Why ask this question?
While building chatbots and voicebots and designing AI tools for contact center agents, I kept seeing cases where more features and better performance still left the customer’s problem unsolved. What is one problem for a customer becomes the work of several teams inside a company. Each team can do its part well while the customer still does not get the result they expected, and reconnecting that work takes considerable effort and cost.
So this research asks less about the performance of a given technology or the efficiency of a single function, and more about whether the company as a whole delivers what it promised customers. Companies have long tried to turn complaints into changes to products and services. I trace what companies could see, who could make changes, what happened to customers afterward, and whether the approach lasted as the organization changed. I am not starting from a conclusion; the questions are refined as I read the cases. About explains how I came to this question.
What period does the research cover?
The current collection spans 1964–2026. It covers statistical methods for finding causes in complaints, the role of consumer affairs departments, customer surveys and voice-of-customer programs, and more recent conversation analytics. I also look at what counted as improvement in each setting. A decline in contacts and a decline in the problems customers experience require different evidence.
What can the sources tell us?
I start with who produced the record and why. A newspaper interview captures what someone said about a practice. A paper lays out its author’s method and analysis. A company record shows what the company reported. For each case, I record the attempt, the outcomes the evidence supports, and the questions that remain. When a study exists in more than one version, I identify the version used.
How are cases compared?
I ask who found the problem, who could authorize a change, where the resources came from, and how the company checked what happened to customers. I also look for attempts that stalled or cases where a different approach worked. Which conditions led to change is a question to investigate across cases.
As a working hypothesis, I compare three states: the customer problem is not recognized as a product or operational problem; the need to connect them is recognized but no working process exists; or a process was built but did not survive changes in sponsorship, staffing, or structure. These can appear in different parts of the same company, so they are not a grade for whole companies. Whether the distinction holds is part of what the cases will show.