 to 20 closed-won and closed-lost deals, not from opinion. Note which criteria came from wins and which from losses.

### 3. Score every candidate account

Score fit separately from timing. A perfect-fit account with no trigger is a nurture account. A mediocre-fit account showing active research is a fast follow, not a Tier 1 account.

Starting model (calibrate it in step 8, the weights are a method, not a law):

| Dimension | Points | What earns points |
|---|---|---|
| Firmographic fit | 0 to 35 | Industry, size, revenue, and geography against the ICP |
| Technographic fit | 0 to 20 | Complementary stack present, integration path clear, competitor due for renewal |
| Intent and triggers | 0 to 25 | Research on relevant topics, visits to high-value pages, trigger event in the last 90 days |
| Relationship | 0 to 20 | Known contacts, past meetings, partner overlap, prior deal history |

Total 100. Worked thresholds to start from: 75 and up is a Tier 1 candidate, 55 to 74 is Tier 2, 35 to 54 is Tier 3, under 35 stays out of the program and gets normal inbound and outbound only.

Score with the helper instead of by hand when the list is longer than a screen:

```bash
python3 ~/workspace/skills/abm-playbook/bin/score_accounts.py accounts.csv --tier1-cap 15
```

Input columns and rules are in the script header. The cap matters: if 60 accounts clear the Tier 1 threshold but the team can run true one-to-one work on 15, the top 15 by score get Tier 1 and the rest drop to Tier 2. Capacity sets the tier, the score only ranks.

### 4. Tier the list and get sales to sign it

- Tier 1 (Strategic, 1:1): few enough that one person can name every account and its owner from memory. Each account gets a named account executive, a marketing owner, and an executive sponsor.
- Tier 2 (Lite, 1:few): group into clusters of roughly 5 to 15 accounts that share one business problem. One play per cluster, adapted content, shared events.
- Tier 3 (Programmatic, 1:many): everyone else who fits. Ads matched to the account list, site personalisation by industry, intent-triggered email, no hand-built assets.

Record for each account: tier, owner, score, the trigger or reason it made the list, and the date it entered. Both sales and marketing leads sign the list in the same document. An unsigned list gets ignored within a month.

Re-tier on a fixed cadence: review intent and triggers monthly for Tier 2 and 3, quarterly for the full list. Promote on a real signal (intent spike, new champion, trigger event). Demote after a full quarter with no engagement and no trigger. Freeze the list during an active play so reporting stays honest.

### 5. Map the buying committee before spending on content

For each Tier 1 and Tier 2 account, fill the committee map (template in `assets/account-plan.md`):

- Economic buyer: owns the budget, signs.
- Champion: feels the problem, sells internally, gives you information.
- Technical evaluator: tests fit, security, and integration, can stall a deal alone.
- Users: live with the outcome, shape adoption and renewal.
- Blocker: prefers the status quo or a rival, has a reason worth learning early.
- Procurement and legal: appear late, set the paperwork clock.

For each person record what they care about, what proof they need, and the one asset or conversation that serves them. A map with only the economic buyer filled in is a single point of failure. Target at least three engaged roles per Tier 1 account before calling the account covered.

Research rule: personalisation must come from a checkable source (the account's own site, a filing, a job post, a public talk, CRM history). Never invent an account's priorities, stack, or headcount plans. If the research comes back thin, personalise to the cluster problem instead and say so in the plan.

### 6. Build the account plan (Tier 1 and cluster leads)

One page per Tier 1 account, one per Tier 2 cluster. Fields are in `assets/account-plan.md`. The plan states: the account's situation in two or three sourced sentences, the problem hypothesis, the committee map, the play chosen, the first three touches with dates and owners, the offer that earns a meeting, and the exit rule if nothing moves.

### 7. Run full plays, orchestrated

A play is a timed sequence of touches by several people on your side aimed at several people on theirs. It is not a single email or ad. The full play library is in `references/plays.md`. It covers:

1. Executive entry (Tier 1, new logo)
2. Intent surge (any tier, signal-triggered)
3. Shared pain cluster (Tier 2)
4. Programmatic warm-up (Tier 3)
5. Opportunity acceleration and multi-threading (open deal going single-threaded)
6. Expansion and renewal (existing customer)
7. Dormant account re-engagement (past loss or gone quiet)

Orchestration rules for every play:

