A/B testing

A/B testing

A/B testing may be used to test various content within the product. You can create experiments using programming or via the interface. For example, the following code will create an experiment that breaks down traffic into three equal parts, and displays the different product values for each third of your customers.

window.airgrow.experiment("experiment-name", ["v1", "v2", "v3"]).then(function(cohort){
    switch (cohort){
        case "v1":
            document.getElementById("price").innerHTML = "$30";
            break;
        case "v2":
            document.getElementById("price").innerHTML = "$40";
            break;
        case "v3":
            document.getElementById("price").innerHTML = "$30";
            break;
    }
})

A weighting coefficient may be used to set a nonuniform distribution. The following code will produce the variant v1 for 50% of customers, v2 for 25%, and v3 for the remaining 25%:

window.airgrow.experiment("experiment-name", {"v1": 2, "v2": 1, "v3": 1})

Traffic break-down percentages may also be specified directly:

window.airgrow.experiment("experiment-name", {"v1": "35%", "v2": "12%", "v3": "53%"})

Not all 100% of the total percentage necessarily has to be set. For example, a value of 10% may be set for one variant. This will establish the selected cohort in 10% of cases, while in the remaining 90% the traffic will not be marked.

Traffic break-down can be set only once. Thereafter you can simply query with what cohort the given visitor is already marked using the experiment name. This query works across all products. In other words, if a user has been marked on the website, this marking may be checked in all your products that are linked to the tracker.

window.airgrow.experiment("experiment-name").then(function(cohort){
    switch (cohort){
        case "v1":
            document.getElementById("price-below").innerHTML = "$30";
            break;
        case "v2":
            document.getElementById("price-below").innerHTML = "$40";
            break;
        case "v3":
            document.getElementById("price-below").innerHTML = "$30";
            break;
    }
})

Analysis of experiment results

The results of the experiment may be analyzed using cohort analysis. To analyze changes in the funnel go to the Funnel section and click on the Filter button.

Click Add to comparison to filter traffic for another cohort. In the result you will see which cohort was most successful in processing the data, and to what degree.

A/B testing
Analysis of experiment results