First, find where the records go stale: open records by last change, owner and field
Start with a staleness read, not a reminder email. Export every open opportunity, and every accepted lead that sales owns, with the date each record last changed and the date each key field last changed: stage, next step and its date, expected close date, amount, and the lost reason on recently closed deals. Rank the records by time since their last change and read from the top. Do not set a cut-off for "stale"; the line between a slow deal and an abandoned record comes from the business's own history by segment and stage, as the weekly pipeline review tutorial says of stuck deals.
Then cut the list three ways, because each cut points somewhere different.
| Cut | If the stale records cluster here | Read next |
|---|---|---|
| By owner | One or two owners, or someone who has left | Whether the owner is present and whether the records were ever handed over |
| By stage | One or two stages hold most of them | Whether those stages have exit criteria and stage-change requirements |
| By field | Some fields always empty or always the same value | Whether anything reads those fields, and when they are required |
| By timing | Records change in bursts just before a meeting, then freeze | Where the meeting's figures come from |
The owner cut catches the case that is not a behaviour problem at all. Records owned by someone who has left still have an owner on paper, so they never appear in a report of unowned records; the tutorial on reassigning CRM records when someone leaves moves them by the written owner rules so every open record has a present owner. Clear those first, or they will inflate every other cut.
One boundary matters. This page is about records that are worked and not updated. Leads that nobody contacts are a different fault with different causes, and the diagnosis of sales not following up on marketing leads starts from them. The test is whether the deal is moving anywhere else, in the forecast spreadsheet, the seller's calendar or their account of it. If it is, the CRM has stopped being the record, which is the problem here.
The stakes are not only internal. Accuracy is one of the principles the GDPR sets out, and contact and deal records describe people, so a CRM that has drifted from reality is wrong about them as well as about the pipeline.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Many fields are empty or hold the same filler; sellers say the record takes too long | Each field on the record against what reads it | The record asks for fields nothing reads, so every update costs the seller time and returns nothing they can see | Retire unread fields in stages; add a rule for adding fields |
| Records are complete at creation and thin later; lost deals carry no reason | Which fields are required at creation and which at each stage change | Requirements sit at creation rather than at the stage changes that need them, so the moments that matter carry no rule | Move requirements to the stage changes; require a lost reason |
| Records pile up at one stage, and sellers disagree about what it means | Whether each stage has a written exit criterion and owner | Stages have no written exit criteria, so sellers cannot tell when a record should move and leave it where it is | Write exit criteria with sales; let workflows move stages a rule can decide |
| The forecast comes from a spreadsheet and the CRM catches up after the meeting, if at all | Where the review's figures come from | The pipeline is reviewed from spreadsheets or memory, so the CRM is never where decisions are made | Run the review from saved CRM reports with a data check first |
| Sellers call the CRM marketing's system and filter its alerts away | What a new record carries when it reaches the seller, and alert volume per seller | The CRM gives sellers nothing back: records arrive without source or reason and alerts fire on every click | Give records source, campaign and reason; keep alerts to events that matter |
| Marketing's service levels are measured from the CRM and sales' data duties are not | What the agreement says about the CRM, and which measures come from it | The agreement between the teams never makes the CRM the arbiter or measures the data side from it | Name the CRM as the arbiter and add a data measure read from it |
Fix, verify and prevent each cause
Fields nothing reads
Every field is a small cost each time a seller touches the record. The tutorial on auditing and documenting CRM fields finds every writer and every reader of each field and treats a field read by nothing as a candidate for retirement. Fields cheap to add, it notes, are expensive to keep. A seller facing a long form whose fields feed no report, workflow or meeting learns that updating is clerical work for somebody else, and stops.
Fix: list each field on the opportunity and lead records with its readers; retire the unread ones in the tutorial's stages, hide, stop writing, archive, then delete; convert free text that reports need into a picklistA CRM field type that accepts one value from a defined list of permitted values, so that records can be reported on instead of typed in many spellings..
Verify: every remaining field has a named reader in the field dictionary, and the stage-change view a seller uses shows only fields something reads.
Prevent: adopt the tutorial's rule for new fields: whoever asks for one names its purpose, writer, reader, owner and type, and the dictionary is checked first.
