Measuring the consideration stage, which leaves almost no trace

5 days ago
3 min read

Updated: 2 days ago

Introduction

Every other stage produces a record. Awareness has proxies, enquiries are logged, quotes are dated, orders are invoiced. Consideration produces nothing: the people in it are anonymous, they take no action you can attribute, and the ones you lose leave no trace at all. It is the least measured stage and, in many businesses, the one where most of the loss occurs.

There is no clean solution. What exists is a set of partial indicators, each weak alone and reasonable together, plus one direct method that most businesses never use, which is asking people. Combining them produces enough to work with, provided nobody mistakes the result for precision.

1. Measuring the consideration stage means accepting proxies

The starting position.

You cannot count people who never identify themselves. Every measure here is an inference, and the discipline is choosing a few, tracking them consistently, and not over-reading them. Four indicators followed for two years beat twelve followed for a quarter.

2. Watch the ratio of visitors to enquiries on key pages

The most available indicator.

Enquiry rate on your service and contact pages. A falling rate with steady traffic suggests something in the consideration stage has changed, even if it does not say what. Check whether the traffic composition changed before concluding the pages got worse.

3. Track how far people get before leaving

The behavioural signal.

Visitors who read three pages and leave were considering and were not persuaded. Visitors who leave from the first page were probably never relevant. Distinguishing them tells you which problem you have.

4. Ask new customers what nearly stopped them

The direct method.

"Was there a point where you nearly went elsewhere?" produces specific, actionable answers and almost nobody asks it. Ten answers are worth more than a quarter of analytics. Ask it in the first week, while the decision is still fresh in their memory.

5. Ask lost enquirers what decided it

The harder half.

A short, non-defensive message to somebody who chose otherwise. The response rate is low and the answers are the only direct evidence about the comparison you will ever get.

6. Watch the questions that recur at first contact

The gap indicator.

Anything asked repeatedly is something your material failed to answer during consideration. The frequency of each question is a rough measure of how many people that gap costs you.

7. Count the enquiries that arrive already decided

The positive indicator.

People who contact you saying they have read your material and want to proceed are evidence that the consideration stage is working. A rising share of these is one of the better signals available.

8. Compare your enquiry rate against periods when you changed something

The experimental approach.

Add prices, add a process page, add reviews — and compare the enquiry rate before and after over a quarter. Crude, slow and the closest thing to evidence available at this scale.

9. Accept the limits and use it anyway

The honest framing.

None of this measures consideration. It indicates. A business that tracks four weak indicators consistently will still see changes it would otherwise have missed entirely.

Be careful about optimising the numbers you can see at the expense of the ones you cannot. Removing prices would raise enquiry volume and worsen the outcome, and a measurement regime that rewards that is worse than none.

Conclusion

Use several weak indicators consistently and ask people directly.

Track enquiry rate on commercial pages rather than site-wide, distinguish visitors who engaged and left from those who bounced immediately, ask new customers what nearly stopped them, contact lost enquirers for the reason, treat recurring first-contact questions as gaps your material failed to fill, count enquiries that arrive already decided as a positive signal, compare enquiry rates before and after deliberate changes, and be honest that all of it is inference.