My colleague Sunil pointed me a neat article on Domino’s Pizza’s Indian operations. While the chain long ago gave up on an explicit delivery time guarantee, their Indian franchisee Jubilant Foodworks still promises 30 minutes or the pie is free. That is not an easy promise to keep in, for example, an old neighborhood with streets running every which way and no really good maps. Still they manage to hit the thirty minute target remarkably often despite not having a whole of lot time for the actual delivery part of the process (Domino’s deadline to deliver, Financial Times, Jan 17).
With preparation, baking and boxing of pizzas taking 12-13 minutes, Indian deliverymen have 8-10 minutes to ferry their piping hot cargo to its destination – leaving a margin of just a few minutes. Riders cannot race to their destinations either: their motorbikes are modified to restrict their maximum speed to 45kph. That means riders must know every street, pothole, traffic light, choke point, construction site and police roadblock in their sectors of fast-changing, densely populated cities. …
Of all Domino’s deliveries in India, less than 0.5 per cent take more than 30 minutes to reach the consumer. Top managers monitor every store’s late rate closely. Rising pizza giveaways are seen as an indicator that a store is being overwhelmed by rapidly growing business – and that the area may be ripe for an additional outlet – or that local congestion is worsening considerably. “We watch that number like hawks,” Mr Kaul says.
Now there are obviously several steps in making these deliveries happen — from making sure that the kitchen staff is well-trained to scheduling enough delivery drivers. The most interesting part to my mind is the last thing hinted at in the quote above: How does Domino’s think about locating stores — and defining their service areas — so they can hit their delivery window?
Here’s what the article says:
It took Jubilant about a year to prepare to make the offer publicly. Every store’s geographical delivery area was first carefully analysed and customers more than a reliable 10-minute ride away in peak traffic were informed they would no longer receive deliveries but receive discount coupons for in-store use instead.
Implementing the guarantee influences Domino’s entire operation, starting with site selection for new outlets. In neighbourhoods where it is setting up shop, the company looks for locations with easy access to the greatest number of buildings within 10 minutes. “We do sorties,” Mr Kaul says. “We get on a motorbike and we do runs at 40-45kph at some of the toughest times of the day.”
Every year or so, each outlet’s geographical service area is reassessed and reduced if fast-growing business or worsening traffic makes it too hard to reach customers on time. “In which other business do you say no to business, to a genuine customer who wants your product?” Mr Kaul asks. “But that is what we do.”
I like this quote because it relates to one of my favorite service operations articles, Managing a sprawling service business (McKinsey Quarterly, December 2002), which present a framework for just this kind of situation. Check out this diagram taken from the article:
Start on the left. This panel plots operating margin for different resource levels as a function of sales. It basically is taking a fixed number of machines and people and looks at what would happen as sales ramp up. Holding resource costs fixed, more business spreads the cost of resources over more thinly and margins climb. Fewer resources here always do better so as we move down the dashed curves, we have added resources and hence lowered margins.
The second panel recognizes that you can only do so much with, say, two vehicles and four delivery personnel. At some point, you’ve maxed out your resources and need to add more equipment, more people, or both. Just where you max out depends on how you’ve defined good service. Better service requires more resources so you fall on a lower resource curve for a given level of business. That’s why the red line corresponding to having 90% of deliveries on time falls below the blue curve for having 80% of deliveries on time.
The key observation is that the operating margin takes a discontinuous jump down when resources are added. This clearly makes sense. Suppose having sales move from an average of 100 orders per hour to 101 orders per day requires an additional vehicle. That means a fairly small change in revenue requires a big increase in cost. That suggests that any increase in demand is not necessarily good news. If you are serious about hitting your promised service level, a small increase in demand could result in a precipitous drop in margins as capacity is expanded but utilization drops. The firm then faces the choice of refusing some demand (as Domino’s does) or trying to scrounge up enough additional sales to raise utilization back up. The good news here is that there are generally economies of scale in operating service systems so that it should be possible to increase sales enough to get utilization back to acceptable levels before having to expand resources again. (As I type this I am waving my hands and making some assumption about there being a singe bottleneck.)
Two last points on this. If instead of plotting margins one plots the cost per transaction, one can then show that the jump in cost per transaction is proportional to 1/N where N is the number of servers in the system. That means these jumps in margin are significant in setting the staffing levels of delivery drivers at a Domino’s or the number of people behind the counter at Starbucks but trivial at call center with 500 agents.
Finally, this approach of turning away all demand from an apartment building that is deemed too distant is arguably crude. Instead, one could choose to turn down any order when the wait would be too long. That may sound crazy, but it has been tried. Super Fast Pizza of Fond du Lac tried to deliver pizzas super fast (note the past tense — I am pretty sure they’ve folded). These guys had a nutty approach; they would bake pizzas in a van while driving to the customer (Super Fast Pizza slashes delivery times, Pizza Marketplace, Feb 21, 2005).
Matthew set out eight months ago to develop a revolutionary cook-and-deliver pizza system that would slash delivery times and increase product quality. By turning ordinary commercial vans into rolling pizzerias, Matthew’s crews put hot pies before customers’ eyes in 18 minutes on average. …
Menu options are limited to seven pizzas: deluxe (multiple meats and veggies), sausage, sausage and pepperoni, five cheese, four meat, veggie, and a pizza of the month. …
Using a wireless Internet connection, orders are transmitted to vans in the field, an alarm rings and they’re printed out. The driver — who works solo — goes to the van’s kitchen area, pulls pre-made pizzas on parbaked shells from the cooler and places them in one of five concession-stand-style pizza ovens which cook at 600 F. (The well-secured kitchen equipment runs off onboard generators.) He then returns to the driver’s seat and sets off to the delivery point. …
During a typical weekend rush, the computer-controlled system is set up to refuse orders if drivers get backed up. “Our software is written so that when business gets beyond a certain number of orders, it actually refuses orders. … Those customers automatically get $2 off their next order. And invariably, we find an order from those same customers the next day.”
So they turned down orders when things got too busy and they couldn’t commit to being super fast. And that raises an interesting question: Which is better in the long run, telling customers that they are no longer in the delivery area of what they thought was a local shop or occasionally telling someone across the street that you can’t their order tonight? I suspect that the former is easier to manage. You may upset once loyal customers but you are only doing it once and the whole point is that you won’t be doing business with them again. Turning down orders only when you are busy makes sense mathematically. It is probably a better way of assuring reliable delivery for the orders you take. But it only really work if customers understand what is going on and appreciate that they will benefit on nights their order is accepted.
SEE ALSO: How An American Banjo Maker Stood Up To Cheap Chinese Competition