Previous Section
That is, do I really believe this? Do I really believe that my model works? Well, I really do believe that this model is working and the reason is that we've got some experimental data. The retailers designed this grand, classically designed experiment.
They took their stores, divided them into three segments, and randomly assigned a third of the stores to an 8 percent increase pricing treatment, which means they increase their prices by 8 percent. Another they left as a control; they didn't do anything to it. In the other third, they decreased prices by 8 percent.
In the Hi-Lo treatment:
What you see in that case, the Hi-Lo treatment is that there is no decrease in unit movement compared to the control movement. In terms of a price elasticity it means that price elasticity is zero, which if you look at some of my calculations, I'm thinking the price - category price elasticity is somewhere around -.5 or -.8, something in that range. If you look at it in terms of profits, by increasing prices 8 percent, they made 11 percent increase in profits. The difference here is they're thinking about 8 percent -- in the experimental design there were some problems and they didn't do all the prices uniformly increased -- they still have some of these features and promotions. The point is that if we would move everything up we would be doing better. It's not the 20 percent we were predicting, but it's still moving in the right direction.
In the EDLP treatment: The other thing is if they were to decrease their prices by 8 percent, the result is they would get a 3 percent increase in movement. If you compare this to what my models say, that somewhere around -25 is the category loss, here you are thinking about a category loss of 3/8ths. So it's not perfect, but it's still, this is 16 weeks and I've got 120 weeks.
Let's consider all category pricing changes and think about increasing all the prices in the categories, not just orange juice, but we'll do cereals and we'll do detergents and what happens is that for 40% of all the grocery products in the store, the prices were increased by 8 percent on average. The results are still consistent with this. Now, I'm not saying that it means that this always -- you can always come in with some catastrophic event or something that maybe causes consumers to start changing their behavior. But the point is that people are just get getting lost here - it doesn't really matter what prices. You can get away with some price increases here.
I'm going to step away from this and just take a look at the implications in the terms of the posterior profit, and then I'm going to ask if we can come up with some more reasonable pricing strategies. Because it's not going to be really satisfactory from my micro-marketing standpoint to say let's increase all the prices. In the beginning I said micro-marketing pricing strategies are very rich and they can allow you to change the gaps between products, they can allow you to change prices at the individual store level and have a lot of control. If I would just increase all my prices by 10 percent, my competitors might be able to take advantage of that. So I'm going to try to come up with some better ways to get around this problem.
But before I do that, I'd just thought I'd illustrate this difference in moving from the smaller to the weak prior. What's happening is that as you become less confident about these random effects at the store level, the profits are increasing; the reason is that as you become less confident, there tend to be more store differences and you tend to get more extreme points in elasticity estimates. That's one reason why you see this overall increase, but the point is that it's consistent across all these different categories.
Go to written version of paper