The Science of Sales Forecasting
Category : Blogs
Predictive analytics is not a new technique. It has been around for years – especially in research and mathematical fields. Predictive analytics is an area of statistical analysis in which information is extracted from data, then analysed to identify patterns and make predictions about forthcoming events.
The End Of Guesstimating
A combination of statistics, machine learning, data mining and modelling is used in order to achieve this.For a long time marketeers have tried to predict what will happen in the future. They’ve used their knowledge of past events, and personal experiences, to try and reduce costs and increase sales.The rise of predictive analysis has meant that this process of forecasting is becoming less about guessing and more about calculating. This makes predictive analysis highly valuable for marketing and sales.Imagine having the ability to accurately predict, not only who the best leads and prospects will be, but when and through which channel they will most effectively be reached. In turn you could better understand when your current customers are most likely to order again. You could also assess the impact that changing your promotions will have on a particular sale.This empowers marketers and sales teams to be more productive and, ultimately, more profitable, than they ever have been.Done properly, predictive analytics can transform the science of sales forecasting from a game of dart-throwing to an exercise in precision.
In Data We Trust
Aberdeen Group’s research (Predictive Analytics for Sales and Marketing – seeing around corners) confirms that companies who invest in predictive analytics outperform those who do not in two key areas: incremental sales lift and high engagement (click through) rates.There are a number of different modelling techniques that can be used to help predict behaviour, for example:
This can be used to extract information from data and then to identify patterns that highlight trends in behaviour. It takes into account previous data points to predict future events.It is currently used widely in stock markets, but can be used in any field in which future movements are important. For marketers, having an understanding of when demand for certain types of products is likely to peak, or dip, helps optimise spend and staff levels.
This can be used to make predictions about the future by estimating the relationship between two or more variables, and is used readily in medical and scientific fields.Using this in marketing would mean building a regression model based on sales, prices and promotional activities. The output from the model could provide precise answers related to what would happen to sales if, for example, prices were increased by 5%.This would offer insight into the profitability of an increase in price – would it balance out any potential loss in sales, for example?
This can be used to group customers, or other subjects, into similar groups to help predict behaviour. It is often used as a way to segment customers into similarly marketable cohorts who then receive different marketing content based on that group’s specific behaviours.It can also be used to make recommendations based on what other people in the same cluster have bought or are interested in.
This is a form of predictive analysis that can be built to, not only understand the current attribution of marketing spend, but to predict future strengths, weaknesses and trends across different marketing channels.This plays an important role in optimising marketing spend and maximising returns.Consider, for example, a marketing campaign that sits across a number of different channels – display, TV and radio. It is entirely possible that someone is initially driven to the site through listening to the advertisement on the radio. They then search for the site and click on the organic search result. The next day they click on the PPC link and purchase from the site.By tracking, and using last click methodology, it will look like the conversion came from the PPC advert – but in fact it was the radio advertisement which drove the customer to purchase. Modelling all of these interactions effectively will provide a better understanding of which marketing channels are most effective.
Despite all these positive uses for predictive modelling there are still a number of ‘watchouts’ which should be kept in mind when thinking about how these techniques can be employed.Firstly, you need to have a relatively rich amount of customer data to be able to perform predictive analysis well. And the results, of course, will only be as good as the data that is inputted.Secondly, if the predictions are based on human behaviour (as they would be for many marketers) one has to consider that humans are not completely predictable. There will always be environmental variables and fluctuating human responses to them, which are inherently unpredictable due to the (often) irrational behaviours of individuals.This is where an analyst’s understanding and experience of the market is an essential counterpart to any work in predictive analytics. We must watch out for these unpredictable elements and ‘sense check’ any models results. There is no substitute for common sense.Predictive modelling can be used to help optimise a marketeer’s time and spend. It can help to anticipate trends in the market and point to which variables are most likely to effect a given promotion or campaign element.To maximise the potential of predictive modelling you also need to understand the market and to be able to interpret the data that is being used. It takes good data plus good judgement to create the perfect predictive model.