Why stages set by hand drift
The guide to lead stages and pipeline turns stages from labels into tests: each has a definition of what must be true, an exit criterion that moves a record on and an owner. Once those three are written, it says, "a stage change becomes a fact that a workflow can act on". The glossary entry for lifecycle stage draws the consequence: "Stages should change by workflow, not by memory", and the field moves backward only by a deliberate rule. Many CRMs fall short of both in the same few ways.
A stage typed by hand changes late or not at all. A stage written by two systems has no owner: the guide to syncing the CRM and the marketing platform says lifecycle stage is the field both sides write most often, so it holds whichever value was saved last. An import set to update the records it matches resets customers to leads, the failure the tutorial on importing contact lists is built to prevent. And a stage changed without a date cannot be read by cohort. The lead stages guide notes: "Stage dates recorded in the CRM are what make cohort reading possible." A report built on the current value counts where records are today, not when they got there.
The marketing funnel assumes an ordered set of stages that records enter one after another; the automation's job is to keep that order true in the data. When it is not, the CRM usually shows it.
| What you see | Where it shows | What it suggests |
|---|---|---|
| Customers still receiving prospect journeys weeks after their first order | Contacts at lead or below with a won deal | The customer stage is not set when a deal is won, or something sets it back |
| Stage counts that change after a list load or an overnight sync | Stage history grouped by the user or integration that wrote it | More than one writer, or an import allowed to overwrite |
| More records at a later stage than at the one before it in a cohort report | Records with a later stage and no date for an earlier one | Records skip stages and nothing records the skipped dates |
| A stage report that cannot be rebuilt for last quarter | Stages with no date field at all | Reports read the current value, not the date entered |
| Sellers and marketing quoting different counts for the same stage | Who may edit the field, and how often it is edited by hand | Manual edits with no reason and no log |
Before you start
- Written stage definitions with exit criteria and owners, agreed by marketing and sales, from the lead stages guide; a rule cannot automate a stage nobody has defined. The comparison of MQLs and SQLs sets out who decides each of the two qualified stages.
- The owning system for the stage, from the sync guide's field table, with the CRM sync change checklist as the gate for changing it.
- The automation inventory from the automation governance guide, to find every workflow that writes or reads the stage, and its naming convention.
- Agreement from whoever administers the CRM to add fields, restrict who edits the stage and build workflows; the CRM fundamentals guide says which fields marketing owns.
- Seed contactsAn inbox or test contact the business controls, added to a send or run through a workflow so the team sees exactly what real recipients would receive., unmistakably named and excluded from reporting.
- A place to record the change, such as the measurement change log, because every stage report moves on the day the rules go live.
The steps
List every way the stage changes today
Export the stage history for a recent period and group every change by what made it: each form, the scoring workflow, the sync, each import, each integration user and each person who edited the field by hand. Then list the workflows in the inventory that read the stage, because those misbehave when it is wrong. A live workflow with no row in the automation inventoryThe single list of every active automated workflow, kept outside the platform, with its purpose, owner, trigger, exits, dependencies, measure and review dates. is easy to miss here, so compare the inventory with the platform's list of active workflows first. Look for more than one writer, changes that arrive in batches on the day of an import or a sync, and backward moves nobody can explain. Keep the list: later steps remove writers from it, and the test step checks that only the intended ones remain.
Write a rule for each stage from its exit criterion
For every stage, write in a sentence what sets it: the event that shows the previous stage's exit criterion has been met, the fields that must be true on the record, and the single workflow that makes the change. A record becomes a lead when it identifies itself and shows interest in something the business sells. It becomes marketing-qualified when it reaches the fit and intent threshold sales agreed to, which the lead scoring guide ties to the stage definitions. It becomes sales-accepted when a seller records acceptance, sales-qualified when the outcome of a first conversation is recorded, and an opportunity when a deal is created and associated with it. The subscriber stage needs care: the list building guide states from the GDPR that an organisation must be able to show that consent was given, and points to the ICO's direct marketing guidance for when consent is needed, so the rule that sets subscriber reads the consent record, not the mere presence of an email address. The customer stage already has its own workflow, which a later step connects rather than rebuilds.
Add a date field for every stage
Give each stage a field that records when the record first entered it, written once by the same workflow that sets the stage and never overwritten afterwards, the way the fundamentals guide treats original source. Add separate fields for the events that can happen more than once: the date and reason of the most recent recycling, and the date a recycled record qualified again. These dates are what a cohort report reads; without them, stage conversion is this month's arrivals at one stage against this month's arrivals at the next, which the lead stages guide warns is a ratio of two unrelated populations. Add each field to the dictionary that the tutorial on auditing and documenting CRM fields keeps.
