Measuring Fried Chicken, or how perverted are your incentives?
It is sometimes a curse to know stuff about management. I was reminded of this earlier today when I purchased lunch in a branch of KFC. I ordered on my phone, and my order number appeared on the screen above the counter, under a message saying “Preparing your order”. Two seconds later it disappeared, and reappeared under “Ready to collect”. I am too experienced at ordering a Twisted Wrap to get excited though, because I have watched this game too many times. But if I weren’t, I would be disappointed at this point.
KFC managers clearly want to ensure customer satisfaction. Measuring that is difficult. Even if it was practical to ask every single customer to fill out a customer satisfaction survey, only a small fraction of them would. And the ones who do fill it out would overwhelmingly be unhappy with the service, because unhappy customers want a chance to vent. (One of the distinctions between a product and service is often held to be that a service customer is often not able to articulate the quality of the service received.) So KFC use a proxy metric which I will call TTC: Time To Chicken.
TTC: Time To Chicken
TTC is the time it takes for an order to go from being placed to being fulfilled, through all of the steps in the order fulfilment process. Related to TTC is throughput; how many orders are fulfilled in a unit of time (the lower the TTC, the higher the potential throughput, although however fast you serve the customer, if you only get one order per hour your throughput will still be one order per hour). (Some sources call this lead time, some cycle time, some elapsed time; I am sidestepping this linguistic nightmare by calling it TTC.)
In theory, the shorter the TTC, the happier and more satisfied the customer, if you assume that the main contributor to customer satisfaction is speed of service. If I had to guess, I’d put money on there being an inter-store league for lowest mean TTC or highest throughput, likely with bonuses or other rewards for best performing staff or teams. This might even be distinguished in terms of TTC relative to size of order if managers are being clever (spoiler: they probably are not).
To track this KFC staff have a digital board (similar to a kanban board) which tells them what to prepare next. Orders come in and are displayed on the screen. Staff then pick the products from the hot shelves and drawers or fridges, and prepare drinks and other products from the dispensers. There’s probably an expectation of a quality assurance step to ensure everything is in the bag. Then they press an “Order Ready” button to signal the order is ready, which updates the customer information screen, and removes the order from their screen. The idea is that they turn around and call out the order number and hand the customer their food. Customer is happy (according to the speed = satisfaction proxy), and the team takes pride in fast delivery.
Predictably, of course, TTC gets gamed: every metric is gamed as soon as people know what is being measured and what’s expected of the metric (ie. should it go up or down). I’ve spent too much of my life in KFC waiting on a Twister Wrap watching orders come in; immediately the server will press the “Order Ready” button. The new order disappears from their screen and at the same time it appears Ready for Collection on the customer display. Staff then use a “Recall Order” function which brings the order back onto the screen so they can start fulfilling it. Sometimes there are five or six orders they’re keeping a track of using this recall order functionality. This has several consequences.
Mismanaged Chicken Expectations
The net result is that first of all, the customer’s satisfaction immediately reduces (unless they’re as cynical as me) because the customer display board, designed to manage our expectations, starts lying to us. This may be passive (in the sense that the customer notices and is a bit confused) or active (in the sense that the more anxious customer is worried about whether they’ve missed their order because someone else has collected it leading to increasing concern the longer it takes). This effect is, of course hidden from the system, which has already recorded its metric.
Forgotten Sides and Dips
Second, the order process takes longer and has more errors because the system as it operates relies on staff remembering unfulfilled orders, rather than having them on the screen at a glance; this further reduces customer satisfaction (increased wait time, forgotten orders, forgotten items; why can KFC never remember to include the spicy mayo you paid for?!). Again, this effect is hidden from the system.
Rewards are Predicated on Lies
Third, TTC is absolutely recorded as being fractions of a second, which signals to managers that the system is super efficient, and that, therefore, customers must be happy and satisfied. This likely means that people who do “immediate fulfilment” hit targets, and get rewards. Their managers probably get bonuses for being the local fastest TTC franchise or branch. This leads to a positive feedback loop which reinforces immediate fulfilment practise, therefore further reducing customer satisfaction by the mechanisms noted above.
