Skip to main content

How do I track where my customers come from?

Prepared by the Flawian team · Reviewed .

Give each enquiry one row and list its sources there, so a person naming two sources counts once. In the example, three enquiries gave four mentions. A booking link shows the route used to make contact, which may differ from how the person first heard of you. A source record does not prove what would happen without it.

Principle and explanation

A customer can meet your business in several places before getting in touch. This example shows why counting those places as separate customers gives the wrong total, and why a recorded source is not proof that the customer would disappear without it.

A first mention and a recorded route answer different questions

The first source a person recalls, every source they mention and the route your own record shows are separate columns. Keep an unknown value when evidence is missing. Do not replace a customer's answer just to make it agree with a link report.

Attribution means assigning credit to interactions along a journey. Google Analytics describes different rules for that assignment. This small handwritten record does not reproduce its models; it simply preserves the observations before you choose how to summarise them.

A source record does not show what would happen without that source

If the shop pauses leaflets and next week gets two enquiries instead of three, the difference alone does not prove that the leaflets caused the lost enquiry. Weather, demand, the days observed and earlier leaflets can differ too. One changed activity makes the comparison easier to interpret; it does not remove these other explanations.

Google Ads distinguishes conversions assigned to ads from additional conversions estimated with a treatment and control group. The distinction matters here even without those tools: recording a contact route and establishing the effect of removing it are different tasks. A small before-and-after count is a clue to investigate, not a guaranteed causal result.

Keep the total beside the source breakdown

Use the same enquiry rule and period when comparing records. Show the total number of distinct enquiries, unknown sources and later buying outcomes separately. If a channel attracts fewer mentions after a change, check the overall total too: people may simply use another route to reach you.

Worked example

Three people describe different parts of their journey

A fictional repair shop records three new enquiries during one week. Each person has one enquiry ID. When someone mentions more than one source, the owner keeps both answers on that same row.

The leaflet was named first twice: by A and B. It was mentioned at all twice. Search was named first once, by C, but mentioned twice, by A and C. There are four source mentions and three enquiries. Adding the mentions would count A twice.

The booking records show search links for A and C. That does not make A's leaflet answer false: the leaflet may explain how A first heard about the shop, while the link records the later route used to make contact. B's phone call gives no independently observed earlier source.

None of these three people has yet agreed to buy. The table therefore counts enquiries, not new customers, completed repairs or revenue.

  • Invented records from one week, with three distinct enquiries and no missing rows in this example.
  • The remembered journeys are reports, not independently verified histories. A link shows the recorded route, not every earlier interaction.
  • No purchase, comparison group or estimate of additional sales is supplied.
One row per new enquiry; remembered sources and observed routes remain separate
EnquiryWhat the person says, in orderWhat the shop can observe
ALeaflet, then searchBooking through a search link
BLeafletPhone enquiry; earlier route unknown
CSearchBooking through a search link

Common misconceptions

Four source mentions mean four potential customers.
There are three distinct enquiries. A mentioned two sources; neither mention creates a fourth person.
The booking link proves the leaflet did nothing.
For A, the remembered first contact and the observed final route can both be true.
Enquiries fell after the pause, so the paused activity caused the fall.
The two weeks can differ in several ways. This record supplies neither a controlled comparison nor enough information to isolate the cause.

First step

Rebuild one enquiry before judging a source

Give it one ID. Write the remembered first source, other mentioned sources and the independently observed route separately. Mark what is unknown and whether it became a customer. Before changing an activity, record which total you will compare, over what period, and what else could explain a change.

Sources