Allow moves forward only, with named exceptions
Write the rule that a workflow may only move a record to a later stage than the one it holds, and list the few routes by which a record may go back: recycling for timing, budget or need; disqualification with a reason; the reversal of a won deal marked by mistake or cancelled; and a correction by a named person, with a reason and a log entry. Anything else is stopped by a condition inside each workflow: read the current stage before writing, and do nothing if the new stage is not later. That condition removes most of the damage an import, a sync or a stray workflow can do. Decide at the same time what happens to a record that skips a stage, such as a demo request routed straight to a seller: stamp the skipped stages with the date of the stage reached and flag them, or leave them blank and build every report to expect the gaps. Write the choice into the stage document.
Build the workflows that set each stage
Build a workflow for each stage, or a single workflow with a branch for each, named under the inventory's convention and given an inventory row before it is switched on. Trigger each on the change that every path into the stage makes, not on a single form; the diagnosis for a CRM workflow that is not running shows how a trigger tied to one path misses records that arrive by import or sync. Each workflow checks the forward-only condition, sets the stage, stamps the date if the field is empty, and stops. Where a separate marketing platform calculates the lead scoreAssigning points to leads for fit and behaviour to decide which are ready for sales., follow the sync guide: the marketing platform sets a request field of its own, and a CRM workflow makes the change by the CRM's own rules. The customer stage is the job of the closed-won workflow, which sets it with its date for every contact on a won deal and handles reversals; connect to it rather than adding a second writer. If companies carry a stage of their own, derive it from their contacts and deals by rule, using the records that the guide to matching contacts to company records keeps straight.
Make disqualification and recycling deliberate moves
The lead stages guide gives every disqualification a reason from a short 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. and lets the reason decide what happens next. Build that literally. A fit or junk reason moves the record to disqualified and stops its workflows, and the record is kept only as long as the retention rules allow, which the guide to data retention across marketing tools sets from the GDPR's storage limitation principle. A timing, budget or need reason moves the record back to lead, stamps the recycling date and reason, and enrols it in a nurture sequence written for that reason. Lost opportunities follow the same routes with the lost reason. Re-entry is a rule too: a recycled record can qualify again when its behaviour changes, while a record closed for fit cannot, however many pages it reads, which the scoring model enforces by scoring closed reasons down or capping them.
Decide what a returning enquiry does at each stage
A known contact who submits another form is not a new lead, and what should happen depends on where the record stands. A subscriber or lead moves forward if the action meets the next stage's criterion; its most recent source updates while its original source stays. A record that already has an owner keeps its stage, and the owner hears about the action through an alert, which the guide to sales alerts from marketing activity keeps for single events on records a seller already owns. A contact at a customer keeps the customer stage, goes to the account owner as the matching guide routes it, and stays out of prospect journeys, the failure the diagnosis for customers receiving prospect messages traces. A recycled record re-enters by its reason; a record closed for fit stays closed unless the action is a declared one, such as a demo or quote request, which the scoring guide routes whatever the score. Set each workflow's re-enrolment to match these answers, deliberately rather than by default.
Keep imports, syncs and manual edits from rewriting the stage
Remove every writer from the first step's list that is not a stage workflow. An import may set the stage only on the records it creates, never on a record it matches, as the importing tutorial requires. The sync carries the stage in one direction, from the CRM, as the sync guide's field table says. Workflows that write other fields must not set off the stage workflows in a loop, which the tutorial on standardising CRM field values prevents by triggering only on the raw field. Restrict direct edits of the stage to a few named people for corrections, with a reason recorded each time; sellers record their judgements in the acceptance, outcome and disqualification fields that the workflows read. If a value still changes back, the diagnosis for CRM values changing back after a sync finds the remaining writer from the record's history.
Backfill existing records and their dates
The records already in the CRM need the stage and dates the new rules would have given them, taken from evidence rather than guessed: customer from won deals, with the close date as the date entered; opportunity from open deals; sales-accepted and sales-qualified from recorded acceptances and conversation outcomes; marketing-qualified from the score history where it exists. Where nothing records a date, leave it empty, and flag every backfilled date as estimated so a report can include or exclude it. Run the backfill in tagged batches with the stage workflows paused, check how many existing records each workflow would enrol before it goes live, as the workflows hub describes, and keep history out of the journeys a stage starts. Record the backfill's date and rules in the Change LogA dated record of changes made to tracking, accounts, reports, lists or rules, with who made each and why, so later movements in the figures can be traced..
Test with seed records and read stage conversion by cohort
Walk a seed contact through every path: each stage reached in order, a skipped stage, a recycling and a re-qualification, a disqualification for fit followed by more reading and then a demo request, a returning enquiry at each stage, a reversed win, an import that matches an existing customer, and a change arriving from the marketing platform. Each should end at the stage and with the dates the rules say, with only the intended writers in its history. Keep the seed contacts as fixtures and repeat the walk after any change to the sync, the imports or the scoring. Then build the report the rules were for: the records that reached a stage within a period, followed forward, which the lead stages guide calls reading by CohortA group of customers who share a starting event and period, tracked together over time.. The conversion rate calculator converts every pair of counts to a rate, and the guide to reading small numbers explains why every rate should carry its counts. Mark the go-live date on every chart, and check on a schedule the team sets for records with a stage and no date, or dates out of order.
