Enhanced network load allocation and variant selection for multivariate testing

1. A computer-implemented method comprising:

identifying a control interface representing a control version of a user interface for multivariate testing;

identifying one or more interface variants, each interface variant of the one or more interface variants representing an alternative version of the control interface, wherein the one or more interface variants differ from the control interface;

allocating network traffic to the control interface and each interface variant of the one or more interface variants;

collecting, for the control interface and each interface variant of the one or more interface variants, metric data representing one or more measured metrics, each measured metric of the one or more measured metrics measuring at least one aspect of an interaction between a visitor browser and either an interface variant of the one or more interface variants or the control interface;

determining that a size of the collected metric data for each interface variant of the one or more interface variants has reached a minimum threshold;

in response to determining that the size of the collected metric data has reached the minimum threshold, executing a variant selection protocol, the execution of the variant selection protocol being iteratively performed, and including:

calculating a selection metric for each interface variant of the one or more interface variants, the selection metric is calculated by computing a difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by an uncertainty value, wherein the uncertainty value is determined based on a combination of a variability metric of the one or more measured metrics calculated for the selected interface variant and a second variability metric of the one or more measured metrics calculated for the control interface;

automatically selecting one of the interface variants from the one or more interface variants, the selection metric being used to determine which interface variant to select from the one or more interface variants;

reallocating a predetermined percentage of the network traffic to each of the one or more interface variants and the control interface;

collecting metric data for each of the selected interface variant and the control interface;

determining a performance associated with each of the selected interface variant and the control interface, the performance being based at least in part on the collected metric data and a recalculation of the selection metric for each of the selected interface variant and the control interface; and

completing the variant selection protocol in response to any interface variant of the one or more interface variants being associated with a determined performance that reached statistical significance.

2. The computer-implemented method of claim 1 , wherein the one or more measured metrics includes a combinatorial metric representing a combination of two or more measured metrics.

3. The computer-implemented method of claim 1 , further comprising:

determining an uplift value for each interface variant of the one or more interface variants, the uplift value being calculated based on a difference between a measured metric calculated for the interface variant and the measured metric calculated for the control interface; and

executing the variant selection protocol only for interface variants for which the uplift value is positive.

4. The computer-implemented method of claim 1 , wherein the selection metric is calculated by computing the difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by the uncertainty value, wherein the uncertainty value is determined based on a sum of a first variance of the one or more measured metrics calculated for the selected interface variant and a second variance of the one or more measured metric calculated for the control interface.

5. The computer-implemented method of claim 1 , further comprising:

generating a reporting interface for displaying a status of each interface variant of the one or more interface variants, the status representing whether or not the determined performance of the selected interface variant has reached statistical significance.

6. The computer-implemented method of claim 5 , wherein the reporting interface depicts a plurality of objects, and one of the plurality of objects displays the status of each interface variant of the one or more interface variants.

7. The computer-implemented method of claim 1 , further comprising:

determining that a particular interface variant has reached statistical significance;

determining that the particular interface variant performed worse than the control interface with respect to the one or more measured metrics;

generating a notification indicating that the particular interface variant has performed worse than the control interface with respect to the one or more measured metrics; and

displaying the notification on a reporting interface.

8. The computer-implemented method of claim 1 , wherein the one or more interface variants comprise a modified versions of the control interface that comprises at least one of a different color and different text from the control interface.

9. The computer-implemented method of claim 1 , wherein the reallocating the predetermined percentage of the network traffic to each of the selected interface variants and the control interface enables a server to bias the network traffic towards best-performing interface variants.

10. A non-transitory machine-readable storage medium, including instructions configured to cause a processing apparatus to perform operations including:

identifying a control interface representing a control version of a user interface for multivariate testing;

identifying one or more interface variants, each interface variant of the one or more interface variants representing an alternative version of the control interface, wherein the one or more interface variants differ from the control interface;

allocating network traffic to the control interface and each interface variant of the one or more interface variants;

collecting, for the control interface and each interface variant of the one or more interface variants, metric data representing one or more measured metrics, each measured metric of the one or more measured metrics measuring at least one aspect of an interaction between a visitor browser and either an interface variant of the one or more interface variants or the control interface;

determining that a size of the collected metric data for each interface variant of the one or more interface variants has reached a minimum threshold;

in response to determining that the size of the collected metric data has reached the minimum threshold, executing a variant selection protocol, the execution of the variant selection protocol being iteratively performed, and including:

calculating a selection metric for each interface variant of the one or more interface variants, the selection metric is calculated by computing a difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by an uncertainty value, wherein the uncertainty value is determined based on a combination of a variability metric of the one or more measured metrics calculated for the selected interface variant and a second variability metric of the one or more measured metrics calculated for the control interface;

automatically selecting one of the interface variants from the one or more interface variants, the selection metric being used to determine which interface variant to select from the one or more interface variants;

reallocating a predetermined percentage of the network traffic to each of the one or more interface variants and the control interface;

collecting metric data for each of the selected interface variant and the control interface;

determining a performance associated with each of the selected interface variant and the control interface, the performance being based at least in part on the collected metric data and a recalculation of the selection metric for each of the selected interface variant and the control interface; and

completing the variant selection protocol in response to any interface variant of the one or more interface variants being associated with a determined performance that reached statistical significance.

