The output

A weekly summary from a small account.

Not a big one. Deliberately — if you are doing forty demo requests a month, you want to know what this looks like at forty demo requests a month.

Sourced weekly summary18 answers from 31 responses · week of 20 July
Worth knowing

Five people named ChatGPT this week. All five landed as (direct) in your analytics. It is your top named source for the third week running, and your analytics has recorded zero referrals from it.

Where they came from
  • ChatGPT5
  • Lenny's Podcast3
  • A colleague or friend3
  • Google2
  • MeasureSlack1
  • Demand Curve newsletter1
  • Couldn't say3
What they asked the AI

how do you find out where your direct traffic actually comes from

cheaper alternative to the attribution tool we already pay for

Why they came — in their words

we're rebuilding our whole reporting stack and I can't defend the numbers

direct is 60% of our traffic and my CEO keeps asking what's in it

trying to work out if the podcast sponsorship did anything

someone in a slack group said you were cheaper than the alternative

Identified leads
  • priya@example.comChatGPT
  • d.okafor@example.comLenny's Podcast
Illustrative example. The layout, volume and answer quality here are representative rather than one customer’s data, and the addresses are fictional. We would rather show you a thin week honestly than a busy one that is not yours.

Note the denominator

Thirty-one people saw the question. Eighteen gave an answer worth keeping. That is the number most attribution tools do not print, and it is the one that tells you whether to believe the rest of the page.

It also tells you something the top-line list does not. The people most likely to answer are the ones who came through a channel they remember — a warm referral answers more readily than an ad click does. So read the list as evidence rather than as a census. It is a truer picture of your dark channels than you have today, and it is not the whole picture.