Sturdy Statistics Case Studies: Telecom Provider

Takeaway:  Answers were readily available in the call transcript data, but without Sturdy Statistics, the company had no practical means to uncover them. Schedule a demo to unlock your unstructured data today.

A telecom provider was optimizing its sales process. A key initiative for this company was to enable its customers to “self-serve,” or to complete purchases online, without enlisting the help of a sales agent. The company had noticed a proportion of its sales calls originating from cart pages — in other words, many sales calls were coming from customers who were already very close to completing a purchase. These calls represented an obvious opportunity for improvement. However, despite having a team of data analysts to study the problem, the company did not know why its customers were picking up the phone instead of simply clicking “complete purchase.” Like the insurance company described above, this company was only analyzing the metadata associated with the calls, and had no method to analyze the call transcripts. This was, however, precisely where the answer to their question was to be found.

Sturdy Statistics ingested one year’s worth of calls, along with metadata about the page each call originated from. Our technology automatically transformed each call into a discrete set of topics, and we ran a regression analysis to identify which topics associate with calls from each page. This analysis immediately surfaced a number of topics distinctive to calls from the cart page. The most significant of these had to do with credit locks: since the company performs a credit check when opening new accounts, callers who had locked their credit were unable to register for service. The telecomp provider had migrated a portion of its human call center agents to AI agents. Callers with locked credits was an operational inefficiency when the call was routed to a human agent: they ask the customer to unfreeze their credit and try again in a few hours. However, when routed to an unprepared AI agent, the callers were redirected to irrelevant departments, put on extended holds, and given no information about how to proceed, effectively losing a new customer's business. Armed with this information, the company decided to include instructions to unlock credit on their registration page for new accounts. Additionally, they took steps to ensure their AI agents were able to clearly communicate this issue when it arose to customers.

The next most significant topic was about fiber availability: before upgrading their internet plans, customers were calling to confirm that fiber would actually be available at their address. Similarly, many calls were to confirm whether customers’ favorite channels would be included with TV subscriptions. Such questions could be preempted with small changes to the website. The last major reason was that the customer may have been a previous customer and needed to pay an outstanding debt before reinstating service.

Each of these insights enabled the company to improve its efficiency and to advance its goal of helping its customers self-serve. And each of these insights was plainly visible in the phone conversations — the callers of course described why they were calling! However, because the company’s analysts had no means to dig into the call transcript data, they were unaware of every one of these issues. Sturdy Statistics was able to surface them, easily and virtually immediately.