How we measure it, so you can check our working.
Retention work is easy to over-claim. Every channel takes credit for the repeat order. So the method is built to be tested, and to be wrong in public where it has to be.
The Loyalty Profit Score
A score out of 100 across five dimensions, so you can see where the business is strong and where the money is leaking.
- Customer retention62
- Customer profitability48
- Customer data41
- Customer journeys55
- Commercial discipline64
An illustration only. It is not a real client's result.
Customer retention
How many customers come back, how soon, and what you do when they don't.
Customer profitability
Whether repeat customers are profitable after discounts, rewards and the cost of reaching them.
Customer data
Whether you can see a customer across your store, email platform and, for stores, the till.
Customer journeys
The messages and moments that follow a first purchase, from welcome to win-back.
Commercial discipline
Whether decisions about offers, budget and channels are made on numbers, and checked afterwards.
Five leaks, each sized in dollars
For each leak we show the formula, the inputs and the assumption, so you can change any of them.
Customers you never email again
Subscribers or past buyers with no follow-up. Sized from list size, a conservative response rate, and your average order value and margin.
Carts left behind
Abandoned checkouts with no recovery flow. Sized from cart volume, a conservative recovery rate and margin.
One order and done
Customers who never place a second order. Sized from your repeat rate against a stated benchmark, and the value of a second order.
Not knowing who your customers are
Missing data that stops you personalising. Reported as a gap and a plan rather than a made-up dollar figure.
Every message looks the same
Blanket discounts. Sized from the discount cost on customers who would have bought anyway.
For stores and venues we add a sixth: customer details that never get captured at the till or booking.
The rules we work to
- Fixed calculations. The same formulas for every client, so results are comparable and not tuned to flatter.
- Assumptions stated. Every dollar figure sits beside the assumption behind it.
- Margin-adjusted. We count profit, not revenue, and subtract discount cost.
- Starting figures recorded. In week one, before we change anything.
- A hold-out group. Where practical, a random 10% of customers don't receive the new messages. The difference is what the work produced.
- Reported in dollars. Opens, clicks and impressions are diagnostics, not results.
Estimates are not promises. Actual results depend on your products, margins, offers and how the plan is carried out.
Where AI fits, and where it doesn't
We use AI to do the heavy lifting: reading exports, spotting patterns, drafting. We don't use it to make the decision. A model can tell you which customers have gone quiet. It can't tell you what you can afford to spend to bring them back, or whether an offer suits your brand.
Identifiable customer lists are never pasted into public AI chatbots. We work under a written data agreement and delete files when the work ends.
For the same reason, a platform or CRM can report the numbers, but someone still has to make the commercial call. That is the part we do.
Questions
Why a hold-out group?
Because customers who were going to come back anyway will look like a win. Holding back a random slice shows what happened without the work, so the gap is real.
What if a hold-out isn't practical?
With small lists it can be too noisy. Then we compare against the recorded starting figures and say plainly that the comparison is weaker.
Do you guarantee results?
No. We state our aims, show our workings and report honestly against the figures we recorded at the start.
Want to see the method on your own numbers?
The Profit Leak Detector uses it in a couple of minutes.