- One shared calendar holds every touch for the account: ads, email, calls, events, direct mail, executive outreach. If a touch is not on the calendar, it does not go out.
- Sequence by stage: air cover and familiarity first, then problem content, then a specific offer, then sales outreach that references what the account has already seen. Sales never opens cold on an account marketing claims to be warming.
- Signal follow-up SLA: when a Tier 1 account crosses the engagement threshold defined in step 8, the account owner acts within one business day and logs what they did. Tier 2 gets two business days.
- Weekly pod meeting, 30 minutes, per tier or cluster: accounts moved, accounts stuck, next touches, one blocker removed. No status theatre, the calendar and CRM are the record.
- Handoff is a shared state, not an event. Marketing and sales both see account engagement, both can trigger the next touch, and neither waits for a baton pass.

Use the orchestration sheet in `assets/orchestration-template.md` to lay out a play week by week before launch.

### 8. Measure at the account level, against a comparison group

Define these before launch, in writing, with the formula and the data source for each (full sheet in `references/measurement.md`):

- Coverage: engaged key contacts divided by identified key contacts, per account and per tier.
- Penetration: how many buying roles per account have engaged at least once.
- Engaged accounts: accounts crossing a stated threshold (for example, 3 or more touches from 2 or more contacts in 30 days). Pick a threshold and keep it fixed for the quarter.
- Marketing qualified accounts (MQA): engaged accounts that also fit the ICP and show a trigger or intent.
- Pipeline: sourced (program was the first recorded touch), influenced (program touched a committee member before the opportunity opened), and open pipeline per tier.
- Win rate, average deal size, and sales cycle length for program accounts.
- Pipeline velocity: (number of deals x average deal size x win rate) / average sales cycle length in days. The result is dollars per day, useful for comparing periods.

Always report program accounts next to a similar set of non-program accounts from the same period. Without that comparison the numbers describe activity, not effect. Review engagement weekly, pipeline monthly, win rate and cycle length quarterly. Engagement with no pipeline movement after two quarters means the plays or the list are wrong. Change one of them, on purpose, and note the change.

### 9. Calibrate and prune

After each quarter, and after every five closed deals from the program:

- Re-score a sample of won and lost accounts with the current model. If winners cluster on one dimension, raise its weight. If losers share a trait the model missed, add it as a disqualifier.
- Check tier honesty: any Tier 1 account that got no custom work this quarter drops a tier next quarter.
- Kill plays with touches sent and no committee engagement across two runs. Keep the record, retire the play.
- Refresh data: intent decays fast and people change jobs. Re-verify committee contacts before each new play.

## Output Contract

A finished run delivers all of these:

1. Fit gate verdict, one paragraph, with the reason if ABM is the wrong tool.
2. ICP criteria sheet with disqualifiers, tied to won and lost deal evidence.
3. Scored, tiered account list (CSV or table) with owner, score, tier, entry reason, and entry date per account, signed off by sales and marketing.
4. Buying committee maps for all Tier 1 accounts and Tier 2 cluster leads.
5. One-page account plans for Tier 1, per `assets/account-plan.md`.
6. At least one full play per active tier, laid out on `assets/orchestration-template.md`, with named owners, dates, and exit rules.
7. Measurement sheet with definitions, formulas, sources, the engaged-account threshold, and the non-ABM comparison group named.
8. Review cadence and the date of the first re-tiering review.

## Operating Rules

1. No list without capacity math. State the hours and people behind each tier before assigning accounts to it.
2. No Tier 1 play launches without a named sales owner per account who has agreed to work it.
3. No personalisation claim without a source. Unsourced insight gets cut or downgraded to a cluster-level problem statement.
4. No reporting on lead counts as the headline. The headline is accounts covered, accounts engaged, and pipeline from target accounts.
5. No mid-play list changes except promotion on a documented trigger. Quiet edits wreck the comparison group and the team's trust in the numbers.
6. Budget follows tiers, and the split is written down and totals 100 percent. Tier 1 normally takes the largest share because custom work costs the most per account. If a plan gives Tier 3 the most money, it needs a written reason.
7. One account, one coordinated experience. If two teams want to run different plays on the same account in the same window, the account owner picks one and the other waits.
8. Respect contact rules: business contacts only, honour opt-outs everywhere at once (one suppression list across email, ads where the platform allows it, and sales sequences), and follow the privacy law that applies in the account's region.
9. If quality and the calendar conflict, ship the smaller, better program: fewer Tier 1 accounts done properly beats a long list done thinly. Report the cut plainly.