Forecasting propensity to buy - Microsoft Excel Tutorial

- [Narrator] Ultimately, one of the main goals when we're doing marketing research is simply to understand whether a potential customer or representative customer would actually buy our product. Would you buy a product with these features? Would you buy a product at this price? Would you buy a product from us verses our competitors? All of this comes down to, essentially, the propensity to purchase a product. Now, there's a key metric that's often used in marketing research, related to this and it's called the purchase probability score. This comes from some research that was done back in the 60s by Juster but the purchase probability score is essentially a scale that ranks the likelihood of a customer purchasing something, anywhere from zero, where there's almost no chance of them buying it, up to a 10, where it's virtually certain they will buy it. And the idea is, this is supposed to encapsulate some sort of a probability that a particular consumer would buy this product. We would ask a customer, very simply, in a survey, given this particular product and these features and this price, what's the likelihood that you would buy this product on a zero to 10 score? Now, an alternative approach to this is to use what's called a propensity score model. Propensity scores are, rather than something we directly ask a consumer about, instead derived from data. So the propensity score is used with a regression. It tells us the likelihood of something occurring, like the likelihood of a consumer making a purchase of this particular product. So essentially, it's capturing a similar concept to the purchase probability score but in a different fashion. When we look at various studies and analysis that's been done, about purchase intent verses actual purchases, we do see some interesting behavior. One of the big dings on surveys, from those who are skeptical about marketing research and in particular marketing surveys is they say well, people will answer in a survey however they think you want them to answer. That doesn't actually mean that's how they behave in the real world. Well, studies that have looked at this have found that's not quite true. In particular, when we look at different types of products, purchase intent fits more closely with what the customer says they will do in existing products than for new products. So if we've got an existing mop, as an example, and we ask a consumer whether they will buy it or not, we get a more accurate response rather than if we're talking about a new type of mop. Similarly, we get a better response for durable products than for non-durable products. If we're asking about say, a household appliance, we get more accurate results from our surveys than if we're asking about likelihood of buying software, for instance. Similarly, we get higher results or better results from our surveys when we're talking about specific brands or product models rather than broader categories. Asking about the likelihood of buying a Ford or a Tesla or a Chevy, we're going to get more accurate results rather than asking about the likelihood of buying a car. And finally, we get better and more accurate results when we're looking at short term time horizons, rather than long term time horizons. The closer that customer is to actually buying something, the more accurately we can capture their intent. Keep these rules in thumb in mind as you build surveys of your own, trying to capture purchase intent and model that for your purposes.