Thus, the TTC as the only metric results in a perverse incentive, which does not reward the desired behaviours, but rewards staff lying to their managers.
I should note, and I imagine much to the relief of the reader, that my order was delivered in a timely manner, with all components present and correct (and I’ve learned to check they’ve included the gravy and popcorn chicken). And yet this still annoys me every time; I’m not quite at the anxious level, but I do sometimes wonder if someone else has stolen my saturated fried goodness.
How To Stop Your Chicken Metrics Being Perverted
How could a KFC manager ensure that this metric is not being gamed?
Sta-chick-stical Veri-fry-cation
At a system level, KFC could calculate realistic fulfilment times (eg. verified statistical ranges) for different size orders, and check recorded TTCs against the standard deviation of that to see if they are “realistic”. For example, a Boneless Box meal requires 45 seconds under ideal conditions, so if it gets fulfilled in 5 seconds, that is a significant outlier. One or two preparation times might come in shorter (or longer), but the vast majority will not be outliers; this would be an indicator that gaming is occurring.
Watch How Orders are Ful-fillet-ed
Gemba: go to the actual place. Branch managers should be verifying that the metrics are recorded accurately and that systems are being used appropriately. This means watching staff and ensuring they’re using the digital system correctly, for example, and not doing the immediate fulfilment statistical hack. However, staff will know your face: what about using secret shoppers, or having other managers verify your activities?
Monitor Other System Events for Zingers
Track order recalls on the display system. Once fulfilled, an order recall signals a quality problem (ie. a customer has said something is missing and staff are checking to see if it was on the order). If every order gets recalled repeatedly, it suggests that the system is not being used in the expected way, and brings the effectiveness of the process into question. Of course, high quality problem signals would trigger investigations: why are so many orders being recalled?
Track How Often Spicy Mayo is Forgotten
Record the number of quality events. Ensure that replacement items get recorded, so you also have a quality indicator for fulfilment, not just speed. This would of course require some kind of non-by-passable tracking system, although I have seen, for example, pubs tracking the amount of beer dispensed from pumps and checking the volumes against the sales figures. So perhaps if you’re concerned about customer satisfaction there is an option for tracking prepared food, which would also help address any stock loss problems, and could also help manage food quality and safety by tracking when hot food should be discarded. I also note that in combination with checking statistical deviation (see “first” above), it would ensure nobody is able to fulfil an order if there is insufficient product ready to serve.
Talk To (or at least watch) the Finger-Lickin’ Customer
Check customer satisfaction aligns with your metrics. If you’re on the shop floor, verify systems are being properly followed, talk to customers, and sense the vibes of the people waiting. You could also do formal surveys with win-free-food rewards for every tenth digital customer. Another option is to introduce something that verifies delivery, such as presenting a QR code to collect the order (much the same as providing a code to delivery drivers on popular apps). Regardless, this and tracking quality events will give a better indicator of the efficacy of the lead time metric as a proxy for customer satisfaction.
Check How Fry-able Your Systems Are
More senior managers should carry out periodic “viable system” audits to ensure your understanding of what’s happening two layers below you aligns with the design of the system that is in place. If the problems being faced by staff serving customers are different to the problems you thought they had, then your systems may be making their lives more difficult, and thus reducing customer satisfaction. This might well mean working shifts, and understanding the pressures at the point of delivery, which are probably not what you imagine they are from a distance, but this is a small sacrifice for high service quality!
Final Sharing Bucket
The key takeaway (pun carefully intended) here is to ensure your systems are fit for purpose, by verifrying your assumptions, and checking that everyone is playing the same game.
I still love a cheeky KFC, but I do wish this didn’t annoy me quite so much every time I order…
