ABM measurement is about influence, not attribution

ABM measurement works when it records what marketing touched on the way to each deal, by named person and by date, and compares the accounts you reached with the ones you didn't. That's influence. It answers the question finance is actually asking, which is what the spend produced and how you know, and it does it without asking anyone to agree on how credit should be split between a webinar and a retargeting ad.

We've written about why attribution models undercount marketing(opens in a new tab) in the first place. This post is the replacement: a working framework for measuring ABM as influence, in five decisions, plus the rules that keep it honest.

TL;DR

• Influence records what marketing touched before a deal opened. Attribution assigns credit for it. Only the first one is checkable.
• Five decisions make a framework: what counts as a touch, what counts as reached, the window, the comparison, and the reporting level.
• The comparison that matters is reached accounts against unreached ones, at program level. Channel isolation inside an ABM program never settles.
• Every figure on the report should open to a name, a date, and a deal.
• Say what the framework leaves out. It undercounts brand and word of mouth by design, and that's fine.

Influence and attribution are different questions

Attribution asks which touch deserves the deal. Multi-touch models answer by splitting credit according to a rule chosen in advance, and last-touch models answer by giving it to whatever came last, usually the demo request. Both are built for a lead funnel where one person converts once.

ABM doesn't work like that. You picked the account before it showed interest, you reached six or more people inside it through ads, outbound, events, and content in the same few weeks, and the opportunity opened a quarter later. Asking which of those touches "caused" the deal is a question with no stable answer, and every answer you give will be argued with by sales.

Influence asks something narrower and more useful: which people on this buying committee did marketing reach, through what, and when, before the deal opened? That's a record rather than a model, and anyone can open it and check it, which is the property attribution lost the moment it started weighting things.

The framework: five decisions

1. What counts as a touch

Write the list down before you measure anything. A workable set: an ad served to an identified contact at a target account, a site visit by an identified person, an email reply, a webinar or event attendance, a piece of content opened. Account-level activity counts as timing evidence, not as a touch; "someone at Acme visited pricing" tells you the account is in motion and nothing about the committee.

The discipline here is identification. A touch on a guessed identity is a touch on nobody. Report fewer people and stand behind every one.

2. What counts as reached

Reaching one person at an account is a start. Reaching the committee is influence. Set a coverage threshold, for example two or more of the roles that have to approve the purchase, and count an account as reached only when the program clears it. What a buying group is(opens in a new tab) covers which roles to map.

Coverage is also the one number in this framework marketing fully controls, which makes it the leading indicator for everything below.

3. The window

Count touches in a fixed window before opportunity creation. Ninety days is a reasonable default for a mid-market cycle; longer for enterprise. Say the window out loud on the report, because it's the assumption most likely to be questioned.

Two rules inside the window. Engagement goes stale in weeks, so a visit from last quarter is context rather than a current touch. And history doesn't get rewritten: when a champion changes jobs, their engagement stays credited to the account where it happened, so the story holds up months later.

4. The comparison

This is the decision that makes the framework defensible. Compare accounts the program reached with similar accounts it didn't, on meetings, pipeline, and win rate. The gap is the program's effect. Our ABM playbook(opens in a new tab) measures the whole program this way rather than trying to isolate channels, because inside a surround-sound program the channel question never settles. The program behind it reported 60% more meetings in the ABM cohort than in the control.

Set the control before launch: hold back a group of accounts that match the target list on fit, leave them in your normal demand programs, and leave them alone. Holding accounts back feels like a cost. It's the only way to show what the program added.

5. The reporting level

Report at program level, with deals openable. The headline is influenced pipeline from reached accounts, next to pipeline from the control, next to pipeline-to-spend. Underneath, every opportunity in the number lists the contacts reached, the programs that reached them, and the dates. The ABM ROI math(opens in a new tab) covers the ratio and what to count as spend.

Don't report channel-level influence as if it were a finding. "LinkedIn influenced 40% of pipeline" is the attribution argument wearing a new name.

What it looks like on one deal

Take a hypothetical opportunity that opened in June. The record shows a VP of Operations who was served ads from March and visited the integrations page twice in April. A director on the same account joined a webinar in May. A finance lead saw the ads and never clicked. Sales first spoke to the account in late May; the opportunity opened three weeks later.

The influence record says marketing reached three of the five people on the committee before sales did, names them, and dates each touch. It doesn't say the webinar deserves 30%. It doesn't need to. The person who asks "which people?" gets an answer, and the person who asks "would this have happened anyway?" gets the cohort comparison.

The objections, and what to say

• "Finance wants a number." Give two: pipeline-to-spend for the program, and the lift of reached accounts over the control. Both are single numbers, and both can be opened.
• "Sales will say they did it." They did. Influence doesn't claim the deal; it records what marketing touched on the way to it. That framing is the one that survives the room, because it asks nobody to give up credit.
• "This undercounts brand." Yes. Brand, word of mouth, and the podcast someone heard in the car don't show up as touches on identified people. Say so on the report. A framework that admits what it misses is trusted more than one that claims to see everything.
• "Our attribution tool already does this." It does something adjacent. Check whether it records touches on identified people or infers them from UTMs, whether it even-weights by default because nobody set it up, and whether your opportunities have contact roles filled in at all. A model can't credit a committee member who was never attached to the deal.

Rules that keep the record honest

1. 1
   Don't add signal types together. Company-level reach and person-level engagement are different populations.
2. 2
   Don't rewrite history when people move.
3. 3
   Report the ratio with the biggest deal in and out, and label anything under two quarters old as directional.
4. 4
   Never claim one touch caused the deal, in the report or in the room.
5. 5
   Keep the slide and the evidence together. A number that can't be opened is a number that gets cut.

How Vector measures influence

Vector keeps ad engagement, site visits, meetings, and the deal on one timeline per account, with people attached only when a real person has been confidently identified. When an opportunity opens or closes, the account's story lists the marketing moments on the way there, by contact and by date. It is deliberately not multi-touch attribution: no model carves the deal into percentages, and the real touches show in order. The same timeline lets you compare deals your ads touched with deals they didn't, which is the cohort comparison above without the spreadsheet work.

Reporting is included in every plan, so the influence record arrives with the campaign rather than as a separate purchase. If your ABM measurement produces a different number every time someone changes the model, see how Vector handles ABM(opens in a new tab).

Book a demo →(opens in a new tab)

Frequently asked questions

What is ABM measurement?

It is how an account-based program shows what it produced. The approach that holds up with finance records what marketing touched on the way to each deal, by named person and date, and compares accounts the program reached with similar accounts it did not.

What is the difference between influence and attribution in ABM?

Attribution assigns credit for a deal to touches using a model chosen in advance. Influence records which people on the buying committee marketing reached, through what, and when, before the deal opened. Influence is a record anyone can check; attribution is a rule people argue about.

How do you build an ABM measurement framework?

Make five decisions: what counts as a touch, what counts as a reached account, the window before opportunity creation, the comparison group, and the level you report at. Then keep every figure openable to a name, a date, and a deal.

How do you measure ABM without attribution?

Compare the accounts the program reached with a control group of similar accounts on meetings, pipeline, and win rate, and report influenced pipeline with the contacts and dates attached. The gap between the groups is the program's effect and it does not depend on a credit model.

What does ABM measurement miss?

Brand, word of mouth, and anything that does not show up as a touch on an identified person. Say so on the report. A framework that admits what it leaves out is trusted more than one that claims to see everything.