An exciting product that we have been developing here is our 'Keeping the Customer' API.

Customer churn is a problem for any business, and given that it costs a lot less to keep an existing customer than it does to get a new one, something that should be foremost.

Why do customers leave? Truth is, often we don’t know until it was too late. Then, browsing the CRM system for the customer's records, we can often see in retrospect all the clues: the customer downsized their plan, they had a run in with someone from customer support, then there were a couple of credits issued for a billing mistake, and then when the customer's agreement expired, they left.

Often businesses run reports with fixed criteria, or set up workflows on common items that can cause the retention team to spur into action and try and save the customer before it's too late. However, often the reasons why the customer leave are ambiguous and not able to be fixed, and then even if they can be fixed, policy change within the organisation can mean that the rules need to be continually updated in order to remain relevant.

What if we could find the reasons people leave, and then continually adapt these reasons as the business evolves over time? With machine learning, we can.

The 'Keeping the Customer' API works currently with Salesforce (other CRM can access on request), interfacing our Machine/Deep learning systems through the Salesforce Bulk API.
We draw data down from your CRM system and then use deep learning to work out why your customers are leaving. The machine learning is able to find combinations of trend that would otherwise be impossible for traditional reporting/BI to discover.

Then, on a continual basis, we monitor your CRM data, looking for any customers that may fit the learned conditions. We update, again through the Salesforce API, the status of any customer that looks to be 'at risk', highlighting them in CRM so that the retention team can try and amend any problem before the customer decides to find another provider.

The API is in early release but has shown in trials that depending on the quality of CRM data, accuracies of over 90% in predicting customer churn.