Conversion rate: how to find your baseline and move it
What conversion rate actually measures on a store, how to establish yours by segment, and the changes that move it in order of return.
The short answer
Conversion rate is the share of visits that end in an order, and as a single number it hides everything useful. A store’s rate is an average across phones and desktops, cold social traffic and warm email, product pages and blog posts, and moving the average requires knowing which parts are low and why. Establish the baseline by device, by source and by landing page type, over a period long enough to smooth campaigns and seasons, and compare against yourself rather than against published averages. Then work the levers in order of return: mobile speed, product page clarity and trust, checkout friction, shipping and returns visibility. Change one thing per template at a time, measure against the baseline, and keep a log.
Building the baseline
| Cut | Why it matters | Typical finding |
|---|---|---|
| By device | Mobile is most traffic and usually the weakest experience | A large gap between mobile and desktop |
| By source | Intent differs: search, email, social, direct | Email and search convert; cold social browses |
| By landing page type | Where people arrive shapes what they do | Product-page landings convert; homepage landings wander |
| By new versus returning | Trust differs | Returning customers convert several times better |
| Over time | Campaigns and seasons distort short windows | Weekly cycles; seasonal peaks |
The levers, in order of return
- Speed on mobile. Every second on the product page and checkout costs orders; it is measurable and fixable.
- Product page clarity and trust: price and availability visible, the right images, honest reviews, shipping and returns stated, one clear add-to-cart.
- Checkout friction: guest checkout, express payment methods, address autofill, no surprises in cost.
- Shipping and returns visibility before the checkout, not inside it.
- Collection page findability: filters that match how people choose, sensible defaults.
- Search that tolerates typos and never returns nothing.
- Everything else: copy, banners, pop-ups, which usually move less than the items above and are tested more often.
Measuring a change
Pick one template and one segment. Record the baseline over a stable period. Make one change. Run long enough to cover full weekly cycles and enough orders for the difference to mean something. Compare against the baseline for that segment, not the store-wide rate. Log the change, the dates and the result, whether it won or lost. Over a year that log is the most valuable document in the store’s marketing, because it says what actually moves the number for these customers.
What this means for you
Stop watching one blended number. Build baselines by device, source and landing page, compare against yourself, and work the levers in order: mobile speed, product page clarity, checkout friction, shipping and returns visibility. One change at a time, measured long enough, logged. Conversion rate moves when the store gets easier to buy from on a phone, and everything else is a smaller number after that.
Frequently asked questions
What is a good conversion rate?
Better than your own last quarter, by segment. Published averages mix categories, price points and traffic sources so broadly that they say little about your store. A store selling considered purchases from cold social traffic will convert far below one selling repeat consumables to email subscribers, and both can be healthy. Build your baseline and improve against it.
Why is our mobile conversion so much lower than desktop?
Partly because mobile traffic includes more early-stage browsing, and mostly because the mobile experience is worse: slower pages, harder forms, galleries that do not swipe, checkout buttons below the fold. The gap between mobile and desktop conversion is one of the clearest signals of where to work, and closing it is usually the largest available gain.
How long should a test run?
Long enough to include full weekly cycles and enough orders for the difference to be meaningful, which in a small store is weeks, not days. Ending a test the moment one variant looks ahead produces conclusions that reverse the following week. Decide the duration and the minimum order count before starting.