From the Pulse plan
Try a change on part of the visits before you show it to everyone. Change texts, buttons and pictures in a visual editor, or test two pages or your own code. Visitics evaluates the goal, guards errors and speed, and shows the winner with its probability.
What it is for
Instead of guessing which version sells more, half of the visits sees each and the result shows which one wins — or that there is no difference.
Guardrails watch JavaScript errors, frustration and speed of every variant and pause a test that breaks the page.
Fields where people give up, dead clicks or a leaking funnel step become test ideas with a hypothesis.
How it works
The running tests with their progress, the drafts and the library of finished tests with their lessons — and ideas from your data.
Open your page, click an element and change its text, style or picture, or hide it — without code. The variant can be checked on a phone too.
The goal of each variant with the probability that it is better, the guardrails and the effect on forms, feedback and the chatbot.
Screenshots of the Visitics dashboard with demo data.
In detail
A small script in the page head asks our server which variant the visit gets. The choice comes from the same daily fingerprint as the measurement, so nothing is stored in the browser and a visitor sees the same variant all day. Changed elements wait at most 600 ms; the rest of the page is drawn at once.
Visual tests change texts, styles, pictures and links or hide elements, up to 40 changes. Page tests send part of the visits to another address. Code tests (from Aura) run your own CSS and JavaScript. A test has up to three versions, the original included, and can be limited to pages and devices.
The chance that a variant is better comes from a Bayesian model with 20,000 simulations; a winner needs 95 %. The verdict waits for the planned sample, at least seven full days and 25 conversions in every variant — only an overwhelming difference may end a test earlier.
When a variant has three times as many JavaScript errors or twice the frustration of the original, the test pauses by itself. A chi-square check warns when the split of the visits is off, and a test whose changed element is missing on more than a fifth of the pageviews is marked faulty.
A goal is an event, a submitted form, a page or an engaged visit. Revenue per visit uses order values capped at the 99th percentile, so one huge order does not decide. Before the start a calculator shows how many visits and days the test needs.
Every variant can be checked on your live site through a preview link, also by a colleague without an account, for seven days. The visual editor opens your page through a link valid for eight hours.
Everything it does
Privacy
Plans
The values come straight from the price list. Every new organization tries Aura free for 14 days, without a payment card.
| Novafree for good | Pulse€9 a month | Aura€29 a month | Luna€79 a month | |
|---|---|---|---|---|
| A/B tests at the same time | Not in this plan | 1 | 3 | 10 |
| Tests with your own code (CSS, JavaScript) | Not in this plan | Not in this plan | Included | Included |
| Start free | Choose Pulse | Choose Aura | Choose Luna |
Questions
The variant is chosen on our server from the same daily fingerprint as the measurement, so a visitor sees the same variant all day. With “remember the variant” the choice can last longer on the visitor’s device (from Aura).
No. The small test script belongs high in the page head and applies the changes before the page is drawn: only a changed element stays invisible until it is changed (at most 600 ms), never the whole page.
Visitics shows the probability that a variant is better and waits for the planned number of visits. A result without a difference is a result too — it goes to the library with its lesson.
From Pulse (one test at a time, without code); Aura adds tests with code and more tests at once (see the table above).
At least seven full days, so every weekday is in it, and until the planned sample is reached. The calculator estimates the days from your traffic; a test that would need more than eight weeks is marked as not worth it.
Yes, from Aura: “remember the variant” keeps the choice on the visitor’s device for up to 13 months, the limit the French authority CNIL sets for analytics. Without it nothing is stored in the browser.
Goes well with
One line of code and the first data within seconds. Aura free for 14 days, then Nova for good — no payment card.