Account-based marketing (ABM) guide — how to run it without a big stack

Account-based marketing means treating a named set of companies as the unit of work instead of a list of individuals, and coordinating everything you do toward one outcome per company. The expensive half of the ABM stack — intent data, display retargeting, orchestration platforms — is the optional half. What actually produces meetings is multi-threaded outreach carrying one coherent account narrative, and that needs exactly three data structures rather than six tools. This is the version a small team can run.

Where the term came from, and why the history matters

ITSMA, the Information Technology Services Marketing Association, named the practice, with the term generally credited to Bev Burgess there in 2003 and the first substantial study published shortly after. ITSMA became Momentum ITSMA when Momentum acquired it in 2021.

The history is not trivia. ABM was invented by and for technology-services marketers selling large, complex engagements to a handful of accounts, where the deal size justified genuinely bespoke marketing for one buyer. Every technique in the canon assumes that economics. When a company with a £600 annual contract value copies the playbook wholesale, the arithmetic does not survive contact, and the programme quietly becomes ordinary outbound with a longer planning document.

The transferable part is the unit of work. The non-transferable part is the budget per account. Keep the first, scale the second to your deal size, and most of the confusion about whether ABM "works for small companies" resolves.

The three data structures ABM actually needs

Strip away the vendor category and ABM needs three things to exist in whatever system you use.

An account object that is real. Not a text field on a contact record. A row with its own identity, its own custom fields, and a stable key. In Autocloz, companies are workspace-scoped rows keyed on domain, unique per workspace so two tenants can independently track the same domain, carrying industry, size band and range, country, region, city, technology stack, description, founding year and a workspace-defined set of custom fields — the account-tier, renewal-owner and contract-value fields an ABM programme needs.

A roster attached to it. Every contact you know at that company, visible in one place, with the ability to add and remove them. Without this the account is a label rather than a unit.

One merged activity feed across the roster. This is the structure people skip and the one that makes the difference. When you can see, in one chronological list, that the VP of Engineering opened an email on Tuesday, the CTO accepted a LinkedIn invite on Wednesday, and a call to the CFO went to voicemail on Thursday, you are running account-based work. When you have to open three contact records and reconstruct it, you are not.

Autocloz's account activity endpoint merges two sources — activity rows attached directly to the account, and the full six-source per-lead timeline for every contact on it — and de-duplicates rows that arrive from both, keeping the lead-attributed copy because it carries more context.

Those three structures are the whole technical prerequisite. Everything else in an ABM stack is an efficiency layer.

Why the expensive half is the optional half

Take the standard enterprise stack and ask what each layer changes about a meeting getting booked.

Intent data tells you which accounts are researching your category. Genuinely useful at scale, because it reorders a list of two thousand accounts. On a list of forty, you already know why each one is on it — you put it there for a reason you can state. Buying intent data to reorder forty rows is buying a sorting algorithm for a list you can read.

Display retargeting puts your brand in front of an account's employees. It measurably lifts recall. It does not produce a reply, and its attribution is the hardest in marketing. Worth adding once the outbound motion is working and you want to raise its yield; a poor first purchase.

Orchestration platforms coordinate touches across channels and teams. This is real value — and it is the layer a multichannel sequence tool already provides, which is why the join between ABM and outbound tooling is closer than the two vocabularies suggest.

What none of them replaces is the research. Somebody has to read the account's recent announcements, work out what changed, and write the sentence that shows you did. That is hours, it does not compress, and it is the actual constraint on how many accounts a programme can carry.

The decision rule for adding any layer: name the decision it would change. If you cannot say what you would do differently with the data, you are buying reassurance.

Picking accounts: the tier rule, and what disqualifies one

Selection is where most programmes are decided, long before any message is written.

Score each candidate on three dimensions, each with an observable definition.

• Fit — do they match your ideal customer profile on fields you actually hold? Headcount, industry, country, technology stack. If a dimension is empty on most of your candidates, it cannot be part of the score. The ideal customer profile generator is a faster starting point than an internal debate.
• Value — realistic first-year contract value, not the theoretical ceiling. Use your median closed deal in that segment.
• Trigger — is there a reason this quarter rather than any quarter? A funding round, a relevant hire, a product launch, a leadership change, a public complaint about the tool you replace.

Then tier. Tier one gets a genuinely bespoke narrative and several hours of research. Tier two gets a segment narrative with two or three account-specific details. Tier three gets your standard sequence with the account's name and one observed fact. That is the honest version of "one-to-one, one-to-few, one-to-many", expressed as research hours rather than as a diagram.

The disqualifiers matter as much as the score, and each one is a reason to remove an account that otherwise scores well:

• No reachable contact in a role that can say yes.
• An incumbent contract with more than nine months to run and no dissatisfaction signal.
• A previous loss inside twelve months with no personnel change since.
• A company that is genuinely too small to need what you sell, however well it fits on paper.

