Marketing & Growth Systems
Personalization: use what you know, and only where it helps
What to personalise, and from what
One email, three customers
Hi R.T., your last furnace tune-up was on October 4 last year. Weekday slots are still open this month. Book online or reply with a time.
3 of 3 slots filled
Hi J.P., your plan covers this visit. We'll book it alongside your November 14 renewal so it's one trip. Book online or reply with a time.
3 of 3 slots filled
Hi there, most furnaces need a check once a year, and fall is the time to do it. Weekday slots are still open this month. Book online or reply with a time.
2 slots fell back: no name, no service history
Where personalization breaks
01
Empty fields
“Hi ,”
Every slot needs a fallback written by a person and tested with the field empty.
previews checked only with complete records.
02
Wrong facts
“Last service: 2019”
A slot is only as good as its source. Old or wrong data is worse than none.
fields nobody owns.
03
Too many versions
Every combination
Four slots with three options each make 81 versions nobody has read. Keep slots few and independent.
slots that depend on each other.
04
Uncomfortable detail
“How did they know that?”
Use what a customer expects you to know from working with them. Never payment trouble, household details or guesses about them.
data bought from elsewhere.
05
Untested gains
Relevant isn't proven
If a slot matters, test it against the general version before building more.
personalization added because it's possible.
What personalization hands off
In
Out
Worth it when you know something useful
Good fit
- Your records hold service history, equipment or plan status.
- Customers would expect you to remember their last visit.
- General messages are being ignored.
Not the right fit
- Records are too sparse to fill slots reliably. Start with data quality.
- One message is right for everyone.
Questions about personalization
Is using someone's first name personalization?
Barely. It's the easiest slot and the one most likely to break. The useful slots change what the message says, like the last service or the plan.
Can AI write a different message for every customer?
It can, and then nobody has read what was sent. We prefer a few slots with written fallbacks. If AI drafts a slot, it works from approved facts and a person reviews the patterns.
What data should we never use?
Anything a customer wouldn't expect you to know from working with them: payment problems, household details, or guesses about age, health or income.
Does this work on the website too?
Yes, with the same rules: a known source, a fallback and a reason. A signed-in plan member can see plan pricing; everyone else sees the standard page.