Customer Return Probability

Last post, I talked about customer types, buying habits/frequency and the products that they could be upsold with. We went a little bit further in the detail of how each of those aspects of the barbershop business looks, but this week, we wanted to touch on some of the fundamentals of probability as it relates to the barbershop.

One thing that we have identified with me is that I go to the barbershop every time I feel my hair gets shaggy. If you look at my visitation habits, you can apply probability and realize that I’ll be back through the door every 5 weeks or so. This gives a probability graph that looks like this, with the highest probability of me returning being cantered around the 5 week mark :

What the visitation pattern of one customer will look like (if they got their hair cut at time=0)


This graph tells us that I’m most likely to come back through the door of my barber shop around 5 weeks after I visited last time. If I’ve gone 7 or 8 weeks without coming through the door, the barber may have lost me as a customer; if he really wants me as a customer, he should start asking some questions.

The astute barber who’s looking to grow their business, and who understands something about me will realize that he can achieve both of our goals (him a thriving business, and me to continually look decent without having to worry too much about looking too shaggy 6 weeks after my last cut) by inviting me back into the shop say 4 or 4.5 weeks after my last cut. This moves me from a customer who comes in 10 times in a given year, to one which visits 12 times a year. Do I care about the extra two visit costs over the year? No. I care about not looking unkept during business meetings.

Now, let’s consider two other customers which frequent the same barber shop. The first has hair which grows slower than mine, he typically comes in on average every 6 weeks. The second is younger and single. He wants to be looking his best all the time, and gets a trim on average every 4 weeks. If we plot the three of us together, you get the following (they’re the blue and green patterns):

Three customers plotted together, assuming that they all got their hair cut on the same day (at time=0)

Let’s walk through this graph together. If all three of us come in on the same day (time=0 on the above graph), then the barber can expect to see us again with the various probabilities which each of us have defined with data from previous visits.

You see that the three of us have different frequency patterns. Based on the three of us, the barber can predict how busy the shop is going to be based on the visitation patterns for these three customers. The probabilities of the three individuals returning are additive, and the resultant curve is in magenta below:

The magenta line shows the additive probability of the three customers visiting during the time period on the graph


If somehow only the three customers outlined above kept the doors open for our example barbershop, and IF those three customers came in on the same day (at time=0 on the above graph), the graph shows us that it would be safe for the barber to take then next couple of weeks off, and return to open the doors before our “blue” customer is likely to return.

Now, it isn’t likely that a successful barbershop has only three customers, and this is where things start to get really interesting! If you look at the visitation habits of all the frequent customers (hopefully a few hundred people) of the barbershop, the probability distribution for all of them is simply the addition of all the individual customers. This may give a probability pattern that might look like this:

A two month prediction of how regular customers will flow into the barber shop

The barber logs into our software on Monday morning, and this is the dashboard which is presented to him, showing what the predicted customer visitation pattern might look like over the next 2 months.

A graph like this gives all sorts of insights to the barber. During the down time, it might be good to grant some vacation to the other barbers in the shop – get some vacation off the books while the shop is going to be slow. During the uptime, it might mean that the 3 seats which are typically open on any give day might not be enough. At that time, it might be wise to have all the barbers back from vacation and open 1 or 2 extra seats…

Without this type of visibility, our barber might get into the “dip” period and think that he has done something to upset his customers, thinking “no one is coming in – have we lost the majority of our regulars?”

Clearly, walk-in customers are not accounted for in this graph. Only return customers. If the barber is starting a mail-out campaign in 6 weeks, and he knows from past experience that this gives a 10% bump in business, he might want something in the software which allows for this type of predictive bump in activity.

In my final post, I’ll help to explain how to use this same type of analysis with cross-sell activity (to ensure that the hair product is moving and the customers are being taken care of), and how these fundamentals work in larger operations when we put it all together.