Which rule sets which stage
The table sets the rules side by side for a common set of stages. It is a pattern, not a prescription: names and stages follow the business's own definitions.
| Stage | Set by | From what evidence | Routes back | Date fields |
|---|---|---|---|---|
| Subscriber | The consent workflow | A consent record the organisation can demonstrate | None; a withdrawal is recorded in the consent fields and the history stays | Date subscribed |
| Lead | The enquiry workflow | A form or action that identifies the person and shows interest | Recycled records return here | Date became a lead; date last recycled and reason |
| Marketing-qualified (MQLA lead that marketing judges ready for sales follow-up against agreed criteria.) | The CRM workflow reading the score or the request field | The fit and intent threshold sales agreed to, or a declared action | Rejection by sales, with a reason | Date marketing-qualified; date qualified again |
| Sales-acceptedA lead sales has reviewed and agreed is worth working, recorded as a stage between marketing-qualified and sales-qualified with a reason when rejected. | The workflow reading the seller's acceptance | Acceptance recorded within the service level | Recycling or disqualification | Date accepted |
| Sales-qualifiedA lead that sales has reviewed and accepted as a genuine opportunity to pursue. | The workflow reading the conversation outcome | A recorded first conversation confirming need, a route to a decision and timing | Recycling or disqualification | Date sales-qualified |
| Opportunity | The workflow reading deal creation | A deal associated with the contact | Lost-reason routes | Date became an opportunity |
| Customer | The closed-won workflow | A deal won under the agreed definition | Reversal of a won deal only | Date became a customer |
| Disqualified | The workflow reading the disqualification reason | A reason from the picklist | Recycling, for timing, budget and need reasons only | Date disqualified; reason |
Separate pipelines for very different products each get their own stage document, as the lead stages guide advises, and the rules follow each one. Lifecycle stage describes the relationship with a person and deal stages a single sale; keeping them apart, as the glossary entry explains, lets a customer hold the customer stage while a new deal moves through stages of its own. The field's accuracy matters beyond the pipeline, because lifecycle is the first cut of a list in the email segmentation guide. Teams that hand over accounts rather than people adapt the qualified stages as moving from MQLs to account qualification describes; the dates and the forward-only rule apply unchanged.
A cold room manufacturer sets its stages by rule (illustrative)
Qelmarund Cold Rooms sets its stages by rule
Qelmarund Cold Rooms, an invented maker of modular cold rooms for restaurants, bakeries and food producers, has 4 sellers, a showroom coordinator and 11,200 contacts in its CRM. An export of 6 months of stage history shows 9,400 changes from 10 writers: 2 forms, a scoring workflow in the marketing platform, the sync, a nightly import from the quoting tool, and the 4 sellers and the coordinator by hand. 1,320 of the changes moved a record backwards, 870 of them on nights when the import ran. 214 contacts with a won deal still sit at lead, and the stage report shows 380 opportunities in the year against 290 sales-qualified leads, because 140 opportunities were opened from records that never passed through the earlier stages.
With the sales manager, the team writes a rule for each of its 8 stages, adds 10 date fields and lets workflows move records forward only, with recycling, disqualification, won-deal reversals and logged corrections as the only routes back. Skipped stages are stamped with the date of the stage reached and flagged. The scoring workflow now writes a request field for a CRM workflow to read, the import may set a stage only on the records it creates, and direct edits are limited to 2 people who pick a reason from a list.
The backfill moves the 214 contacts to customer from their deals, gives 2,630 records dates taken from deal and activity history and leaves 3,900 dates empty where nothing records them, each flagged as estimated or missing. 15 seed contacts walk every path before go-live. In the next quarter the history shows 31 backward moves, all through the named routes. The first cohort report built on the new dates shows that 41% of the leads created in its first month reached marketing-qualified and sellers accepted 63% of those, and that the largest loss sits between accepted and sales-qualified rather than at the threshold the team had been arguing about.
The workflows section places this tutorial beside the closed-won workflow, the scoring model and the nurture sequences that read the stages it keeps.
Common mistakes
- Automating stages nobody has defined. A rule can only enforce a definition both teams have agreed; write the stage document first.
- Letting two systems write the stage. The value becomes whatever was saved last; one system owns it and the other requests changes.
- Recording the stage without the date. A funnel built on current values cannot be read by cohort or rebuilt for a past period.
- Overwriting the first date when a record recycles. Keep the original entry dates and record recycling and re-qualification in fields of their own.
- Letting any workflow move a record backwards. Forward-only with named exceptions removes most of the damage imports and stray workflows do.
- Triggering on a single form. Records that arrive by import, sync or lead form never meet a stage rule that listens to one form.
- Backfilling by guesswork. Take stages and dates from deals and recorded outcomes, leave the rest empty and flag every estimated date.
- Switching on with history enrolled. A stage workflow activated on the whole database sends every old record through the journeys that stage starts.
