Deciding how many funnel stages to track, and why fewer wins

3 days ago
3 min read

Introduction

Published funnel models come with four, five, six or seven stages, all confidently labelled, and none of them derived from your business. Adopting one wholesale is how a business ends up tracking a stage it cannot define, populating it with guesses, and abandoning the whole exercise within two months.

The number of stages is a practical decision rather than a theoretical one. Each stage has to be countable from something that actually happens, and each one has to be maintained by somebody. More stages mean more resolution and more upkeep, and the point at which the upkeep exceeds the insight arrives earlier than most people expect.

1. Deciding how many funnel stages to track starts with what you can count

Countability is the constraint.

A stage you cannot populate from a record is not a stage; it is a label. If nothing in your systems marks when someone entered it, either create that record or delete the stage. Most abandoned funnels contain at least one of these, and it is usually the stage that sounded most strategic.

2. Four is enough for most small businesses

The default recommendation.

Aware, enquired, quoted, bought. These map onto records almost every business already keeps, they can be counted in an hour, and they are sufficient to locate a leak. The awareness figure will be a rough proxy rather than a real count, and that is acceptable as long as the proxy stays consistent.

3. Add a stage only where you suspect a problem

Resolution follows evidence.

If the enquiry-to-quote drop is large, split it: enquiry, qualified, surveyed, quoted. The extra detail is earned by a question you actually have, rather than added in advance because a model has five boxes.

4. Longer sales cycles justify more stages

The main exception.

A nine-month business-to-business sale genuinely has more distinguishable steps, and collapsing them hides where deals stall. Six or seven can be right there, provided each is still an event.

5. Every stage needs an owner

The maintenance cost.

Somebody has to move records into and out of it. Stages nobody owns fill with items that entered and never left, which makes the counts meaningless within a quarter. Ownership is also what makes the review meeting productive, because there is someone who can explain the number.

6. Watch for stages that never lose anyone

A sign of redundancy.

If ninety-eight per cent of a stage progresses, it is not a decision point and it is not telling you anything. Merge it into the neighbouring stage and reduce the upkeep.

7. Watch for stages everybody sits in

The opposite sign.

A stage holding hundreds of records that neither progress nor close is usually two stages, or a stage with no exit rule. Both are worth splitting or defining rather than tolerating.

8. Keep the same stages across quarters

Comparability is the point.

Redefining stages every few months destroys the ability to compare, which is most of the value. Change them deliberately, note the date, and expect a break in the series.

9. Test the definitions on two people

The check.

Give the same ten records to two people and ask which stage each is in. Disagreement means the boundaries are not events, and the counts are opinions rather than measurements.

Be careful about mapping stages to internal departments rather than to buyer behaviour. Stages that describe who is handling something tell you about your organisation; stages that describe what the buyer has done tell you about the sale.

Conclusion

Track the fewest stages that still locate a problem, because every stage has an upkeep cost.

Start with four that map onto records you already keep, add resolution only where a drop tells you to look closer, allow more stages where the sales cycle is genuinely long, give every stage an owner responsible for moving records through it, merge any stage that never loses anyone, split or define any stage that everyone sits in, keep the definitions stable so quarters remain comparable, and test each boundary by asking two people to classify the same records.