The ways a crisis might evolve

A large brand finds itself having to recall a significant number of its products. They want to estimate how long this topic will attract attention and its magnitude, in order to understand the damage to the brand's public image and to successfully resolve the crisis.
We worked on the estimation of the possible outcomes of the event. We compared the current properties of the event with historical data. We calculated the number of mentions of the event every day (or hour) since the beginning of the crisis. And then we searched for past cases that exhibited similar patterns. From the current state of the system and the history of similar cases, we were able to propose different values of confidence about the possible outcome.
The crisis remained active for several days as predicted until it finally subsided. The actual magnitude of the event in the following days was close to the mean value of the distribution that we had used for estimation (technically within one standard deviation). The company preferred a quick, honest, full disclosure response to the crisis in order to inform its customers and to minimize the damage.