Email that tests itself.

We write several versions of every email, learn which one your customers respond to, and send more of the winner. Automatically, every day.

Live experiment0 sent
  • A“Quick question about your spring orders”
    25% of sendsno replies yet
  • B“The 3 emails your best customers open”
    25% of sendsno replies yet
  • C“Still thinking it over?”
    25% of sendsno replies yet
  • D“Your account manager has a suggestion”
    25% of sendsno replies yet

A simulation with made-up reply rates, run by the same decision code as our real campaigns.

  • Works with your current email platform
  • You approve every version
  • Scored on replies and bookings, not opens

Why does most email stop improving?

A stack of identical envelopes
  1. 01

    One version of every email.

    Someone picks a subject line, it goes to the whole list, and nobody ever finds out what would have worked better.

  2. 02

    Tests that end.

    A/B tests need someone to set them up, wait, read the result and act. Most get run once and forgotten.

  3. 03

    The same email for everyone.

    Your best customers and your coldest leads get identical messages, so the email is tuned for nobody.

What if every email ran its own experiment?

It is A/B testing that never stops. A learning model (a multi-armed bandit, the same family of methods used to tune ads and recommendations) decides which version each person gets, and keeps adjusting as results come in.

Four envelopes in a row: three white, one black
Four versions go out. One earns the rest of the list.
  1. 01

    Write variants

    We write several versions of each email: different subject lines, angles and lengths, in your voice.

  2. 02

    Send a little of each

    Every send is a small experiment. Nothing goes out without your approval of the versions.

  3. 03

    Learn from real results

    Replies, clicks and bookings are the score, not opens. The model updates after every result.

  4. 04

    Shift and retire

    More of your list gets the winners each day. Clear losers retire themselves and new versions take their place.

Where is your email leaking?

The leakEvery shopper gets the same discount email, so you either give away margin to people who would have bought anyway or under-sell the ones who needed a nudge.

Version A
Subject
The email your team is probably sending to everyone

Hi Alex,

You asked about smarter email for your team. One pattern shows up in almost every list made of past buyers, abandoned carts and newsletter sign-ups: every shopper gets the same discount email, so you either give away margin to people who would have bought anyway or under-sell the ones who needed a nudge.

The fix isn't writing better emails by hand. It's letting your list tell you. Our system sends a few versions of each email, watches who replies, clicks and books, and moves more of your list onto whatever is working, every day, without anyone running the test.

For a business like yours I'd start with:

  • discount vs. no-discount win-back for lapsed buyers
  • product-led vs. story-led subject lines on new drops
  • send time per customer, learned from when each one actually opens

If you pick a time for a 20-minute call, I'll walk through what that would look like on your list: [link to pick a time]

Version B
Subject
Alex, which of these would you test first?

Hi Alex,

Thanks for asking about AI email for your team. If we switched on a learning campaign for you tomorrow, these are the three experiments I'd queue first:

  • discount vs. no-discount win-back for lapsed buyers
  • product-led vs. story-led subject lines on new drops
  • send time per customer, learned from when each one actually opens

Each one runs on its own: the system keeps sending the version that's earning replies and quietly drops the ones that aren't. You don't pick winners; your customers do.

Which one matters most to you? Grab 20 minutes and we'll go through it: [link to pick a time]

Two real follow-ups from our own campaign, written for your industry. The system sends both, then sends more of whichever one gets people to book.

Book a 20-minute call.

Tell us a little about your list and pick a time. Here's what you'll leave with:

  • Your list, read properly

    Which emails you send today, who they go to, and where the replies are coming from.

  • The first three experiments

    The specific tests we'd run first for your industry, and why.

  • A straight answer on fit

    If your list is too small or the timing is wrong, we'll say so.

An opened envelope with a letter
  1. 1. About you
  2. 2. Pick a time
No spam. You can unsubscribe from any email in one click.

What else do you need to know?

Do I have to switch email platforms?

No. We work with the platform you already use. The learning layer decides which version each person gets; your platform still does the sending.

Will it email my customers without me seeing it?

No. You approve every version before it can be sent. The system only decides which approved version goes to whom, and when.

How quickly does it learn?

It depends on how many people you email and how often. Bigger lists learn faster. On the call we'll estimate it for your list honestly, including if it's too small to be worth it yet.

What does it cost?

It's scoped to your list size and how many campaigns you run. You'll have a clear number after the call, with no obligation.

Is this cold email?

No. We work with people who already know you: customers, sign-ups, enquiries and leads who gave you their address. Every email carries a one-click unsubscribe.