Which Ad Did My Customer Come From?

The phone rang, the sale happened. So which ad brought that customer in? The question sounds simple, but the answer is usually “not one ad” — and knowing that changes how you split your budget.

The real journey isn’t one step

A typical purchase runs like this: someone sees an ad on Instagram, is interested, does nothing. A few days later they search your company name on Google. They land on your site, look at prices, leave again. A week later they come back directly and fill in the form.

So: did that customer come from Instagram, from Google, or from “direct”?

All three are defensible, and the answer you get depends on which panel you ask. That doesn’t mean your measurement is broken — attribution is precisely this question of how to divide the credit.

Why the panels contradict each other

Every platform sees its own contribution and naturally credits itself. Meta sees its impression, Google sees its click. If the conversion totals across two panels exceed your actual number of sales, that’s why — the same sale is being counted in both places.

This is why adding up platform panels and calling it “total sales” gives the wrong answer. I covered why GA4 and Google Ads numbers differ on a separate page.

Why last-click misleads

By default, most reports credit the last click. In the example above, that means crediting “direct” or “brand search” — so the Instagram ad that introduced the customer scores zero.

The result is predictable: the campaigns that do the introducing look like they “don’t work”, get switched off, and a few weeks later brand searches drop too. Because the thing feeding them was turned off.

What you can do

Choose the attribution model deliberately. Last-click isn’t the only option; there are models that spread credit across the journey. Attribution model optimisation sets this according to how long your customers take to decide.

Ask one question on the form. A “how did you hear about us?” field gives you something no panel can. When it contradicts the panel data, both are still worth having.

Don’t base a shut-off decision on one panel. Pausing a campaign and watching what happens to total sales is a more reliable test than what any panel reports.

Fix measurement first. If calls and WhatsApp messages aren’t counted, the attribution debate is academic — conversion tracking comes first.

Want me to map your journey? I can report which channel introduces customers and which one closes them. Get in touch.

Frequently Asked Questions

Can I see where each individual person came from?

No — panels don’t give person-level data. Everything you see is aggregated; I covered those limits on the who is clicking my ads page.

Is “direct traffic” really direct?

Usually not. Visits whose source was lost land there too: in-app taps, copied links, some email clients. A rise in direct traffic is normally a sign of measurement loss.

Isn’t asking the customer enough?

It’s valuable but not sufficient on its own: people rarely remember the first touch, and tend to name the last one they recall. It works best alongside panel data.