Removing accounts is the highest-return editing pass in the whole exercise, because every account you keep is spending research hours that a better one could have had.

Mapping the committee without buying intent data

You need three roles covered per account, and public sources reach all three more often than people expect.

The champion. The person whose job the problem makes harder. Usually findable from a job title plus a public post, a conference talk, or a question they asked somewhere.

The economic buyer. The person whose budget pays. Often obvious from the organisation chart, and often the person who engages with nothing you send — which is precisely why an engagement-scored funnel misses them.

The evaluator. Security, IT, procurement or the end user who has to live with the choice. The one most likely to kill the deal quietly.

Sources that cost nothing: the company's own team page, public professional profiles, conference speaker lists, published job advertisements which name the tools and the reporting line, regulatory filings, and support forums where employees ask questions using their work email. Where to find B2B leads covers the provenance of each of these and the notice obligations attached.

The practical target is three to five named contacts per account with a role label on each. Below three you are single-threaded and one job change ends the deal. Above five or six you are guessing at relevance and the personalisation quality falls.

Forrester's argument for why this is the right unit is worth reading directly: its B2B Revenue Waterfall guide describes the shift from managing individual leads to "identifying, prioritizing, and advancing opportunities with connected buying groups", and its analyst Terry Flaherty wrote on 14 April 2022 that "the MQL era is antiquated and needs to end immediately", on the grounds that scoring individuals both double-counts one deal and misses the members who never cross a threshold. The difference between an MQL and an SQL works through what that means for your handoff.

Sequencing a committee so you do not look like a machine

This is the operational skill that separates ABM from spraying one list harder, and it comes down to one rule: contacts at one account are not independent.

They talk. Four identical emails arriving at one company on one morning is the single most common way an ABM programme announces itself as automation, and the damage is to the account, not to one recipient.

Four practices that prevent it:

Stagger within the account. Two to three days between contacts at the same company. Enough that a forwarded message reads as a colleague's note rather than as a batch.

Vary the opening line by role, not by name. The champion's opening is about the operational problem. The economic buyer's is about cost or risk. The evaluator's is about the integration or the security posture. Same account narrative, three genuinely different first sentences.

Reference the account, never the person's colleague. "I saw you are hiring three data engineers" is fine. "I emailed your colleague Sarah last week" is not, unless Sarah told you to.

Change channel between contacts. A LinkedIn approach to one person and an email to another looks like a company reaching out. Four emails looks like a tool.

Autocloz's step palette makes the staggered shape buildable — a LinkedIn profile view before a connection request, a wait, an email, a branch on whether the connection was accepted. Multichannel outbound without getting banned covers the pacing each channel enforces on you.

The LinkedIn company gate will block your multi-threading by default

This is the single most important operational fact in this guide, and it is the opposite of what an ABM playbook leads you to expect.

Autocloz's LinkedIn dispatcher runs a company-level de-duplication gate before every messaging step. If any other lead at the same company has been touched on LinkedIn inside the window, the step is skipped with the stop reason cross_company_dedup and an error naming the offending touch id. The default window is 30 days. There is a second, longer gate at the person level: once any LinkedIn sender in the workspace has touched a lead, no other sender in the same workspace may touch them for 90 days.

Read that against a three-contact ABM roster. Approach the champion on LinkedIn on the first of the month and the economic buyer's LinkedIn step is skipped for the next thirty days. That gate exists for a good reason — five sellers connecting with five people at one company in one week is the pattern that produces compounding spam reports and risks account restrictions across the whole workspace — but it is calibrated for volume outbound rather than for deliberate account work.

The window is workspace-configurable through channel_defaults.linkedin.company_dedup_days, and one detail matters: a value of zero or below is ignored and the 30-day default applies, so the gate cannot be switched off, only shortened to a positive number of days. A LinkedIn step is also skipped entirely if the lead replied on any channel in the last 14 days, which is correct behaviour and worth knowing before you conclude a step is broken.

The practical design that follows: multi-thread across channels, not within LinkedIn. Champion on LinkedIn, economic buyer by email, evaluator by phone. That is better ABM practice anyway — three identical LinkedIn requests to one company in a week is precisely what marks you as automation — and it means the gate is enforcing a discipline you wanted rather than blocking work you needed.

For email there is no company-level cap. Enrolments are unique per campaign and lead, so one person cannot be double-enrolled in a campaign, but four contacts at one account in one email campaign will all be dispatched on their own schedules. There, the staggering has to come from how you build the sequence and enrol the roster, not from a setting.

Autocloz's free plan covers 5 users and 100,000 contacts, with companies, the account roster and the merged activity feed all included — start free and build one account's roster properly before you build forty.

