At a glance
Before scoring your target accounts, exclude those outside the campaign brief. Then assess business fit, the quality of your information and evidence of a current buying decision separately. This guide helps ABM managers agree priorities with your sales team without treating complete data or frequent activity as proof of readiness.
You’ve got more potential accounts than your team can support. Your sales colleagues have their favourites, marketing has engagement data and your data supplier offers a ranking. When those lists disagree, another combined score may hide the disagreement rather than help you resolve it.
An account can be a strong fit without a current project. Another can have an active evaluation but gaps in your research. A third can generate plenty of activity while sitting outside your service geography. Your team needs to see those differences because each calls for a different action.
Decide what the ranking is allowed to change
Our account-prioritisation approach starts with the commercial objective, evidence and delivery feasibility, then makes the shortlist and reassessment conditions explicit. Use the three dimensions below to expose uncertainty within that discussion, not to replace commercial value or feasibility with a score.
A practical decision record might read: “Relevant operating model; no confirmed project; one missing technical fact; retain in cluster activity and commission only the research needed to resolve compatibility.” Another might read: “Confirmed evaluation; suitable scope; account owner available; allocate individual preparation before the next customer meeting.” These are illustrative decisions, not descriptions of client accounts.
The output is an allocation of effort: research, shared activation, individual account support or no investment for now. It is not a probability of purchase. An account can be first for research and lower priority for paid activation because the unanswered fact determines whether the offer fits.
Record an override against the same dimensions. Sales knowledge can legitimately change priority when it supplies new evidence or commercial context. A request to favour an account without an explanation should remain a discussion, not silently become model validation. Keep person-level response scoring outside this account investment decision.
Exclude ineligible accounts before scoring
Some criteria should be gates rather than points. If the service can’t be delivered in a market, a high engagement score shouldn’t make that account eligible. If the brief excludes existing customers or particular business types, apply that rule explicitly.
Create an eligibility status with a reason and a review date where needed. Distinguish excluded from unresolved. Missing company-size information shouldn’t automatically become a positive match or a permanent rejection.
Only then compare accounts within the eligible population. This prevents the scoring model from solving the wrong problem: ranking organisations the programme can’t or shouldn’t target.
Keep fit, evidence and readiness visible
| Dimension | What it describes | What it doesn’t establish |
|---|---|---|
| Fit | Relevance of the business, use case and delivery conditions | A current purchase process |
| Evidence strength | Reliability, recency and specificity of supporting information | Commercial attractiveness by itself |
| Readiness | Direct or contextual evidence of a decision under way | Certainty of winning or confirmed budget |
Within fit, choose criteria that connect to the service: operating model, company characteristics, compatible environment and plausible use case. Within evidence strength, assess source quality, entity accuracy and freshness. Within readiness, separate directly stated plans from weaker behavioural clues.
HubSpot’s lead-scoring documentation supports separate fit and engagement measures. We suggest recording evidence strength separately so the team can see where a high score rests on an uncertain assumption. You may need a custom field or a working note to do that.
Agree what each score means
A simple working scale might run from zero to three. Zero means no usable evidence. One means an indirect or ambiguous clue. Two means a relevant, recent source with a clear limitation. Three means direct, well-scoped evidence appropriate to the decision.
These values are illustrative. The important part is a written rule and an example for each level. A researcher shouldn’t award three because a source feels persuasive while another uses three only for direct confirmation.
Avoid averaging unrelated evidence into false confidence. Three sources repeating the same original announcement aren’t three independent confirmations. A recent source about the wrong subsidiary remains weak for the target entity.
Record contradictions alongside the score. The explanation is often more useful than the number when sales decides how to approach the account.
Why the highest total may suggest the wrong next step
Consider three fictional accounts. Account A has strong fit, well-verified operating information and no evidence of an active initiative. Account B has strong fit, a directly stated evaluation and an uncertain stakeholder map. Account C has frequent category activity but falls outside the service geography.
Account A belongs in a relevant education or relationship programme. Account B merits focused research and a timely response suited to the stated evaluation. Account C remains excluded from this campaign unless the delivery scope changes.
A combined total might rank A above B because A has more complete data, or rank C highly because it has more recorded activity. Separate dimensions reveal why those rankings wouldn’t support the intended action.
You may need more than one priority list. Account B could be first for stakeholder research, while Account A is ready for a shared educational campaign. A single master ranking only helps if it serves the decision you’re actually making.
Treat intent as one input
6sense’s discussion of contextual signals cautions against treating isolated activity as sufficient context. In an account model, preserve the signal’s topic, timing and scope rather than converting every observation into the same readiness points.
A broad category surge may justify investigation. A direct request about a specific project may justify a different response. Neither should overwrite a known exclusion or erase uncertainty about the operating entity.
Use first-party and third-party intent data for the questions each can reasonably answer. Account-level activity doesn’t prove that a particular named person performed the research.
Validate against decisions and outcomes
Before activation, review a sample of accounts with your sales team and delivery colleagues. Ask whether the model produces sensible actions and whether the explanations match known account circumstances. This checks usability, not predictive accuracy.
After the programme runs, compare outcomes across score bands using a consistent cohort and enough elapsed time. Keep track of how accounts were treated. If high-scoring accounts received much more effort, their better outcomes can’t automatically be attributed to the score’s predictive quality.
Inspect missed opportunities and false priorities. Which promising accounts scored poorly because data was missing? Which high-priority accounts consumed effort without a relevant need? Adjust specific rules rather than changing weights until the historical results look impressive.
Prevent missing data from becoming a hidden penalty
An account can score poorly because it is unsuitable or because the available data is incomplete. Those cases shouldn’t have the same label. Add a completeness or unresolved-evidence status alongside the fit assessment, without allowing completeness itself to become a proxy for buying potential.
A listed company may have a rich public record while a private business with an equally relevant use case is harder to research. If data completeness drives the ranking, your team can end up pursuing the easiest companies to describe. Keep the research gap visible so it doesn’t quietly become a judgement about commercial value.
Use a research queue for accounts where one missing fact would materially change the decision. State the fact, why it matters and how it can be checked. Don’t commission open-ended research on every incomplete record; the investigation should have a decision value.
Also define a limit on overrides. Sales knowledge can legitimately correct a model, but an override should include the evidence and reason. “Strategic account” without explanation isn’t a substitute for selection criteria. Review repeated overrides to see whether the model is missing an important dimension.
Check how easily the ranking changes
Change uncertain weights within a plausible range and inspect which accounts move substantially. If small changes completely reorder the priority list, the apparent precision of the ranking is misleading. Use broader priority bands or gather better evidence before making high-cost commitments.
Check concentration as well. A model may place most of the budget in one sector or region because a particular data field is more available there. Determine whether that concentration reflects strategy or measurement coverage.
Document the intended use of the score. A model built to prioritise research shouldn’t later be presented as a probability of winning. A model built for one service may need different criteria for another. Versioning should record both rule changes and changes in purpose.
Finally, keep the unscored facts accessible. The best review question is often not “Why is this account a 74?” but “Which evidence would make us choose a different action?” A transparent model should make that answer easier to find.
Turn the score into an action
The output should include a recommended action, owner, reason and review trigger. “High fit, uncertain need: validate the problem through the account owner” is more useful than “82 out of 100”.
Refresh the relevant dimensions when circumstances change. A new conversation can strengthen readiness. An old technology observation may weaken evidence confidence. A change in service scope may alter eligibility.
Our guide to ABM intelligence and account selection provides the broader selection context. ABM Logic can help translate scoring criteria into a programme where the next action remains visible.
