Skip to content

Marketing & Growth Systems

Personalization: use what you know, and only where it helps

“Hi {first_name}” isn't personalization; it's a mail merge waiting to break. Useful personalization changes what a message says because of something you know: the last service, the plan, the right next step. Each slot has a source, and when the source is empty it falls back to words that are still true.
Fall tune-up emailgreeting ← first namelast service ← service historynext step ← plan + seasonHi R.T.,Last tune-up: Oct 4 last year.Weekday slots are still open.Hi J.P.,Your plan covers this visit.We'll book it with your Nov 14 renewal.Hi there,fallbackMost furnaces need a check once a year.fallbackWeekday slots are still open.Book online or reply with a time.R.T.first name: R.T.last tune-up: Oct 4 last yearplan: noneJ.P.first name: J.P.plan: renews Nov 14plan covers tune-upsNew customerfirst name: —service history: —plan: nonenever usedpayment historyhousehold detailsguesses about the person

What to personalise, and from what

A few slots, each with a source someone owns and a fallback someone wrote.
GreetingFirst name on the record“Hi there”
Last serviceService historyA general line that's true for everyone
Next stepPlan status and the seasonThe standard offer
Nearest officeService addressThe main number, never a guess

One email, three customers

Emailto R.T. · tune-up due

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

Where personalization breaks

  1. 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.

  2. 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.

  3. 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.

  4. 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.

  5. 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

Customer facts
name, services, plan
Season and offers
what's current
Slot rules
source and fallback per slot
Personalization
fill each slot → or fall back

Out

The message
each slot's source logged
Fallbacks
counted, to show missing data
Versions
few enough to read

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.