Introduction
Everyone agrees CRM data quality matters, and almost nobody measures it. What gets measured instead is a completeness percentage — the share of fields that have something in them — which is worse than measuring nothing, because it produces a number that improves while the pipeline gets no more trustworthy.
This article gives five metrics that actually predict whether your pipeline can be relied on. For each: what it is, how to compute it, a sensible target, and what to do when it is bad. Then a worked example scoring a single pipeline across all five, and why the completeness metric deserves to be retired.
Why Field Completeness Is the Wrong Metric
Field completeness treats every field as equally important. Filling in the industry dropdown counts the same as recording a next step, so the number goes up when a rep bulk-edits a field nobody reads, and stays flat when they schedule a follow-up that decides the deal.
It also rewards the wrong behaviour. Told to raise completeness, a team fills fields — selecting the first dropdown value, entering placeholder text, copying values across records. Completeness rises and accuracy falls, and you now have a worse problem, because bad data that looks complete is harder to spot than data that is obviously missing.
The five metrics below share a property completeness does not: each one, when it moves, changes a decision somebody makes.
Metric 1: Share of Open Deals With a Next Step and a Date
Definition: of all open deals past qualification, the percentage with a scheduled future activity that has a date on it.
How to compute: a filter on open deals where the next activity field is empty, expressed as a share of open deals. Most CRMs can produce this in a saved view with no configuration.
Target: above 90 percent. This is not an aspirational number — a deal without a next step is a deal nobody has committed to advancing, and the honest interpretation of 60 percent is that four in ten of your open opportunities are unattended.
When it is bad: do not start with a reminder campaign. Work the list itself: every deal without a next step gets either a real scheduled action or a close-lost decision, this week. That exercise usually removes a chunk of pipeline that was never real, which is uncomfortable and correct. Then make the rule permanent and enforce it at stage transitions rather than on every save.
This is the single most predictive of the five. If you track only one, track this.
Metric 2: Share of Open Deals With Activity in the Last 14 Days
Definition: the percentage of open deals with a logged, substantive interaction — call, meeting, email exchange — within the last fortnight.
How to compute: filter open deals by last activity date. Exclude automated touches if your system logs them, or the number will flatter you.
Target: above 70 percent for pipelines with a normal cycle. Adjust for your business: long enterprise cycles have legitimate quiet periods, and a two-week window may be too tight if your median cycle runs to several months.
When it is bad: distinguish two causes, because they have opposite fixes. Either the deals are genuinely dormant, in which case the pipeline is overstated and needs pruning; or the work is happening and not being logged, in which case you have a capture problem and the answer is email and calendar sync rather than nagging.
The diagnostic is quick: pick ten deals with no recent activity and ask the rep. You will know within an hour which of the two you have.
Metric 3: Share of Closed Deals With a Recorded Loss Reason
Definition: of deals closed lost in the period, the percentage with a structured reason recorded — not free text, a value from a defined list.
How to compute: count closed-lost deals with a populated reason field against all closed-lost deals in the period.
Target: above 95 percent. This is one of the few places where near-total compliance is realistic, because it is a single field, recorded once, at a moment when the rep is already updating the deal.
When it is bad: the cause is almost always the list rather than the discipline. Twenty options mean people pick at random; a free-text field means fifty spellings of the same three answers. Cut to six or so meaningful values — budget, timing, competitor, no decision, product fit, champion left — plus one optional free-text box for context.
This metric is different from the others in that it does not describe your current pipeline at all. It determines whether next year’s pipeline benefits from this year’s losses.
Metric 4: Duplicate Rate on Organizations
Definition: the share of organization records that are duplicates of another record.
How to compute: use your CRM’s duplicate detection, then verify a sample by hand — automated matching both over- and under-reports, so an unverified number is not worth acting on. Sorting alphabetically and scanning is a crude but effective cross-check.
Target: under 2 percent. Duplicates are not merely untidy: they split the history of a relationship, so two reps can work the same account without either seeing the other, and they quietly corrupt every account-level report you produce.
When it is bad: merge, then find the source. Duplicates are almost always created by a process rather than by carelessness — an import without matching rules, a web form that creates rather than matches, an integration writing records blind. Merging without fixing the source means doing this again next quarter.
