How predictive modelling works?


predictive analytics

What we are trying to do is predict the future by analysing the past.

How we do that is by going back in time. 

We need historical sales data. Things like time frame geography and market. For example to analyse all the homes sold. We choose a time frame such as a year. Next geography locations of where these homes are sold or not sold. We need to have a database that has all that information. We need know if single family homes. In other words as many recorded events as we can obtain.

We obtain last year’s data with the geography of everything that was sold and homes that did not sell. In addition, our data needs data points. Such as mortgage history, land registry (deeds, lease or freehold), solicitors, taxes paid.

We begin by analysing both sellers and buyers. We will break both buyers and sellers into a couple categories. One variable we may choose are investor’s the other home buyers another sellers. 

We break buyers and sellers down by their age their job, income lifestyle.

Even by what vehicles they drive.

How much disposable income they have. What kind of stuff they purchase. If they are married. What their interests are such as music, where they shop for groceries.

The aim is to understand both buyers and sellers motivations. Why they may be interested in homes that we sell. The frequency of properties purchased. We want to know whether it was a cash sale. Are they conventional people?

Have they purchased property to renovate then sell and move up the ladder?

If they are investors what type of investor are, they. Do they purchase in volume?

What type of properties are they interested in? 

Are they investing for a fund or small group or family members?

These are called data points.

When the data is added to the database it is referred to as fully enriched set.

We need data on the properties in the geographic of interest. New builds, properties rebuilt following a fire. Properties plagued by crime. Renovated properties. Properties with additions, conservatory, loft or cellar or conservatory added. Or garage converted into living space or living space built on top of the garage. Did these properties have planning permission? 

Assuming you can collate all this data. You might assume you cannot collate such data? You can. What you need is time and access multiple data sources. Some data will incur costs some is freely available.

So, what does all this really mean?

Next, for example we mark investor sales as orange. This variable we call our targets. Having all this historical data is like knowing what questions are going to be asked in an exam.

We build models for example our interest remains investors. We run an algorithm against our database. The first attempt predicted 15%. We know that prediction is not good enough. We sold more properties to investors than that. We build another algorithm and run that. We will need to reiterate this process until we end up with something that is the best result we can expect. This is referred to as optimised.

Next, we take all the properties everything that exists and run the algorithm. The algorithm spits out a list of scored properties. What that means every property has a score. This is known as a completely statistically optimised set. Next, we need to trim it down. 

Reviewing ROI property sales. We know that the top two percent of these sales refer to a statistically small part of the population. Obvious most are investors. We split this group off. This is called optimised target. If we choose, we can focus our marketing on this group.

An optimised target is not a mailing or telephone list or pay for click (PPC). Think of it this way.

Seventy percent of business will come from you list. Twenty percent from your offer, ten percent from your copy.

It is all about the targets. 

Think of it as a wheel. The target is the hub of the wheel. The target is the core. The most important thing. The target indicates where your marketing spend should focus. It may be mailing or telephone or pay for click (the spokes of the wheel). The percentage of time and money you invest in each spoke of the wheel will be determined by your target.

That is what predictive modelling does for us.

Key learning point.

The model is based upon historical data. The future may be very different to the past for example COVID 19 has thrown our whole economic model into the air. Nevertheless predictive modelling remains a powerful tool. 

About the author 

Christopher Bird

Building your own Power App, BI solution, or automated workflow can be a mind-blowing experience. It can also be a nightmare. Particularly when you begin with a blank screen. My advice, get professional help as and when you need it. That's what successful people do.

{"email":"Email address invalid","url":"Website address invalid","required":"Required field missing"}
Subscribe to get the latest updates