Conversion rate optimisation is the discipline of turning more of your existing traffic into customers: understand what a conversion is, log it properly, then optimise the flow. I rate it the highest-leverage work in marketing, because every improvement compounds across all of your acquisition spend.
Define Your Conversions
Primary Conversion
The main action that drives your business:
- E-commerce: Complete purchase
- SaaS: Start paid subscription
- Media: Subscribe to newsletter
- App: Complete signup & first key action
Secondary Conversions (Micro-conversions)
Smaller actions that lead to primary:
- Add to cart
- Start free trial
- Download resource
- Watch demo
- Create account
Not to Be Confused With
🎯 Primary vs micro-conversions
The primary conversion is the action that drives the business; micro-conversions are the steps along the way. Optimise the primary, diagnose with the micro.
📝 Logging vs attribution
Logging records what happened; attribution argues about who gets credit. Get logging right first: I have seen teams argue for quarters over attribution models built on event data nobody had validated. Attribution built on bad logging is fiction.
🧪 Testing vs tweaking
A test has a hypothesis, a control, and a pre-registered read window. Changing button colours on a hunch while watching the dashboard is not testing.
The Three Essentials
LOG IT
Record every conversion event:
- User ID
- Timestamp
- Source / channel
- Device type
- Value
And define the conversion before you log it. At eBay we tracked confirmed registered users, then realised that was the wrong conversion: we switched to activated confirmed registered users: registered, confirmed, and having bid on, bought or listed an item. Defining the conversion is the logging problem.
ATTRIBUTE IT
Understand which efforts drove the conversion. But beware: last-click attribution is broken.
Use incrementality testing where possible.
MEASURE IT
Key metrics:
Map Your Conversion Path
▼ 40%
▼ 67%
▼ 40%
▼ 33%
Illustrative example: your numbers will differ, but the shape is always a funnel.
Optimisation Quick Wins
Testing Principles
Isolate each test
I run multivariate tests, but I never cross-contaminate: each experiment gets a clear control and variant.
Pre-register test length
Don’t peek and kill early. Require sustained significance over a pre-registered period.
Business significance matters
Know Its Limits
My position: optimisation has a ceiling, and the ceiling is set by your offer and your audience. You can test your way from 2% to 4%, but you can’t optimise your way out of a product people don’t want or traffic that was never qualified. When my tests stop producing winners, the problem is usually upstream: in the targeting or the offer itself.
Beware the local maximum, too. Relentless optimisation of the current flow can blind you to the fact that a completely different flow would convert twice as well. When I hit that wall, I stop testing the flow and redesign it.
And keep the business-significance rule: if you need a microscope to see the impact, you didn’t move the business.