Measuring by account, and the four numbers that matter

Contact-level metrics mislead at the account level in a specific way: an account where one junior person clicks everything looks better than one where the CFO read a single email and forwarded it. Measure the unit you are working.

Account coverage. The share of target accounts with at least one contact who has engaged — replied, accepted, answered a call. This is your reach.

Depth per engaged account. The average number of engaged contacts within accounts that have any engagement. This is the number that says whether you are actually multi-threading or just finding one friendly person per company.

Meeting rate. The share of target accounts that reached a first meeting. This is the outcome the programme exists for.

Opportunity rate and cycle time. The share that opened an opportunity, and how long it took compared with your non-ABM motion. If the cycle is not shorter and the deal is not larger, the extra research hours are not being repaid, and that is a finding rather than a failure.

Autocloz's account rollup returns four fields per company — contact count, open deal count, open pipeline value and last activity timestamp — each computed under the same row-level scope as the lists beneath them, so a member with restricted visibility sees a header consistent with the table rather than an aggregate that leaks.

A worked forty-account programme, end to end

Illustrative arithmetic for one seller over a quarter, chosen to show the shape rather than measured from any customer.

Forty accounts. Four contacts each, so 160 people. Tier one is eight accounts at three research hours each; tier two is thirty-two at forty-five minutes. That is 24 plus 24, so about 48 hours of research across a quarter — roughly four hours a week, which is real and schedulable.

Outreach: a six-touch multichannel sequence per contact over three weeks, staggered so contacts at one account start two days apart. At 160 contacts and roughly four emails each, that is 640 emails plus LinkedIn actions and calls across a quarter. Against Autocloz's shipped default of 40 sends per mailbox per day, that is well inside the capacity of two mailboxes.

Outcome shape, using deliberately unremarkable rates: 12% of contacts engage in some way, giving about 19 engaged people. Because they cluster in accounts, that is perhaps 14 accounts with at least one engaged contact, so 35% account coverage. Depth of 1.4 engaged contacts per engaged account tells you most accounts are single-threaded — which is the actionable finding, and it points at the follow-up work rather than at more accounts.

Say six of those 14 reach a first meeting and two open an opportunity. Five per cent of the account list became an opportunity in a quarter. Whether that is good depends entirely on your contract value, which is exactly why account selection and value estimation come before anything else. Run the same arithmetic with your own numbers in the CAC and LTV calculator before committing a quarter to it.

What changes between forty accounts and four hundred

At forty, one person holds the whole programme in their head, and the constraint is research hours. At four hundred, three things break.

Research does not scale linearly, so tiering has to become real. At four hundred, tier three is genuinely a segment play with light personalisation, and pretending otherwise produces interchangeable copy with a swapped company name — which is worse than an honest segment message, because it reads as an attempt at personalisation that failed.

Account ownership needs rules. Four hundred accounts across several sellers needs a precedence order for who owns what and what happens when a new enquiry arrives from someone else's account. Lead routing best practices covers the assignment and visibility halves of that.

The merged feed hits its bound. Autocloz's account activity feed fans out across at most 25 leads per account, ordered most-recently-active first so a truncated feed loses the coldest contacts rather than the newest. The response reports whether the bound bit rather than hiding it. For a buying committee of three to five that is irrelevant; for an enterprise account with sixty tracked contacts the feed is a recent-activity view rather than a complete history, and you should know that before relying on it.

What Autocloz does not do for ABM

Being specific is more useful than being encouraging.

There is no intent data. No third-party signal telling you which accounts are researching your category, and no integration that supplies one. If intent is central to your programme, it comes from somewhere else.

There is no advertising layer. No account-targeted display, no audience sync to ad platforms for ABM purposes. Ad lead capture brings leads in; it does not push audiences out.

There is no automatic lead-to-account matching by inference. Contacts attach to companies through the person record, and a lead imported with a company name but no domain does not silently join an existing account. Domain hygiene on import is yours to maintain.

There is no per-account send cap on email. On LinkedIn the opposite is true and it constrains you: a company-level de-duplication gate defaults to 30 days, cannot be disabled — a configured value of zero or below falls back to the default — and will skip a step aimed at a second contact at an account a colleague approached this month.

The account activity feed is bounded at 25 leads per account, and it says so rather than truncating quietly.

And no tool decides which forty companies deserve your quarter. That judgement uses information that mostly is not in any database — who returned a call last year, which competitor just annoyed whom, which champion moved where. The software's job is to make an account the unit of work and to keep the record honest. Choosing the accounts stays yours, and it is the decision with the highest leverage in the entire programme. If you are weighing which system should hold that record, how to choose a CRM without being sold to sets out an evaluation that survives reality, and the Autocloz and Salesforce comparison covers the account-object differences directly.