11. The non-transitory machine-readable storage medium of claim 10 , wherein the one or more measured metrics includes a combinatorial metric representing a combination of two or more measured metrics.

12. The non-transitory machine-readable storage medium of claim 10 , wherein the operations further comprise:

determining an uplift value for each interface variant of the one or more interface variants, the uplift value being calculated based on a difference between a measured metric calculated for the interface variant and the measured metric calculated for the control interface; and

executing the variant selection protocol only for interface variants for which the uplift value is positive.

13. The non-transitory machine-readable storage medium of claim 10 , wherein the selection metric is calculated by computing the difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by the uncertainty value, wherein the uncertainty value is determined based on a sum of a first variance of the one or more measured metrics calculated for the selected interface variant and a second variance of the one or more measured metrics calculated for the control interface.

14. The non-transitory machine-readable storage medium of claim 10 , wherein the operations further comprise:

generating a reporting interface for displaying a status of each interface variant of the one or more interface variants, the status representing whether or not the determined performance of the selected interface variant has reached statistical significance.

15. The non-transitory machine-readable storage medium of claim 14 , wherein the reporting interface depicts a plurality of objects, and one of the plurality of objects displays the status of each interface variant of the one or more interface variants.

16. The non-transitory machine-readable storage medium of claim 10 , wherein the operations further comprise:

determining that a particular interface variant has reached statistical significance;

determining that the particular interface variant performed worse than the control interface with respect to the one or more measured metrics;

generating a notification indicating that the particular interface variant has performed worse than the control interface with respect to the one or more measured metrics; and

displaying the notification on a reporting interface.

17. A system, comprising:

one or more processors; and

a non-transitory computer-readable storage medium containing instructions which, when executed on the one or more processors, cause the one or more processors to perform operations including:

identifying a control interface representing a control version of a user interface for multivariate testing;

identifying one or more interface variants, each interface variant of the one or more interface variants representing an alternative version of the control interface, wherein the one or more interface variants differ from the control interface;

allocating network traffic to the control interface and each interface variant of the one or more interface variants;

collecting, for the control interface and each interface variant of the one or more interface variants, metric data representing one or more measured metrics, each measured metric of the one or more measured metrics measuring at least one aspect of an interaction between a visitor browser and either an interface variant of the one or more interface variants or the control interface;

determining that a size of the collected metric data for each interface variant of the one or more interface variants has reached a minimum threshold;

in response to determining that the size of the collected metric data has reached the minimum threshold, executing a variant selection protocol, the execution of the variant selection protocol being iteratively performed, and including:

calculating a selection metric for each interface variant of the one or more interface variants, the selection metric is calculated by computing a difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by and the selection metric including an uncertainty value, wherein the uncertainty value is determined based on a combination of a variability metric of the one or more measured metrics calculated for the selected interface variant and a second variability metric of the one or more measured metrics calculated for the control interface;

automatically selecting one of the interface variants from the one or more interface variants, the selection metric being used to determine which interface variant to select from the one or more interface variants;

reallocating a predetermined percentage of the network traffic to each of the one or more interface variants and the control interface;

collecting metric data for each of the selected interface variant and the control interface;

determining a performance associated with each of the selected interface variant and the control interface, the performance being based at least in part on the collected metric data and a recalculation of the selection metric for each of the selected interface variant and the control interface; and

completing the variant selection protocol in response to any interface variant of the one or more interface variants being associated with a determined performance that reached statistical significance.

18. The system of claim 17 , wherein the one or more measured metrics includes a combinatorial metric representing a combination of two or more measured metrics.

19. The system of claim 17 , wherein the operations further comprise:

determining an uplift value for each interface variant of the one or more interface variants, the uplift value being calculated based on a difference between a measured metric calculated for the interface variant and the measured metric calculated for the control interface; and

executing the variant selection protocol only for interface variants for which the uplift value is positive.

20. The system of claim 17 , wherein the selection metric is calculated by computing the difference between the one or more measured metrics calculated for the selected interface variant and the one or more measured metrics calculated for the control interface, multiplied by the uncertainty value, wherein the uncertainty value is determined based on a sum of a first variance of the one or more measured metrics calculated for the selected interface variant and a second variance of the one or more measured metrics calculated for the control interface.

21. The system of claim 17 , wherein the operations further comprise:

generating a reporting interface for displaying a status of each interface variant of the one or more interface variants, the status representing whether or not the determined performance of the interface variant has reached statistical significance.

22. The system of claim 21 , wherein the reporting interface depicts a plurality of objects, and one of the plurality of objects displays the status of each interface variant of the one or more interface variants.