Measuring the impact of network deployments

1. A method comprising, by one or more computing systems:

determining one or more network metrics for a plurality of geographic regions connected by a communication network, wherein the plurality of geographic regions comprise a geographic region of interest impacted by a network deployment;

identifying a set of candidate control regions by clustering two or more geographic regions of the plurality of geographic regions into a first group of geographic regions based on at least one metric feature of at least one network metric shared by the two or more geographic regions, wherein each metric feature measures a change in the value of the corresponding network metric during a particular time period;

selecting one or more control regions from the set of candidate control regions by identifying a second group of geographic regions from the first group of geographic regions based on a comparison between the region of interest and each of the first group of geographic regions;

comparing a first network metric associated with the geographic region of interest to a first network metric baseline, wherein the first network metric baseline is determined based on the one or more control regions; and

determining whether the first network metric associated with the geographic region of interest has a change greater than a first pre-determined threshold after the network deployment.

2. The method of claim 1 , wherein the one or more network metrics are determined based on network performance data collected during a first time period before the network deployment and a second time period after the network deployment.

3. The method of claim 2 , wherein the at least one metric feature shared by the first group of geographic regions is determined based on the network performance data collected during the first time period before the network deployment.

4. The method of claim 3 , wherein each of the second group of geographic regions and the geographic region of interest have a similarity being above a similarity threshold, and wherein the similarity is determined based on one or more metric features associated with the geographic region of interest and that geographic region before the network deployment.

5. The method of claim 2 , wherein the first network metric baseline is determined based on a weighted combination of the selected one or more control regions.

6. The method of claim 5 , wherein the second group of geographic regions are identified by a machine-learning model, and wherein the selected one or more control regions are used by a synthetic controls method.

7. The method of claim 5 , wherein the first network baseline is determined based on the network performance data collected during the second time period after the network deployment.

8. The method of claim 1 , wherein the second group of geographic regions are hold-out regions excluded from being impacted by the network deployment, wherein the selected one or more control regions are randomly selected from the second group of geographic regions, and wherein the first network metric baseline is determined based on the selected one or more controls regions.

9. The method of claim 1 , further comprising:

comparing a second network metric of the geographic region of interest after the network deployment to a second network metric baseline; and

determining whether the second network metric of the geographic region of interest has a change greater than a second pre-determined threshold after the network deployment.

10. The method of claim 9 , wherein the second network baseline is determined based on network performance data of the geographic of interest before the network deployment.

11. The method of claim 9 , wherein the second network metric comprises one or more of:

a churned user number;

a user churning rate;

a new user number; or

a resurrected user number.

12. The method of claim 1 , wherein the at least one metric feature comprises one or more of:

a metric trend;

a metric value;

a metric range;

a metric changing slope;

a metric error level;

a metric change; or

a metric confidence level.

13. The method of claim 1 , further comprising determining a cause-effect correlation between the change of the first network metric and the network deployment when the change of the first network metric is greater than the first pre-determined threshold.

14. The method of claim 1 , wherein the one or more network metrics comprise one or more of:

an incremental network metric;

a monthly active people metric;

a network speed metric;

a network latency metric; or

a time spent metric.

15. The method of claim 1 , wherein the geographic region of interest is a kilometer-level small-scale region.

16. The method of claim 1 , further comprising:

determining a first distribution of a third network metric of the geographic region of interest over a set of network metric buckets before the network deployment;

determining a second distribution of the third network metric of the geographic region of interest over the set of network metric buckets after the network deployment;

comparing the first distribution and second distribution of the third network metric; and

determining whether the third network metric has a change greater than a third pre-determined threshold after the network deployment.

17. The method of claim 16 , wherein the third network metric is determined by a carrier-specific method, a carrier-agnostic method or a user-based method.

18. The method of claim 1 , further comprising:

aggregating the one or more network metrics across two or more carriers impacted by the network deployment; and

determining whether the one or more network metrics have changes greater than respective pre-determined thresholds over the two or more carriers after the network deployment.

19. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:

determine one or more network metrics for a plurality of geographic regions connected by a communication network, wherein the plurality of geographic regions comprise a geographic region of interest impacted by a network deployment;

identify a set of candidate control regions by clustering two or more geographic regions of the plurality of geographic regions into a first group of geographic regions based on at least one metric feature of at least one network metric shared by the two or more geographic regions, wherein each metric feature measures a change in the value of the corresponding network metric during a particular time period, wherein each metric feature measures a change in the value of the corresponding network metric during a particular time period;

select one or more control regions from the set of candidate control regions by identifying a second group of geographic regions from the first group of geographic regions based on a comparison between the region of interest and each of the first group of geographic regions;

compare a first network metric associated with the geographic region of interest to a first network metric baseline, wherein the first network metric baseline is determined based on the one or more control regions; and

determine whether the first network metric associated with the geographic region of interest has a change greater than a first pre-determined threshold after the network deployment.

20. A system comprising: one or more non-transitory computer-readable storage media embodying instructions; and one or more processors coupled to the storage media and operable to execute the instructions to:

determine one or more network metrics for a plurality of geographic regions connected by a communication network, wherein the plurality of geographic regions comprise a geographic region of interest impacted by a network deployment;

identify a set of candidate control regions by clustering two or more geographic regions of the plurality of geographic regions into a first group of geographic regions based on at least one metric feature of at least one network metric shared by the two or more geographic regions, wherein each metric feature measures a change in the value of the corresponding network metric during a particular time period;

select one or more control regions from the set of candidate control regions by identifying a second group of geographic regions from the first group of geographic regions based on a comparison between the region of interest and each of the first group of geographic regions;

compare a first network metric associated with the geographic region of interest to a first network metric baseline, wherein the first network metric baseline is determined based on the one or more control regions; and

determine whether the first network metric associated with the geographic region of interest has a change greater than a first pre-determined threshold after the network deployment.