Spreadsheet on one screen. Website copy on the other.
That was the last two weeks.
No launch. No new funnel. No clever agent doing something nobody's seen before.
Just checking data points... then changing things because of what the numbers said.
The Briefing
Meta has a new conversion setting on by default, and it starts counting November 12.
It's called "Measure conversions on customer devices." It skips the EU and the UK, sites with no login, and pixels sitting inside an iframe.
Your move: go and look in Events Manager before the 12th. When your numbers shift next month you'll want to know it was the setting, not your copy.
Meta switched product browsing on inside your live ads.
Went live October 28. A tap on your ad can now open a Facebook or Instagram product experience before it opens your website. Meta's own test (August 6 to 21, around 117,700 advertisers) showed 3% more conversions.
Your move: open Advantage+ creative and your Destination settings, then decide whether you want that step sitting between the click and your page.
AI is quoting you without saying your name.
Kevin Indig, working with Semrush data, October 29: 61.7% of brand appearances in AI answers were citations with no brand mention. Only 13.2% got cited and named.
Your move: one small messaging shift. Put your brand name in the same sentence as your claim, not in the byline.
Shoppers arriving from AI tools are worth more than the rest.
Adobe, September US retail data reported October 29: 83% more revenue per visit and a 90% higher conversion rate than other visitors. AI referral traffic up 71% year on year.
Your move: split AI referrers into their own channel before the holiday run, so by December you've got a number instead of a hunch.
The Build
Three things got changed this past fortnight. Website copy, a handful of email sequences, and a couple of small shifts in how the offer gets described.
Nothing on that list is new. Every item on it got checked against data before a single word moved.
Most people think growth comes from the next big swing. The new channel. The clever agent. The thing nobody's doing yet.
But...
What actually moved this week was dull. Open the sheet. Read what the numbers say about one page. Change one line because of it. Do the next page.
GROWTH NEEDS DATA AND A LITTLE BOREDOM MIXED IN.
Not every success story is exciting while it's being built. Sometimes it's just the right work, done the way it's meant to be done.
Here's the prompt. The trick is that it refuses to let the model start writing:
Here is one page (or one email sequence) and the
data behind it.
COPY:
[paste the copy exactly as it is live]
DATA:
[paste the numbers: traffic, conversion rate, opens,
clicks, replies, the date range, the sample size]
Do not rewrite anything yet. Work in this order:
1. List what the data actually supports changing.
For each one, quote the specific number you are
using and say what it tells us.
2. List what you would change on instinct but the
data does not support either way. Label these as
guesses and keep them in their own list.
3. Tell me where the sample is too small to conclude
anything. Say so plainly instead of handing me a
recommendation anyway.
Then, for list 1 only, write the change.
One change per item. Show me the before line and the
after line, nothing else.What didn't work: the model wants to rewrite everything. Show it a page and it hands back a whole new page, with the evidence-backed changes sitting right next to pure preference, and no way to tell which is which. Ask it for a verdict on thin data and it'll give you one, confidently, off forty visitors.
What worked: doing the work line by line. If you hand AI a page, it edits everything. And it backs itself. Goin g line by line and doing the edits yourself, creates results.
When did you last change a page because a number told you to, instead of because you'd got sick of looking at it?
The Play
If you're the itchy type, you already know. No patience. Always half-looking for the next thing that feels like momentum.
This one's for you, and it's going to feel like nothing.
Pick the one page or sequence closest to your revenue. Not the fun one. The one that already gets traffic and already converts, so the numbers mean something.
Pull its last 7, 14, 30, 90, 180 days. Run the prompt above. Ship only what's in list 1.
Then next week, do the next one... and the week after, the one after that. Every change should be backed by a number. That's a change that compounds instead of an exciting story you can describe at a meeting.
Try this today. Open your highest-traffic page, pull its last 180 days of numbers, and run the prompt. Ship only the changes with a figure sitting behind them, and park the rest in the guesses list for next month.
James
P.S. The exciting work and the work that compounds are rarely the same work. One of them you'll want to do. The other one you'll have to put in the calendar.
Growth Partner for founders stuck between $1M and $20M | I find the money you're sitting on and build the machine that collects it | $1B+ in client results.