Wiki

Predictive Analytics

The crystal ball of the 21st century

Making predictions and discovering connections: these are the two objectives of predictive analytics. Companies increasingly use such analysis to predict future customer behaviour. Like for instance, the ability to predict the likelihood that your customer moves over to the competition. 

Doing business is looking ahead. This is exactly what predictive analytics does. Finding out how likely it is that a particular event will occur in the future and the impact upon your business.

The growing popularity of this analysis is due to two IT developments in recent times: cloud computing and open source software. In cloud computing, data is stored and processed through the internet. It means that companies no longer need to buy expensive hardware or software.  Instead, they “lease” external computing and storage capacity at a time that is the most convenient to them. An additional advantage is that companies no longer have to bother about (costly) maintenance.

Predictive analytics starts with data

Predictive analytics depends on data. The data can be put into various categories, as indicated below:

Historical data versus real-time data

Many insights originate from “historical data”, i.e. stored information relating to the past that is studied to detect connections. Increasingly, companies are also analysing real-time information such as GPS data from consumers’ mobile phones.

Internal versus external data

Over time, companies generate large volumes of data about their customers: customer loyalty, number of contracts, total contract value, turnover per customer, customer visits, complaints, etc.

You can supplement this in-house data with external data. For instance, you could collect data about your competitor's new product. Or pin point comments on social media. And don’t forget market information. This is because, for instance, it is of crucial importance to know whether you are in a shrinking or expanding market.

You can make predictions on the basis of this data, so the more data you collect the better. The accuracy of predictions increases with the size of your data collection.

From big data to smart data

How to predict

Broadly speaking, there are two possibilities for making predictions.

  1. You can collect data yourself and conduct predictive analysis.
  2. Or you can increasingly buy predictions and make use of an increasing number of SaaS providers such as, for instance, INfer or Fliptop. They generate a predictive lead scoring through plugins without you having to provide any input yourself.

When making a prediction, you determine the likelihood that a particular event will or will not materialise. For instance, you calculate the probability that customer X will leave you in the years to come.

At the same time, you try to discern connections or correlations among different variables so as to be able to control particular findings. You can, for example, ascertain whether there is any connection between a customer’s turnover and the number of visits they receive.

Scores predict customer behaviour

For instance, you want to find out how likely it is that a particular customer will cease doing business with you. To this end, predictive analytics examines all variables that may possibly affect your collaboration with the customer. As an example, the number of complaints in recent years may have a significant impact, but your future collaboration will also be affected by the number of customer visits and email traffic. A software program examines all these data for connections. The result may be a prediction based on a score, e.g. the chance that customer A will leave you this year is 60%.

Predictions need not be confined to existing customers. Using predictive analytics, you can also predict how likely it is that particular prospects will actually become your customer. For this purpose, the prospect’s profile is compared with that of current customers. The more similarities there are between the two, the greater the chance that the lead will do deals with you.

Increasing accuracy

The more frequently you use predictive analytics, the more reliable the results. This is because you will be able to check whether the predictions have turned out to be correct. You can then add these findings to your current data set. It means that the system will become smarter every day and the predictions more accurate.

Another advantage of predictive analytics is that you do not know in advance what connections will emerge. It may well show that complaining customers are much more loyal than you think.

From predictive to prescriptive

There is no point in just collecting data. Correlations and predictions are of no use if you do not act upon them. This is why you will have to change over from predictive to prescriptive. In other words, you should take action through which you can counter or enhance a particular prediction because you know the outcome in advance. A number of inspiring examples are given below.

Hoarding honey cake

Wal-Mart, the American supermarket chain, has made a very odd discovery. Sales analysis showed that consumers stock up on gingerbread when there is a storm approaching. Wal-Mart had noticed earlier that people buy flashlights and batteries before a heavy storm. Wal-Mart used this information to replenish their stocks. When weather services issue a gale warning, Wal-Mart fills its shelves with extra flashlight batteries and now also with gingerbread.

Predicting company closures = preventing loss

Graydon’s Discontinuation Score predicts for each company the likelihood that it will cease to exist within the next 12 months. This score is computed on the basis of a number of variables, including the company’s past payment record, annual accounts and “demographic” factors such as age, size, legal form and sector. Additionally, the score takes account of “exceptional patterns” that affect the company’s health. Think for instance of frequent changes of management or the involvement of directors in earlier bankruptcies or closedowns.

The information about company closures is of crucial importance to Finance. Customers or suppliers who cease doing business have an impact on the company’s cash flow. In the past, the assumption was that about 5% of companies cease to exist each year. With a turnover of £1 million, this meant that there was a risk of £50,000. The Discontinuation Score is much more refined. The score makes it possible to determine at debtor level which part of turnover is being threatened. The score could show that debtors with an increased chance of closedown on average have a higher turnover. In this case, you would have to take a possible loss of £80,000 instead of £50,000 into account. These findings are also of interest to Sales and Marketing. By not spending any budget on companies that are about to disappear, the departments can raise their profitability.  

Predictive analytics ends with application

What you do with the findings ultimately determines the added value which predictive analytics can offer. The strength of the predictive value partly lies in the quality of the data. This is because it is not easy to keep up the level of quality of your data. It requires expertise to translate findings into opportunities and threats and then translate these into processes and action.

Sign up for the latest downloads