A/B Testing

In order to ensure that AristanderAI is working for your company, we perform extensive A/B testing and validation.

All A/B tests and validation data can be accessed at https://dashboard.aristander.ai

A/B testing can be performed in two ways: split testing, or time series.

Split Testing

Split testing is what most people think of as A/B testing: we have two pricelists, and sessions are assigned one of the pricelists at random (and the pricelist is tied to that session).

We are then able to compare the differences in effects between the two pricelists by comparing sessions.

The price mode in the event tracking API should be set to "split".

The end-user is then required to perform the session splitting + pricelist assignment/pinning.

It is up to the shop owner to decide whether pricelists are pinned for the remainder of the session, or whether the prices will be changed while a user is browsing.

AristanderAI will deliver prices via API for both pricelists.

/prices?source=aristander
/prices?source=original

Time Series

Time series testing is the better option when:

1) It's too much work to implement proper A/B testing

2) You don't want different users to see different prices

In time series testing, only one pricelist is delivered at a time. The caveat is that it requires more testing time to prove that our models are working.

The price_mode parameter should be set to 'timeseries'.

The pricelist_source parameter should be set to match the header returned in the prices API.

Fixed

If at any point during the use of AristanderAI you wish to use only one pricelist - set your 'price_mode' value to 'fixed' for all events.