Statistical predictive modelling can help to take the guesswork out of estimating the true value of a site by optimising the mix of homes.
Take, for example a theoretical site that is going to be a mix of identical three and identical four-bed houses. Your own landbuying staff and local agent or surveyor may suggest a mix of 20 three beds and 20 four beds - but what about the order of building? Should the last 10 plots all be three bed, all four bed or a mixture?
And this is just a single decision. How much more difficult must it be to balance the product mix of a site with two, three or (realistically) 50 variables?
It is simply not possible to make these objective decisions with limited subjective information.
Statistical predictive modelling offers objective data that can be used to improve the ideal product mix.
Combining objective data from statistical predictive modelling with other, more traditional, sources of information can give a much better chance of reaching the ideal product mix for a site.
The appliance of science
Homebuilders begin with a headstart over many other businesses since they know the previous and current address of every one of their customers. Details of income at time of purchase and certain other facts are also usually known. Customers can be categorised in a variety of ways - initially by demographic (lifestyle) group and then sub-divided according to any other information held. The product also needs to be categorised. This means identifying, for every sale made, what the product was. Certain product attributes are obvious: number of bedrooms, square footage, with or without garage, level of specification, size of garden, etc. However at this stage it is not known which attributes actually influence customers’ buying decisions so any other attributes that might be an influence need to be included. Statistical techniques can then identify which of these product attributes has a significant effect on customers’ buying decisions. They can also determine, considering only the attributes that are significant, what the relationship is between each attribute and each customer category. The end result is a mathematical model that can make predictions regarding what a given type of customer, in a given location, is most likely to want to buy. This is the predictive model. The accuracy of predictions will depend on the skill of the consultant selecting the customer and product categories to work with and the skill of the statistician choosing the analytical techniques to use when creating the model. For each site a given percentage of customers will reside within a given radius of the site (typically 80% within five miles) and presumptions can be made regarding the geographical origins of the remaining 20%. From this geodemographic data a prediction can be made for the unconstrained buying pressures on the site. But further constraints need to be added to the model. For example:- the number of plots or the total square footage may have been predetermined;
- planning constraints may limit options;
- company brand position - may preclude building high or low value homes;
- technical or management skills of the company may mean certain building types are not favoured; and
- there may be site-specific constraints.
Source
Building Homes
Postscript
Andy Tebbutt-Russell is principal consultant with ATR Consultants which specialises in business processes, marketing strategy and IT. He can be contacted on: 01425 402315. email: andy@atrconsultants.co.uk