Metric 5: Share of Open Deals Past Their Expected Close Date
Definition: the percentage of open deals whose expected close date is in the past.
How to compute: a filter on open deals with a close date earlier than today.
Target: under 5 percent. Anything above that means the forecast is being assembled from dates nobody maintains.
When it is bad: each of these deals is one of three things — a date nobody updated, a deal that has actually stalled, or a deal nobody wants to close as lost. Resolve them individually: a new date the rep will defend, or a close-lost decision.
Watch the second-order signal too. A deal whose close date has moved three or four times is telling you something the current date is not, and repeated slippage predicts loss far better than any stage percentage.
Worked Example: Scoring One Pipeline
An illustrative example. A team of eight with 240 open deals runs the five metrics for the first time.
- Next step and date: 58 percent. Target 90. About 100 open deals have nobody committed to a next action.
- Activity in 14 days: 64 percent. Target 70. Close enough to be unremarkable on its own.
- Loss reason recorded: 31 percent. Target 95. Two thirds of last quarter’s losses taught the business nothing.
- Duplicate organizations: 6 percent. Target under 2. Roughly 40 duplicate company records.
- Past close date: 22 percent. Target under 5. Some 53 deals are being forecast on dates that have already expired.
How to read this. The instinct is to launch a data quality programme against all five. That is the wrong response, because two of these are causes and three are symptoms.
Metrics 1 and 5 are the same underlying problem: nobody is maintaining the deals. Fix that and both move, and the activity number moves with them. Metric 3 is a five-minute configuration change — shorten the loss reason list — not a behaviour programme. Metric 4 is a one-off merge plus a fix at the source, and it is genuinely independent of the others.
So the plan is: work the no-next-step list this week, shorten the loss reason list this afternoon, schedule the duplicate merge for next month. Three actions, not five, and the largest of them is the one that also fixes the forecast.
What the numbers meant for the forecast. With 22 percent of deals past their close date and 42 percent without a next step, the pipeline value in the CRM was materially overstated — not because anyone was dishonest, but because dead deals had never been closed. The first week of cleanup removed about a fifth of the open pipeline. The forecast that came out the other side was smaller and, for the first time, worth arguing about.
Keeping the Numbers Current
Measuring these five once is a useful shock. Measuring them every week is what changes anything, and that is where doing it by hand quietly fails — the person who built the filters gets busy, and the report stops.
This is the case for automating the monitoring rather than the entry. SalesBond audits how the pipeline and CRM are structured, including data quality, duplicates and next-step rules, and turns the findings into a prioritised Fix Plan, with nothing changed until an authorised user approves it. Deal Health watches active deals continuously and flags stalled deals, missing next steps and overdue activities, which is exactly the raw material behind metrics 1, 2 and 5.
FAQ
What is the most important CRM data quality metric?
The share of open deals with a next step and a date. It is cheap to measure, entirely within the rep’s control, and it predicts both win rate and forecast accuracy better than any other single field.
Why not just measure field completeness?
Because it treats every field as equally important and rewards filling fields rather than recording the truth. Teams pushed on completeness produce complete data that is less accurate than what they had before.
How often should these be measured?
Weekly for the three deal-level metrics, quarterly for duplicates and loss reasons. Weekly matters because these numbers drift continuously rather than in steps.
Should data quality metrics be tracked per rep?
Track them by team first. Per-rep leaderboards tend to produce data entered to look good rather than to be true, which is the opposite of what you are trying to achieve.
What is a realistic target for a team starting from a bad baseline?
Pick one metric and move it, rather than moving all five slightly. Next step and date is the right first choice, and a team going from 60 to 90 percent on it will see the other numbers improve without being worked on directly.
Conclusion
Field completeness measures effort; these five measure whether the pipeline can be trusted. Next step and date tells you whether deals are being worked. Recent activity tells you whether the pipeline reflects reality. Loss reasons decide whether losses teach you anything. Duplicate rate protects the account view. Deals past their close date tell you how much of the forecast is built on expired assumptions.
Run all five once to get the shock, then act on the two that are causes rather than the three that are symptoms. Start this afternoon with the cheapest one: count your open deals with no next step, as a share of the total. Whatever that number is, it is the ceiling on how much your forecast is currently worth.
