Setup together
Install the snippet, create goals, set up the first test, guided instead of alone with a manual.
?Support and documentation
Answers before you
have to ask.
The most common questions from real setups, sorted by topic. Whatever is not here, we answer directly.
- German and English
- From people who run tests themselves
- Setup is guided
The direct route
How to reach us.
No ticket maze. For setup and analysis you get someone who built the product.
Questions about results
Not sure whether a result holds? Send the report link and we will look at it together.
Technical questions
Theme quirks, consent managers, revenue tracking, API and MCP, there is someone with access to the code for that.
A named contact Coming
From the agency plan onwards, a named person instead of a general address.
Status page Coming
A public operational status for hosted installations, including maintenance windows.
Response times in the contract Coming
Guaranteed response times per plan, put in writing.
Documentation
Briefly explained,
by topic.
Each entry answers one question and says where the setting lives.
Setup
How do I install the snippet?
One script tag with your website ID goes into the <head> of every page you want to test on. The ready-made tag sits under Website setup → Installation; Shopify, WooCommerce, Shopware, Webflow and Magento have their own guides.
How do I know it works?
The setup check verifies before your first test whether the snippet reports in, whether goals fire and whether purchase events arrive. It also shows when the website last reported.
What does my own subdomain do?
Under Installation → First-party tracking you enter a subdomain of your own domain, for example t.your-domain.com. The snippet then loads from there, is blocked less often and measures more completely. What is collected stays the same.
How do I test on staging before going live?
Environments under Website setup → Environments. An experiment can be limited to one environment, and a preview link shows you the variation without any result being counted.
Building experiments
When the visual editor, when code?
Text, images, buttons and ordering work by clicking in the editor. As soon as logic is involved, conditions, calculations, inserted sections. JavaScript and CSS is the more stable route. Both can be mixed in the same variation.
How do I test two completely different pages?
As a redirect experiment: the variation redirects to a different path on the same website. The exposure is counted before the redirect and campaign parameters are carried over.
Can I reuse a change?
Yes, through reusable sections and asset management. Upload an image once and use it in several experiments.
How do I see the variation before it goes live?
Through a preview link that bypasses audiences and schedules and counts nothing. It expires after 24 hours or as soon as you change the draft; there is a QR code for your phone.
Measuring
What goal types are there?
Click on an element, form submission, page view, custom event and revenue. An experiment has one primary goal; everything else runs alongside as an additional goal or a guardrail.
How do I measure revenue?
Create a revenue goal and call SplitTest.track('Purchase', { value, currency, eventId }) on the order confirmation. Using the order number as eventId prevents double counting on reload. The snippet alone captures no orders.
What are guardrails for?
For metrics that must not get worse while you improve something else, returns, support requests, drop-offs. abdriven reports when a winner is doing damage elsewhere.
Does the data reach GA4 or Matomo too?
abdriven can forward the exposure to tools already present on the page. GA4, GTM, Matomo, Clarity, Hotjar and others. The switch enables forwarding; it does not install a tool or connect an account.
Analysing
How many visitors do I need?
The sample size planner works it out from your current conversion rate and the smallest difference you want to detect. Progress then sits next to the result.
What does the interval next to the result mean?
It says how large the difference plausibly is. If zero is inside it, you have no result, however good the average looks. With several variations abdriven adjusts the interval, because otherwise the sheer number of comparisons produces winners.
What is the traffic split check?
A comparison of the planned and the actual distribution. A clear deviation means something is wrong with delivery, and the result should not be used.
How do I pass a result to a client?
As a report link: read-only, no login, aggregated numbers only, with an expiry date and revocable at any time. For your own analysis there is also CSV export.
Technical and operations
Does it work in a single-page app?
Yes. The SDK follows pushState and back/forward, reverses variations that no longer match, and waits for an element signalling that the new view is ready. Hash routes such as #/page are not followed.
How does abdriven hook into the consent manager?
With data-consent="required" the snippet stays quiet until your banner calls SplitTest.consent(). If you load the snippet only after consent anyway, leave the attribute out.
Can I automate abdriven?
Through API tokens for your own scripts, webhooks and Slack for notifications, and the MCP server if AI agents should create experiments or summarise results.
How do I back up the data?
Through the built-in backup in the monitoring area or through SQLite’s online backup facility. The database belongs outside the public directory, and automatic off-site copies are not built in yet.
Still open?
Better to ask once
too often.
A badly set up measurement costs more time than any question.