|Tesco interview and demand planning
||[Feb. 2nd, 2016|12:29 am]
I have a major interview coming up, on Friday, which is quite exciting. It's for a job at Tesco trying to forecast demand. In short, Tesco pioneered (at least in the UK) the practise of using engineering and science based graduates to dip into their past data (which they've always been very good at collecting at the checkout) and then using that to commit forecasts about shopping behaviour.
I've been doing research on the topic because they asked me to, but I've been finding some of it quite interesting. There's lots of things you wouldn't expect once you start delving into what they can find and know.
For example, certain goods are very much bought at the start of summer or winter. The first warm period marking summer sees a surge in, amongst other things, meat for barbecuing, while vegetables drop. On the other hand, cocoa and cat litter fly off the shelves in the first freeze of winter (cats hate going outside to pee in winter, something our cats get no choice about).
Equally, ice cream responds to temperature, though it is mainly receptive to changes in temperature from previous days (if it's hotter than the day before) and tops out at about 24 degrees C (whereupon people shift to ice lollies). Also, Glasgow considers 15 degrees C as being very hot for data purposes and you can see the same reaction at that kind of temperature.
They seem to, in particular, want me to do this for fresh food, which I've been looking into. This in itself is maddeningly complex, because fresh food is itself a broad church.
Some things behave a bit like bell peppers: you grow them year round in greenhouses (either in the Netherlands or Spain) and the demand you think you need you communicate to the supplier and they try to grow whatever they need. Some things are like apples: they're grown once, preserved in weird gasses and then trotted out when you need them during the year. Other things merely have a season and that's it. Some things will be flown in from all over the world (pineapples), some will not.
Also included is meat and fish, which also have varied and different ways of being dealt with. Beef, for example, needs 3-4 weeks to cure properly while chickens seem to be slaughtered to order. Fish, on the other hand, will mainly take about 10 days at most to get from the sea to us and may be frozen, may be not (and legally must be if used for sushi to kill pests). It's on ice a lot of the time, whatever happens.
All of these things make the demand process pretty fraught. If you could predict the demand, would it do you any good given that the supply depends on when you predict the demand? I don't know, but it's all turning out to be a fascinating problem.
If that's not enough, then there's also logistics elements that I've been looking into, which seems to be like playing a game of OpenTTD, but for real. So the length of time you take to get your stuff from supplier to shop also matter a lot, as the path you take.
Tesco aim to reduce waste by removing as many unnecessary trips from their supply chain and try to basically run the company as if they were the Toyota of Supermarkets, embracing Just In Time philosophies and reductions in wastage, as well as techniques that try to respond to errors.
Demand planning, of course, plays into that, but it's just a small part overall of what's going on. They seem to have pioneered a major logistics chain, which I'm not 100% certain about how it works, but it seems to be that they can get everything into their major distribution centres and use those to push the goods into where they need to go, using RFI to track shipments through the system and eliminating storage by cross-docking, sending stuff from arrival straight to it's destination. The data challenge involved in all of this just is quite exciting and makes my head hurt just thinking about it.