Root cause discovery engine

1. A computer-implemented method comprising:

receiving operations data from a plurality of networked agents;

storing the operations data in a historical data store;

building a machine learning model using the historical data store and a dependency graph;

receiving an indication of a user experience problem, wherein the user experience problem is associated with observed operations data;

converting the observed operations data into a set of observed features and an observed outcome;

generating, based on the set of observed features, a first predicted outcome according to the machine learning model;

determining that the observed outcome is within range of the first predicted outcome;

identifying a set of candidate causes of the user experience problem when the observed outcome is within range of the first predicted outcome;

identifying within the set of candidate causes a candidate cause that is most heavily weighted;

replacing a value in the observed operations data corresponding for a metric corresponding to the identified candidate cause with a historical value for the metric calculated based on the operations data in the historical data store;

generating, based on the replaced value, a second predicted outcome according to the machine learning model; and

identifying the candidate cause as an actual cause of the user experience problem when the second predicted outcome is different from the first predicted outcome.

2. The computer-implemented method of claim 1 , wherein the operations data includes metrics and events.

3. The computer-implemented method of claim 1 , further comprising providing the set of candidate causes, via a network, to an administrator machine.

4. The computer-implemented method of claim 1 , wherein a candidate cause in the set of candidate causes corresponds to a metric that is correlated to the user experience problem.

5. The computer-implemented method of claim 1 , further comprising providing the actual cause, via a network, to an administrator machine.

6. The computer-implemented method of claim 1 , wherein the historical value for the metric is an average value.

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

identifying the candidate cause as not an actual cause of the user experience problem when the second predicted outcome is the same as the first predicted outcome.

8. A non-transitory computer-readable medium comprising instructions, the instructions, when executed by a computing system, cause the computing system to perform operations comprising:

receive an indication of a user experience problem, wherein the user experience problem is associated with observed operations data;

convert the observed operations data into a set of observed features and an observed outcome;

generate, based on the set of observed features, a first predicted outcome according to a machine learning model;

determine that the observed outcome is within range of the first predicted outcome;

identify a set of candidate causes of the user experience problem when the observed outcome is within range of the first predicted outcome;

identify within the set of candidate causes a candidate cause that is most heavily weighted;

replace a value in the observed operations data corresponding for a metric corresponding to the identified candidate cause with a historical value for the metric calculated based on the operations data in the historical data store;

generate, based on the replaced value, a second predicted outcome according to the machine learning model; and

identify the candidate cause as an actual cause of the user experience problem when the second predicted outcome is different from the first predicted outcome.

9. The non-transitory computer-readable medium of claim 8 , wherein the operations further comprise:

receive operations data from a plurality of networked agents;

store the operations data in a historical data store; and

build the machine learning model using the historical data store and a dependency graph.

10. The non-transitory computer-readable medium of claim 9 , wherein a candidate cause in the set of candidate causes corresponds to a metric that is correlated to the user experience problem.

11. The non-transitory computer-readable medium of claim 10 , wherein the operations further comprise:

provide the actual cause, via a network, to an administrator machine.

12. A system comprising:

a processor; and

a non-transitory computer-readable medium storing instructions that, when executed by the system, cause the system to perform operations including:

receive an indication of a user experience problem, wherein the user experience problem is associated with observed operations data;

convert the observed operations data into a set of observed features and an observed outcome;

generate, based on the set of observed features, a first predicted outcome according to a machine learning model;

determine that the observed outcome is within range of the first predicted outcome;

identify a set of candidate causes of the user experience problem when the observed outcome is within range of the first predicted outcome;

identify within the set of candidate causes a candidate cause that is most heavily weighted;

replace a value in the observed operations data corresponding for a metric corresponding to the identified candidate cause with a historical value for the metric calculated based on the operations data in the historical data store;

generate, based on the replaced value, a second predicted outcome according to the machine learning model; and

identify the candidate cause as an actual cause of the user experience problem when the second predicted outcome is different from the first predicted outcome.

13. The system of claim 12 , wherein the operations further include:

receive operations data from a plurality of networked agents;

store the operations data in a historical data store; and

build the machine learning model using the historical data store and a dependency graph.

14. The system of claim 13 , wherein a candidate cause in the set of candidate causes corresponds to a metric that is correlated to the user experience problem.

15. The system of claim 12 , wherein the operations further include:

provide the actual cause, via a network, to an administrator machine.