Feeds:
Posts

## Trade offs in achieving customer service

One of my sisters-in-law (I have several) runs a frame shop. A trip to my wife’s hometown often means stopping by her sister’s shop and hanging out while mats are being cuts and frames assembled. So I was curious then to read an essay to read in the New York Times about running a frame shop — even more so given its title, The True Price of Customer Service (Aug 21).

The essay is written by the founder of Artists Frame Service, which is based here in Chicago and has generally been very successful. It has been open for more than three decades and (according to the article) is 20 times bigger than the average US frame shop. Part of how it got that way is by promising faster service. Since its founding, it has promised to turn around orders in a week. Consistently delivering on that, however, creates operational challenges.

What I was really asking about was the one-week turnaround: Was it worth the trouble? Did it matter enough to customers? I was asking because it is not an easy commitment to keep. If someone orders framing on Monday afternoon, it will not go into production before Tuesday morning. And it has to be done by Friday for it to be inspected and wrapped and put in the pickup shelves by the following Monday. That means we really only have three days to get it done.

The challenge has always been that some weeks are so busy that my staff members have to work 10 or more hours of overtime, while other weeks are slower and their hours are cut. I considered switching to a 10-day turnaround. This would effectively double the amount of time that we have to complete an order. But I worried: Was this the equivalent of cutting Samson’s hair and losing the magic?

Let me start by saying that in terms of operations, this is a really nice problem to think about. It highlights some important points about managing processes to provide fast service. For example, if you are dealing with uncertain demand, working with just averages is not enough. An average day might mean getting 25 orders but you might go months without seeing any day with exactly 25 orders. If the concern is only about keeping up with demand, staffing to fill 25 orders per day is enough. However, if one cares about response time (i.e., completing orders within some deadline), things get more complicated. The firm will have to carry extra capacity, resort to overtime, or both in order to get the work done on time.

If one thinks only about one day, there is a fairly clean way to think about the use of overtime. Suppose we know the distribution of the number of new orders in each day. Further, suppose frames is frames so that every order takes the same amount of work for each department. Thus if we want to be able to fulfill 30 orders or 15 orders before the due date, we know how many people to staff. To make things simple, suppose that staffing to serve one order costs W and that staffing costs grow linearly — that is, staffing to serve Q orders costs WQ.

So how many people should that be? Well, this turns into a newsvendor problem. If we staff to complete Q orders in a timely fashion, we are committed to spending WQ. If actual demand is Q-1, we’ve spent W unnecessarily. On the other hand, if demand is greater than Q, hitting the deadline is more expensive. Assuming a 50% overtime premium, an order completed using overtime costs 1.5W. If we had known that demand was going to be high, we could have saved 1.5W-W = 0.5W by increasing our staffing level.

So what we really need to do here is balance the cost of paying for unneeded capacity (i.e., W) with the added cost of running overtime (0.5W). It turns out, that we want to set staffing so that we can handle Q* orders where Q* is such that the chance that we can handle all orders without running overtime is

$\frac{0.5W}{0.5W+W} = \frac{0.5}{1.5} = \frac{1}{3} .$

What does this mean? In a nutshell, it says we should be running overtime on the majority of days. Now in reality, things for Artists Frame Service are a little different. They are solving a multi-day problem. Labor scheduled for Wednesday to process orders that were placed on Monday can work on orders that arrived Tuesday if Monday demand was lighter than expected. But still, overtime figures in the equation. It would be impractical (i.e., wicked expensive) to schedule enough labor so that the deadline is always achieved while running straight time.

This gets us back to Artists Frame Service. What the essay focuses on is a set of “value conflicts.” To contain costs, Artists Frame Service wants to run fairly lean. But that means keeping a lid on head count, but that means overtime is inevitable. So here is the first conflict: The desire to honor the one-week promise to customers is in tension with respecting your employees’ lives outside of work. Said another way, my little analysis assumed that the cost of overtime is just a higher wage. That may be true if we are only looking at this week. Run OT every week for three months, and it’s a different story. The extra cash is nice but at some point it wears on people’s nerves to never have a weekend off.

So what can give? One possibility is to relax deadline. It’s a lot easier to hit a two-week deadline than a one-week one. They tried that, but it didn’t feel right. With many repeat customers, it is hard to shift to a lower level of service.

That leaves adding staff but even that is not free from tension.

Hire more people.

Sounds simple enough. But if you own a business and have survived the last five years, you know what that means – two more people to lay off if business drops again. I have 110 employees in all of my companies, and I am happy to say that, in the five or so recessions I have endured, we have cut hours and frozen hiring, but we have never had lay offs.

Here is the conclusion of the essay.

There is a price for giving great customer service – just as there is a price for not giving it. No matter what business you are in, it is a balancing act. Ultimately, you have to decide what is more important – staying lean or having a backup bench. Because sales fluctuate and people get sick or quit, it is not possible to calculate the precise number of people you will need over any given time.

It is time to hire, and that is good news.

As I said, I find this an interesting problem to think about. At a high level, it is an easy setting to analyze and have some intuition for what a good plan looks like. But that model is too abstract and ignore the inevitable challenges of perpetually running overtime. That presents the balancing act between customer service, respect for employees, and the risk of taking on more bodies.

### One Response

1. great article!