When the Score Becomes the Business
By Aldridge Dagos, operations software engineer
A restaurant can turn its tables faster and become a worse place to spend an evening.
Imagine a neighborhood restaurant whose owner notices a long wait at the door. Some tables sit uncleared. Diners struggle to catch a server when they want the bill. People who would happily buy dinner leave because nobody can tell them when they will be seated. Getting tables ready sooner seems like a sensible place to start.
The owner begins measuring the time between a party sitting down and the next party taking its place. Staff get better at clearing empty glasses. The kitchen finds a delay between finishing a meal and sending it out. Customers spend less time trying to pay. For a while, a shorter visit means a better visit.
Then the measure becomes the target. Managers begin comparing shifts. A bonus depends on beating last month’s average. The server who waits for a conversation to pause before offering the bill starts to look slower than the server who places it beside an unfinished dessert. Both have learned what the business wants. Only one of them has learned to ignore what the table wants.
The spreadsheet has an excellent evening. Whether the guests do is another question.
A number can begin as a way to notice something and end as an instruction to neglect something else. The change is easy to miss because the number keeps the same name.
Table turnover still sounds like table turnover. The manager is now using it to decide what kind of evening the restaurant should sell. The staff member who used to be praised for reading a table well must decide whether that skill still has a place in the job.
Jerry Z. Muller makes a useful distinction in the introduction to his 2018 book, The Tyranny of Metrics. He accepts that standardized evidence can improve decisions and correct the limits of personal experience. His concern is the habit of allowing a measure to replace the judgment it was supposed to inform, especially when rewards and penalties depend on the result.1
The difficulty is that a measure has to leave things out. You can record how long dinner took. Recording whether the pace felt right is harder. A couple celebrating an anniversary and someone eating before a train may want very different amounts of time. Compressing those visits into an average makes comparison possible, but it also removes part of the explanation.
There is nothing dishonest about an average. The mistake is asking it to settle a question it cannot answer. If the restaurant wants to help people leave promptly when they are ready, it needs to learn where unwanted waiting occurs. If it wants every customer gone quickly, a shorter total visit may be exactly the desired result. Those are different businesses hiding inside the same report.
Once money depends on the result, the distinction becomes more urgent. A target tells people where to put effort when they cannot do everything. A bonus tells them which effort the owner is willing to pay for. If a careful conversation makes an employee look worse, a speech about thoughtful service leaves that employee with an awkward choice.
This need not involve cheating. Someone can follow the rules perfectly and still make the work worse. A bookseller asked to raise the value of each sale may recommend a more expensive edition to a reader who would have been just as happy with the paperback. The receipt improves. The reader’s reason to trust the recommendation becomes less secure.
The 2016 Wells Fargo enforcement case shows how serious the consequences can become. The US Consumer Financial Protection Bureau described employees opening unauthorized accounts, spurred by sales targets and compensation incentives. The activity could increase the recorded sales figures while harming the very customers those products were supposedly serving.2 That case involved misconduct. It should not be used to suggest that every ambitious target creates fraud. It does show why a business must examine how a reported success was produced.
It would be comfortable to stop there and declare that good work cannot be measured. An owner still has to pay wages. Customers still need orders to arrive. A manager who refuses to examine results can protect poor decisions behind the language of craft just as easily as another manager can hide them behind a chart.
There is evidence for taking incentives seriously. In a study published in the American Economic Review in December 2000, Edward Lazear examined Safelite Glass Corporation’s move to piece-rate pay. The study reported a 44 percent increase in output per worker and an increase in profits.3 That is a result from one company under a particular arrangement, not a promise about what happens when anyone introduces a bonus. It is enough to make a blanket rejection of measured performance unconvincing.
Fitting more windshields and opening more bank accounts may both appear in a column labeled productivity. Whether the extra activity creates value depends on the work, the customer’s need and the conditions under which the activity counts as finished. The label does not do that thinking for us.