Requirements at creation instead of the stage change
The CRM fundamentals guide sets required fields at two points: a short minimum when a deal is created (its name and stage, the contact, and an amount or an explicit unknown) and further fields at each stage change, such as a decision-maker contact before proposal and a reason before closed-lost. It puts most of the value in the second kind, because those fields record why the pipeline moved. Load every requirement onto creation and sellers guess at what nobody knows yet, while the moments when they do know carry no rule at all.
Fix: cut creation requirements to the minimum, set each stage's required fields at the change into it, and make the closed-lost reason a required picklist.
Verify: a test deal walked through every stage stops at each change until that stage's fields are filled, and every deal lost in the period carries a reason.
Prevent: design the requirements alongside the stages, so a new or redefined stage gets its fields in the same change.
Stages without exit criteria
The guide to lead stages and pipeline gives every stage three written parts, its definition, its exit criterion and its owner, so a stage is something a record passes or fails rather than a label. Without them, a stage change is an opinion a seller expresses by dragging a card, and a seller unsure of the rule leaves the card where it is. Definitions should be testable from fields, and a criterion that rests on a conversation is read from the logged activity, its type and its outcome. The tutorial on automating lifecycle stage changes takes the next step for the lifecycle stageA CRM field recording where a contact stands in the relationship, such as subscriber, lead, opportunity or customer.: wherever a rule can decide, a workflow changes the stage and stamps the date, and sellers record only the judgements a person must make.
Fix: write each deal stage's definition, exit criterion and owner with sales, testable from a field or an activity; move the changes a rule can decide, such as deal creation or a won deal, to workflows.
Verify: for a sample of records, two people reading the fields name the same stage, and it matches the CRM. If a stage workflow then stops moving records, the diagnosis for a CRM workflow that is not running traces it.
Prevent: keep the stage document where both teams can see it, and change a definition only through the owner of the agreement between the teams.
A pipeline reviewed from spreadsheets or memory
A seller updates whatever their manager reads. When the meeting runs from a spreadsheet, the CRM is double entry, and the spreadsheet wins. The weekly pipeline review tutorial builds the data pack from saved CRM reports instead of hand-built spreadsheets, so the numbers in the pack are the ones sellers see on their records, and asks for close-date changes and stage timestamps to be recorded automatically rather than typed. Before the meeting, the review's owner checks the records that would distort the pack, those with no next step, next steps or close dates in the past, missing source, stages long unchanged, and sends the list to the sellers concerned, because bad data found in the meeting consumes the meeting.
Fix: with the sales leader's sponsorship, which the tutorial treats as a precondition, run the pipeline reviewA short regular meeting where sales and marketing read the same opportunities from the same data and agree an owned action for each stuck deal. and the forecast from the saved report and retire the spreadsheet.
Verify: the figures discussed in the meeting are the CRM report's, and the spreadsheet is no longer updated.
Prevent: keep the data check before every meeting, and read the share of open opportunities whose next step is current and whose close date is believable, as the tutorial does, as a measure of whether the CRM can be trusted.
A CRM that gives sellers nothing back
A record a seller opens should help them act. The follow-up diagnosis linked above names the context a seller needs on the record: the reason for the enquiry, the campaign behind it and the action the person took, held in required properties and shown in the view sellers work from. A record that arrives with none of it reads as marketing's database, and the diagnosis for a blank or wrong lead source finds where the source went missing. Alerts are the other half. The guide to sales alerts from marketing activity warns that alerts used for every click teach sellers to ignore the channel.
Fix: make source, campaign and the qualifying action required and visible on every new record; prune the alert list to events that change the seller's next action, suppressed where they do not apply.
Verify: a sample of new records all carry source, campaign and the qualifying action, and sellers can name the alerts they act on.
Prevent: require a stated reason before any alert is added, and ask at the regular review which alerts sellers ignore, then retire them.
An agreement that never makes the CRM the arbiter
The guide to sales and marketing alignment makes the CRM the arbiter: a disagreement its records cannot settle is really a disagreement about the records. The guide puts sales' data duties into the service-level agreementA published promise between teams, or with a supplier, about what each type of request needs and how quickly it will be handled, agreed rather than imposed.: required opportunity fields filled, and lost reasons entered within an agreed time. It measures both sides automatically from CRM timestamps, on one dashboard, in the same meeting. An agreement that measures lead acceptance from the CRM but says nothing about the records sales keeps tells everyone that the data side is optional.
Fix: name the CRM as the record in the agreement, and add a data measure read from it, such as the share of open opportunities whose next step is current and whose close date is believable, plus lost reasons recorded.
