Attribution when you cannot track everything, done honestly

5 days ago
3 min read

Updated: 3 days ago

Introduction

Somebody hears about you from a neighbour, looks you up two weeks later on a phone, forgets, sees a van, searches your name on a laptop and calls. No analytics package will reconstruct that, and no amount of tag configuration will make it appear. This is not a technical shortfall; it is what buying actually looks like.

The response is not to give up on attribution or to buy more tracking. It is to accept that you will get a partial, biased picture, understand which direction the bias runs, and use it to make decisions that only need to be roughly right. Most spending decisions do.

The businesses that measure attribution well are not the ones with the best tracking. They are the ones that ask every caller a question and write down the answer.

1. Attribution when you cannot track everything starts by asking

The cheapest method.

How did you hear about us, asked on every first contact, recorded in the same place. It is imperfect and it is still the best data most small businesses will ever have. It also catches everything offline.

2. Expect the answer to be vague

The realistic expectation.

People say online when they mean a search, and they say a friend when they mean a review someone mentioned. Take the vagueness as information rather than error. Record what they said, not what you think they meant.

3. Understand which direction the bias runs

The interpretation rule.

Trackable channels are over-credited and word of mouth is under-credited, systematically and everywhere. Your data will understate the value of reputation. Adjust your reading accordingly rather than adjusting the numbers.

4. Ask what made them get in touch today

The second question.

The source explains how they found you; this explains the timing. The two answers together are far more useful than either alone. It is also a natural thing to ask.

5. Count first touch and last touch separately

The two ends.

If you can identify both, keep both, and do not average them into a single credit. Awareness channels and capture channels do different jobs. Judging one by the other's metric is the commonest attribution mistake.

6. Use the totals when the details fail

The fallback.

If spend went up in one channel and enquiries rose the following month, that is evidence even without per-enquiry tracking. It is weak evidence and it accumulates. Several months of it is a reasonable basis for a decision.

7. Turn something off and watch

The strongest test available.

Pausing a channel for a month tells you more than any attribution model. It is uncomfortable and it is decisive. Do it in a quiet period rather than a busy one.

8. Do not attribute to the fraction of a percent

The proportionality point.

The decision is usually keep, cut or increase, and that needs an approximate answer. Precision beyond that is false and expensive. Stop measuring at the point the decision stops changing.

9. Write down what you cannot see

The honesty record.

Note the channels you know are working and cannot measure, so that they are not cut by someone reading the report literally in a year. One line is enough. This is the most valuable sentence in most funnel reports.

Be careful about attributing a long-cycle sale to whatever happened most recently. In a business where people deliberate for months, the last thing before the call is nearly always the least important thing in the sequence.

Conclusion

Ask every new contact how they heard about you and write the answer down verbatim.

Expect vagueness rather than trying to eliminate it, remember that trackable channels are systematically over-credited and word of mouth under-credited, ask what prompted them to make contact today as well as how they found you, keep first and last touch separate, use month-level totals when per-enquiry data fails, pause a channel when you need a real answer, and record in the report the channels you know matter but cannot see.