One funnel or several by customer type, and how to decide

5 days ago
3 min read

Introduction

Most businesses serve more than one kind of buyer. A builder does insurance work and private extensions. A consultancy sells small diagnostic pieces and long programmes. An online retailer sells to consumers and to trade accounts. These groups find you differently, decide differently and convert at different rates, and measuring them together produces an average that describes none of them.

The question is when splitting is worth the extra work. Every additional funnel is another set of numbers to maintain and another report to read, and in a small business that cost is real. The test is not whether the customers feel different; it is whether the numbers behave differently enough to warrant separate decisions.

1. One funnel or several by customer type is answered by the conversion rates

The empirical test.

Split the last ninety days by group and compute each stage rate. If the rates are within a few points, one funnel is fine. If one group converts at three times another, you have been managing an average that exists nowhere.

2. Different entry points usually mean different funnels

The strongest indicator.

A group arriving by referral enters at a later stage than one arriving from an advertisement. They are not the same process with different results; they are different processes, and forcing them together makes the referral funnel look magical.

3. Different cycle lengths force a split

A measurement necessity.

If one group decides in a week and another in six months, a blended funnel cannot be measured on any sensible period. The fast group's results arrive before the slow group's have begun.

4. Different economics justify separate management

The commercial argument.

A group with twice the average order value and half the servicing cost deserves its own view, because decisions about where to spend acquisition budget depend on it. A blended cost per acquisition sends money to the wrong place.

5. Start with two, not five

Proportionate ambition.

The most useful first split is usually the obvious one: referral versus everything else, or trade versus retail. Two funnels are maintainable and will produce most of the insight that five would.

6. Split by behaviour, not by demographics

The choice that determines usefulness.

How they found you, what they bought, how long they took, what they spent. These predict conversion. Sector, size and location are frequently proxies for nothing in a small business.

7. Keep the stage definitions identical across funnels

Comparability requires it.

If "qualified" means something different in each, the funnels cannot be compared and you lose the main benefit of splitting. Same boundaries, different populations.

8. Watch for a group you are systematically failing

The most valuable finding.

Splitting frequently reveals one segment converting far below the rest, which was invisible in the average. That is either a targeting problem or a service gap, and both are actionable once seen.

9. Merge them back if the difference disappears

Maintenance in both directions.

Segments converge as a business changes. If two funnels have tracked within a few points for a year, merge them and reclaim the reporting effort. Splits should be earned continuously, not permanently.

Be careful about splitting so finely that each funnel has too few records to mean anything. Twenty enquiries a quarter cannot support a stable conversion rate, and noise presented as segmentation is worse than an honest average.

Conclusion

Split when the rates genuinely differ, because a blended average describes neither group.

Compute stage rates by group before deciding, treat different entry points and different cycle lengths as strong reasons to separate, split where the economics differ enough to change spending decisions, start with two funnels rather than five, segment on behaviour rather than demographics, keep identical stage definitions across funnels so they stay comparable, look for the segment you are systematically failing, and merge funnels back when the difference disappears.