Verify: the measure appears on the shared dashboard and is discussed at the agreement's regular review.
Prevent: when a level is missed, follow the guide's defined escalation and review of whether the level was realistic, not punishment, and score the agreement periodically with the lead management audit scorecard.
Prevention: fewer fields, rules at the stage change and a review run from the CRM
The causes share a pattern: the CRM asked for work and gave little back, so the work moved elsewhere. Prevention reverses that in three habits, plus one check on the checks.
Fewer fields. Keep the field dictionary from the audit tutorial current, with a reader and an owner for every field, and route every new field through the rule for adding one. In general practice, a record a seller can update in the moment after a call stays current; a record that needs a session at the desk is updated before meetings or not at all.
Rules at the stage change. Each stage carries a definition, an exit criterion, an owner and the fields required to enter it, designed together and changed together. Stage changes a rule can decide are made by workflow, with dates stamped automatically, so sellers spend their effort on the judgements only they can record.
A review run from the CRM. The weekly meeting, the forecast and the agreement's dashboard all read the same saved reports. The data check goes to sellers before the meeting, and records owned by anyone who has left are reassigned by rule as part of the leaver process, so no open record waits for an owner who will never return.
Checks that read content, not presence. Goodhart's law warns that when a measure becomes a target, it stops being a good measure. Make field completeness the target and fields fill with whatever passes the check: generic next steps, close dates pushed forward, catch-all lost reasons. The checks have to read what the fields say.
| Field | A presence check asks | A content check asks |
|---|---|---|
| Next step | Is it filled? | Does it have a future date, and has its text changed since the last review? |
| Expected close date | Is it filled? | Is it in the future, and how often has it moved? |
| Stage | Is it set? | How long has the record sat there against the business's own history for that stage? |
| Amount | Is it filled? | Is it a real figure, or an explicit unknown where nobody knows? |
| Lost reason | Is it filled? | Is it a picklist value other than the catch-all, and do the reasons vary by deal? |
Read these checks at team level and by stage, as the agreement's measures, and use the owner cut from the first section only to find records that need a conversation. Pay and performance management sit with sales leadership and outside this page. The CRM section holds the field, stage and leaver guides behind these habits, and the marketing automation troubleshooting hub routes a neighbouring symptom, such as a workflow that stopped or a source that went blank, to the page that diagnoses it.
A label printer maker brings its pipeline back into the CRM (illustrative)
Wexlomir Label Printers
Wexlomir Label Printers, an invented maker of industrial label printers for food producers and warehouses, has 7 sellers and a marketing operations lead. Her pipeline report shows 312 open opportunities, 141 of them with close dates in the past and 96 with next steps unchanged for over a month, while the sales director forecasts from a spreadsheet each seller updates before the Monday meeting.
The staleness read finds three clusters. The opportunity form has 38 fields, and the audit finds only 11 read by any report, workflow or meeting. 64% of the stale records sit in two stages, Evaluation and Negotiation, neither with a written exit criterion. And 29 deals belong to a seller who left 4 months earlier and still owns them on paper.
The company retires 27 unread fields in stages, requires a decision-maker contact at the move to Proposal and a lost reason from a picklist of 8 at closed-lost, and writes exit criteria for all 6 deal stages with the sales director. It reassigns the leaver's 29 deals by territory, runs the Monday meeting from a saved CRM report with the data check sent on Friday, and adds a measure to the agreement between the teams: the share of open opportunities with a next step dated in the future and a close date not in the past, which starts at 41%.
Within a quarter the measure reads 87%, the spreadsheet is retired and the forecast in the meeting and the CRM report give the same number. The check also flags 18 next steps that read "follow up" week after week, which a presence check would have passed, and the sellers concerned replace them with real actions.
Common mistakes
- Starting with a reminder to update. Find which owners, stages and fields hold the stale records first; the cluster names the cause.
- Requiring everything at creation. Sellers guess at what nobody knows yet; require fields at the stage change.
- Keeping fields because they might be useful. A field nothing reads is work with no return.
- Leaving stages as labels. Without exit criteria, a record stays where it was last dragged.
- Running the meeting from a spreadsheet. The CRM becomes double entry and loses.
- Alerting on every click. Sellers learn to ignore alerts and treat the CRM as noise.
- Making completeness a target. Fields fill with placeholders; check what they say.
- Forgetting the leaver's deals. Records with an owner on paper never appear as unowned.
