3
$\begingroup$

I'm unfamiliar with OR, and would like to get some advice on how to "think" about this scenario:

  • I work for an organization that primarily has store fronts. Customer's arrive throughout the day and are serviced per their needs. Some needs require less than 10 minutes to service. Some require half an hour. There's probably 10 types of services a customer can come in for.
  • The most frequently used services can be completed in less than 10 minutes. The complex ones - around half an hour.
  • Now, should a customer call the help-line and be scheduled for an appointment (instead of randomly walking in), the time taken to help the customer schedule an in-person appointment for the right reasons, is approx 10 minutes.
  • Most times, the customer just walks-in.

The scenario:

  • The organization is recovering from the Pandemic and would like to switch to an appointments only model.
  • The organization will have self scheduling options in the future, but not on day one of opening up.

The question:

  • How would one go about modeling/comparing the FCFS (original) situation, with an appointment only model?
  • How would I arrive at a reasonable scheduled appointment duration for a given type of service. Meaning: if a frequently used service takes 10 minutes to complete (on average), how much time should I block off for that service, in an appointments only model?

Data:

  • I don't have much data in hand. I have the average time (minutes) to complete a service, by service type. I have the total number of service requests by service type (by month). I don't have customer wait times.
$\endgroup$
4
$\begingroup$

Since you are unfamiliar with OR, I would recommend using discrete event simulation, which I think is the easiest approach for a newcomer (although it may require some programming chops, depending on what software you use). You will need a bit more than just average service completion times -- you will want a distribution of service times. (In the absence of a robust data sample, you will probably have to just assume a commonly used distribution, such as exponential.) Similarly, for the walk-in case you will need a distribution of what are called interarrival times (time between consecutive customer arrivals), which again might be assumed exponential. For the appointment case, you may need a distribution for time between arriving calls for appointments (maybe exponential).

That may be enough to get you started. For a more realistic model, you may need some additional information, including when (and for how long) servers take breaks (you do give them breaks, don't you?), how likely a walk-in customer is likely to wait before bailing out ("reneging"), how likely an arriving walk-in customer is to look at line length and think "no way" ("balking"), how likely a customer with an appointment is to be a no-show or to be late (in which case, how late?), and possibly some other criteria. You can build a fairly simple model for the FCFS case, play with it, and see if people in the know think the results match their experience ("face validation"). Then you can start tweaking the model to reflect policy choices. The reservation-only model can be approached similarly.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.