Pages by path
Exact path, path starts with, or all pages. Only the path is compared, no domain, no ?query, no #anchor.
TATargeting
The right test
in the right place.
A test on every page measures nothing in particular. abdriven lets you say exactly where an experiment runs, who takes part, and when it stops.
- Pages by path
- Audiences with AND/OR
- Schedule and environments
The building blocks
Six controls, and you
need no more.
Targeting should sharpen the experiment, not become a second job. So there are few controls, and they explain themselves precisely.
Device and language
Mobile only, desktop only, or only visitors with a particular browser language.
Campaign and referrer
Rules on query parameters such as utm_source and on the domain the visitor arrived from.
AND or OR
Several conditions must all apply, or one is enough. You decide per experiment.
Schedule and environments
Set a start and an end, and decide whether a test runs only on staging or live.
Traffic share
Not every test has to run 50/50. A risky idea gets 10% first, and more later.
Staying clean
Two tests on one
page do not work.
When two experiments change the same page, neither of them is measuring its own effect any more. abdriven detects overlapping page rules and will not let the second test start there.
- Overlaps are checked at launch, not discovered later in the results.
- The preview bypasses audiences and schedules, so you see your variation even when you do not fit the target yourself.
- The traffic split check reports when the actual distribution deviates from the planned one.
- A winner kept live retains pages, audiences and environments, only the schedule no longer applies.
Afterwards
Segments tell you
who it worked for.
Targeting decides who takes part. Segments show afterwards who the variation worked for, without you having to set anything up in advance.
- Broken down by device, traffic source, and new versus returning.
- Recorded once at first contact, with no identifier alongside.
- Meant as exploration: cells with fewer than 100 visitors per arm are not compared at all.
- A segment finding is a hypothesis for the next test, not a result in itself.
Honestly
What targeting here is not
- Audiences are enrolment filters in the browser, not access control. Anyone calling the page directly still sees it, just without the experiment.
- There is no geographic targeting by country, because that would mean evaluating the IP address. Language and campaign work without it.
- Page rules compare the path only. Hash routes (#/page) are not followed, and a trailing slash makes a difference.
Coming
Already planned
None of this is built yet. It is here so you know what you can rely on, and what you cannot rely on yet.
Country and region Coming
Geographic targeting resolved on your own server, without storing the IP address.
Your own conditions in JavaScript Coming
Audiences from your data layer: cart value, customer group, login state.
Target returning visitors directly Coming
Use new versus returning not just for analysis, but as a condition for taking part.
Mutually exclusive test groups Coming
Several experiments on the same page, cleanly separated through assigned visitor groups.
FAQ
Targeting
Can I target by country?
No, that would mean evaluating the IP address, which abdriven deliberately does not process. For most cases, browser language and campaign parameters are enough.
What happens to visitors who do not fit the target?
They see the normal page and are not counted. They appear neither in the result nor in your booked traffic volume.
Can I ramp a test up slowly?
Yes, through the traffic share. Just bear in mind that a small share takes correspondingly longer to reach the planned sample size.
Product
More of the product
Make your next change an informed one
Less guesswork.
More now we know.
One question is a good place to start.