Something to prove, part 1

Things had been going well at Berger for the last year, with two projects for Volvo in America delivering extremely successful conclusions.

We had optimised Volvo’s sales incentive programme, comparing the effectiveness of their eight or nine types of sales incentive to see which boosted sales most effectively. These included straight payments to the salesperson, payments to their sales manager and the ‘stair-step program’ which encouraged salespeople to aim to sell a number of vehicles up to two different thresholds. Their payment per car if they were below the first threshold was quite low, but the payment per car increased if they reached the first threshold and increased further if they reached the higher threshold. So they could earn a huge boost in their sales commission if they reached the upper threshold. Car leasing has always been popular in the US and another type of incentive we analysed was ‘lease subvention’, where Volvo would pay the finance company to reduce the interest rate on the lease.

It turned out that the best performing type of incentive had the unintended side-effect of reducing the profitability of the dealerships, because the salespeople were so desperate to get their big commission that they discounted the price of the car, reducing the dealer’s margin. The dealers were not happy. So we followed up our first project with a second one, studying this in more detail and tried to find a happy medium that would sell more cars but also keep the dealers happy.

We were almost at the end of the second Volvo project. Suddenly, John appeared looking pumped! He had a new project for BMG Records on optimising the pricing of their CD albums (streaming didn’t exist yet) and he was particularly excited about a collaboration they had arranged with Virgin Megastore to get detailed pricing data. Virgin Megastore which, at that time, was a chain of large shops in the UK selling CDs and vinyl records, had an unusual policy on pricing, which was that the store manager set prices for the entire stock for their store, which were reset each week. This meant that there was far more variation of price for a given record than in other music retailing chains, with variation of the price by week and also by location. That was likely to make a huge difference to the quality of our analysis.

Sidebar on demand curves

When the price of a product changes, the demand for it, i.e. the sales level, changes. For almost all products, when the price rises, sales drop. Luxury goods like diamond jewellery or luxury cars sometimes are an exception to this, with demand actually increasing when the price is increased, because the customers perceive the price as a mark of quality, the poor deluded fools. But for most products, if you have the data, you can draw a graph called a ‘demand curve’, like this.

For those in the know, when drawing a demand curve it is traditional to put the price on the y-axis and sales on the x-axis, but it makes more sense if you put the price on the x-axis because that is the thing we can control, the independent variable. So the picture here is my Roland Berger version of the demand curve, not the traditional version.

When you include the data points themselves it looks more like this, where the dots show the actual sales points.

If we have a lot of separate price points for the same project, we can get a nice smooth curve like this, but if there are just a few points, as is more common, the best we can do for an individual project is to get a straight line for the demand ‘curve’, which is not much use for optimisation.

Back to the project for BMG. Simon Cowell, who went on to create Pop Idol, the X Factor and the various other reality TV shows, was on the Project Board. John kept me away from the clients because, after meeting Simon Cowell, I think he thought I was too naïve to be put in front of him and that he would eat me for breakfast in the presentations! Oh well. My strength at that time was doing the analysis, so why fight it?

We had started to use a separate company, which I’ll call FS, to help us with some of the database work. I was seconded to work there during the second Volvo project with a permanent desk and we were now adding the BMG project to the workload. There were a lot of very clever people there who taught me Excel shortcuts, which I had dreams about: ‘Type in the formula, Enter, Left, Ctrl-Down, Right, Shift-Ctrl-Up, Ctrl-d’.

At this time it would still have been dangerous to invite me to a party, though my statistics jokes were becoming quite good.

I had a meeting with John at FS about the details of the analysis and realised that, although John had a great mind for the perspective of the client and their business questions, his understanding of the details of the numbers and how we would answer the client’s business questions in terms of Excel formulas and database queries was less clear.

The project for BMG aimed to understand the effects not just of the varying price on sales of CD albums, but of the various promotional affects, such as the release of a single by the same artist or the artist being on tour. It seemed to me that we had to take account of the date of the promotional affect, such as the release of the single, and to watch for a boost in the sales shortly afterwards. The dates of the promotional stimulus were the crucial linking field.

My idea was that if we aligned all the sales datasets by the date of the promotional stimulus, we could then aggregate the sales datasets and that would show the causal connection between them. It was this idea that John couldn’t quite grasp.

I discussed the approach with the boss at FS and John even arranged a meeting with Richard Hoptroff, the very analytical author of the 4Thought neural network software we were using, to discuss my approach. Both of them broadly agreed with me, although they pointed out that this would add complexity to the analysis, but John would not be persuaded. Things became acrimonious and John told me that if we were to use my method, he was going back to the Roland Berger office.

‘Gideon, you’re on your own.’

I was really sad about the apparent disintegration of my relationship with John because I liked the guy and respected his judgement. He was my mentor at Berger and had drawn me out of my shell when I first arrived and was still thinking in equations and forgetting to eat lunch. Usually, I am not good at dealing with adversity. But for once I felt sure of myself and was certain that – if I could just put in enough hours of work, with the help of the people at FS – I would get some good results and might, just possibly, be able to repair my fractured relationship with John.

Next time I’ll talk about those hours, the nickname they gave me at FS, my long-suffering friends and the final outcome of the project. You can read that here.