In defense of self-service checkout

I have previously stated that I love self-service checkouts. It may just be that I hate making eye contact (ask anyone who has had to sit through me teaching). I consequently worry every time there is a story suggesting that supermarket chains are eliminating them. Gady had a post on this trend this summer and now the Associated Press is piling on with an article that was picked up by several newspapers (Supermarkets start bagging self-serve checkouts, Sep 26). Here is what the AP says:

Big Y Foods, which has 61 locations in Connecticut and Massachusetts, recently became one of the latest to announce it was phasing out the self-serve lanes. Some other regional chains and major players, including some Albertsons locations, have also reduced their unstaffed lanes and added more clerks to traditional lanes.

Market studies cited by the Arlington, Va.-based Food Marketing Institute found only 16 percent of supermarket transactions in 2010 were done at self-checkout lanes in stores that provided the option. That’s down from a high of 22 percent three years ago.

Overall, people reported being much more satisfied with their supermarket experience when they used traditional cashier-staffed lanes. …

An internal study by Big Y found delays in its self-service lines caused by customer confusion over coupons, payments and other problems; intentional and accidental theft, including misidentifying produce and baked goods as less-expensive varieties; and other problems that helped guide its decision to bag the self-serve lanes.

So one interpretation of this is that the systems are too hard to use and thus unappealing to the average shopper. The counter argument to that is in the United Kingdom. Tesco, the heavyweight champion among British supermarkets, racks up 10 million transactions per week on its self-service tills, fully one third of all its transactions (Tesco Deploys NCR Self-checkouts in Central, Eastern Europe, Sep 16, Progressive Grocer). Perhaps the Brits are more willing to adopt new technology or maybe Tesco aggressively understaffs it regular registers so that customers have no choice but to use the self-service option. In any event, it should be possible to have a decent level of adoption in the US.

The question then is what’s in it for the customer?

I would contend that it is possible for self-service registers to expand capacity cheaply in a way that makes customers better off. This should be true even when self service is an inferior technology in the sense that it takes more time for any part of the transaction. Let me try to give an example. Suppose that a standard checkout requires 2 seconds per item and then 45 seconds to take the payment. (See here for more on the consequences of having a fixed, “set up” time as part of the transaction.) Suppose the average self-service customer needs 5 seconds per item and 60 seconds to pay. If the average basket has 15 items, a standard checkout transaction requires 75 seconds while self-service would require 135 seconds — a 56% increases!

So if it is a straight up question of one technology for the other, customers are better off with all standard checkouts. Note that I am defining better off here as having the most capacity as that will presumably provide the shortest wait. For my example, four standard checkouts results in enough capacity for 192 customers per hour while four self-service checkouts could only handle 106.7 customers per hour.

Of course, cost has to figure into the equation. I had a student report several years ago that put the cost of a four lane self-service checkout in the ballpark of $100,000. Even if the annual cost of a cashier is only $18,000 and the self-service lanes are only good for five years, that is a good trade.

So suppose that store will use both self-service and standard checkouts. How should they be deployed? Using the number I have above, if there are four of each type of register, then the store can process 192+106.7 = 298.7 customers per hour if any customer uses any machines.

But we can actually do better — possibly a lot better. The key observation is that not every customer has the same size basket and when we move a customer from the standard checkout to the self-service one, the hit they take in increased processing time is increasing in the size of their basket. So the smart thing to do is to route those with small baskets to the self-service lines.

Suppose that basket sizes are uniformly distributed between 5 and 25. That is, the chance that the next customer has a basket with exactly 8 items is equal to the chance that his basket is 6 or 16 etc. I know this is unrealistic, but it is simple and illustrates my point. Also, it keeps the average basket size across all customers fixed at 15.

Now let’s impose the policy that only customers with 10 or fewer items use the self-serve lanes while standard checkouts are saved for  big (i.e., greater than 10 items) baskets. The average basket size for those using self-service is then 7.5 and while on the standard lines its 18. The capacity of a self-service line is 36.9 customers per hour while a standard checkout can process 44.4. If we have four of each, we can now process 325.5 customers per hour — nearly a 9% increase. To deliver a similar capacity with just standard checkouts, the firm would need seven cashiers.

This is not the only policy that expands capacity relative to having all customers use all machines. As long as the basket size being routed to the self-service machines is less than 14, we get an increase in capacity.

So what do we get from this analysis? It shows that customers can be better off with a mix of self-service and standard checkouts if they are used smartly. A mixture of checkouts can deliver a level of service that a store would likely not be able to afford if it could only use standard checkouts.

The analysis also shows that the basket mix matters. The gain from routing small baskets to self-service is more pronounced when there is more variability in basket size. Suppose, for example, that shoppers either had baskets in the range of 5 to 10 or in the range of 20 to 25 but that average across all customers was still 15. Now we really want to separate small from big baskets. This suggest one reason why self-service is more successful in the UK. If British customers shop more frequently for smaller baskets, they incur a smaller penalty when moving from standard to self-checkout. Further, retailers have more of an incentive to offer them that option.

It also suggests something to US supermarkets considering killing self-service checkout: Labor costs are going to increase with the move (for a given level of service) and how much they increase will depend a lot on basket mix. If we are looking at a suburban location where most customers only make one trip a week, then there may little penalty for the move. However, in an urban setting with big swings in basket size, a mix of self-service and standard checkout may be best.