A thought experiment 01
Same elapsed time. Different needs.
“Could we have the bill?”
Bring the bill
“May we see the dessert menu?”
Offer dessert
The same time on the clock can call for different service.
I like a measure that makes a decision easier to examine. If the first response to a concern is that the target was met, the report has begun closing a conversation that should still be open. The owner should be able to enjoy a good result and remain curious about how it happened.
Take the restaurant again. Suppose the shorter visits come from bringing the bill promptly after a guest asks. The customer gets time back, and the owner can seat the next party sooner. The same improvement helps both. Suppose instead that the shorter visits come from making people feel uncomfortable ordering another coffee. The owner may still prefer that policy, but should be willing to call it a choice about the service rather than evidence that the service improved.
That distinction gives staff something useful to discuss. A server can describe a table that wanted to leave and could not. Another can describe a table that felt pushed. The stories do not cancel the numbers. They explain which kind of change the numbers contain.
Of course, stories can be selected too. A manager might remember the one delighted guest and overlook a week of complaints. A senior employee might use an unusual case to defend a habit that wastes everyone’s time. Judgment needs challenge as much as measurement does. The owner should be able to ask whether the example is typical and look at visits that nobody selected to make a point.
This is where a business can become more exact without becoming more elaborate. Look at a few completed customer experiences. Compare what was promised with what happened. Ask whether the improvement remains an improvement when you include the inconvenience passed to somebody else. A report gains meaning when the people reading it can follow the result back to the work.
More measures are not always the answer. Add enough of them and a team can spend the afternoon explaining why one number fell because another had to rise. You can also create a score so complicated that nobody knows which judgment it contains. The owner has not escaped the need to choose. The choice has merely moved into the weighting formula.
Some standards should be conditions of success. A dinner served quickly but charged incorrectly is unfinished work. A sale the customer never agreed to cannot become a good sale because it helps a monthly total. Making that boundary clear gives the person doing the work a reason to stop, even when stopping makes the report less attractive.
Other decisions require room for an exception. A server should be able to recognize that a guest needs help understanding the menu, provided the restaurant has made time for that kind of service. If management wants discretion, it must accept that a justified decision may occasionally look untidy in the daily comparison. Otherwise the permission exists only until someone uses it.
There is a cost to that room. Managers have to explain why an exception was reasonable. Staff need enough knowledge to use it well. An owner cannot praise independent judgment and then avoid the work of discussing what good judgment looks like. A rigid target is simpler to administer. Simplicity is a real benefit, but it does not settle whether the target serves the business.
The hardest results to defend may be the ones that prevent a sale. A mechanic who explains that an expensive repair can wait has brought in less revenue today than someone who recommends doing it immediately. If that advice is sound, the customer has received something valuable. The business needs a way to recognize that contribution without pretending it can assign an exact future profit to every honest conversation.
That also means accepting uncertainty. We cannot confidently turn every act of restraint into a claim about loyalty. Some customers will leave anyway. Some will take the advice and buy elsewhere. A business may still decide that sound advice is part of the service it sells, and price that service so it does not depend on recommending unnecessary work.
What the owner chooses to count then becomes an expression of what the business is willing to stand behind. The numbers help test whether the promise can be afforded. The promise helps decide which numbers deserve to be called progress.
Back at the restaurant, a guest has finished dinner and is looking toward the door. At the next table, someone has just asked to see the dessert menu. Both tables have been occupied for the same length of time. A good server can see that they need different things.
The owner should want the report to help that server, and should be curious when the report cannot. There may be a better measure to use. There may also be a moment when the right decision belongs to the person standing beside the table.
Let them offer dessert.
Footnotes
-
Jerry Z. Muller, The Tyranny of Metrics (2018), publisher-provided introduction. ↩
-
US Consumer Financial Protection Bureau, Wells Fargo Bank enforcement action, September 2016. ↩
-
Edward P. Lazear, “Performance Pay and Productivity”, American Economic Review, December 2000. ↩