Build it
Make it sell
Keep it running
Conversion optimisation
The cheapest customer is the one already on the page.
Cross-selling, bundles, abandoned carts, reviews, subscriptions. Raising what an existing visitor is worth is almost always cheaper than buying more of them — and it keeps working after the advertising budget stops.
An unpopular truth
Your shop probably cannot A/B test anything.
A test needs enough orders to tell a real difference from luck. At twenty or fifty orders a month that takes years, and by then the shop, the range and the market have all moved. The industry sells testing anyway because it sounds scientific. We do the things already known to work, in the order of what they are worth, and we measure the outcome against revenue rather than against a test that never finishes.
In order
What we do first, second, and only then.
Everything below works. The order is the part worth paying for.
Group one
Worth the most
If any of these are missing, nothing below matters yet.
- Product pages that answer the questions which stop a purchase — how that works
- Checkout without a forced account
- Delivery cost and date shown before the cart, not inside it
- Abandoned carts recovered by one well-written email, not five
Group two
Worth doing
Real gains, once the first group is in place.
- Cross-selling built from your actual range, not a plugin guessing
- Bundles where the combination genuinely makes sense
- Reviews collected automatically, shown where the decision happens
- Subscriptions for anything people rebuy on a schedule
Group three
Worth doing eventually
Good things. Not the ones to start with.
- Search that still finds the product when the spelling is wrong
- Wish lists and saved carts for considered purchases
- One-click reordering for returning customers
- A loyalty scheme, but only if you would actually run it
If somebody sold you the third group before the first was finished, that is worth knowing about them. It is also the most common thing we find when we look at a shop that has already had help.
Measurement
How we know it worked.
Not with a test. With before and after, measured on revenue, over a period long enough that one good week does not look like a result. That is slower than a dashboard and considerably more honest than one.
Which means the tracking has to be right before anything else happens: conversion tracking that reports actual order value rather than counting form submissions, funnel steps that match your real checkout rather than a template, and a report that can tell a shop that got better from a month that had a public holiday in it.
And we say so when we cannot prove it. Some changes are obviously right and impossible to attribute — putting the delivery date above the button, for one. We do those because they are right, and we do not claim the credit when the number moves afterwards. An agency that attributes every good month to its own work is telling you something about itself.
Which of the three groups are you in?
Send us the shop address. We will tell you which of the first four are missing, roughly what each one takes, and whether the second group is even worth discussing yet.