|
Using 4Thought to model the effect of your sales force In this example a model is built to estimate the sales value of each county, based on factors such as the population density and the number of retailers per capita in that region. In a similar fashion, the value of a particular account (instead of county), efficiency of a process, or performance of a retail network can be estimated. A giftware manufacturer in the UK distributes his products using a direct sales force to promote goods to retailers nationally, with each member of the sales force covering a number of counties. The manufacturer wants to identify which counties show greatest scope for improvement, but this is a difficult task because so many factors beyond a salesperson's control influence the sales potential of a region. Even if the manufacturer does feel that his presence could be increased in a particular region, he has little idea whether this is attainable by seeking out new accounts or whether the salesman should try to increase the value of the existing accounts. This case history shows how 4Thought's analysis of his business enabled him to offer his Sales Force constructive help in building the value of their work and to offer praise where it was deserved. As with any problem that might be solved with the 4Thought approach, the first step towards the solution was to decide what the yardsticks were that were a true measure of business performance (in modelling terms, the output variables of the model). In the giftware manufacturer's case, it was important to understand both the entire sales potential of a county and the expected value of a typical account in that county. The second step was to identify, using business experience and common sense, what the factors were that made these yardsticks vary from county to county (the input variables). Of the many factors affecting the manufacturer's business, the key factors driving variations in sales seemed to be: population, population density, tourism spending, number of retail outlets in the county and consumer expenditure. In deciding on the best input factors, it's always a good idea to standardise to remove systematic variations. For example, if you're dealing with a long time series, it's best to divide prices by an inflation index; in this case, it was best to work with per capita population rather than total numbers. Thus, data that was gathered and fed into 4Thought for each county was:
In this case the giftware manufacturer wanted to model sales per capita and sales per account as a function of the other variables. Using state-of-the-art mathematical modelling techniques, 4Thought identifies patterns in data, representing a big advance over standard statistical techniques because it can identify what is a real pattern and what is just a pure coincidence. Initially, 4Thought makes no assumptions about relationships between the input factors and the output. After a while, it identifies the broad features in the data. All this time, it sets aside a proportion of the data (the test set) which it uses, every step of the model building process, to test the predictive ability, and thus the usefulness, of the picture it's putting together. As 4Thought progresses it narrows down on finer and finer levels of detail, until it gets to a point where it begins to think it can see patterns in the random fluctuations in the sales figures. The problem of seeing such mirages in data is common to all statistical techniques (and a large number of sales managers!). Note, in the diagram below, how the test fit starts deteriorating after 53 steps. The neural network starts to "overfit" by finding patterns in the modelling set of data that are unrealistic, and are not applicable to the test set.
4Thought, however, is unique in being able to overcome the problem. Combining the broad-pattern-to-fine-detail approach with constant monitoring of predictive ability, 4Thought identifies exactly when the most information has been drawn out of a set of data, and thus when to quit while the going is good. Any other approach will yield a model which looks good statistically, but doesn't translate into achievable results when you use it for real-world management decision making. (4Thought's approach is particularly powerful in applications like Sales Force optimisation, where there are many factors affecting the business and there are only a handful of figures to work on). The value of 4Thought's analysis is first realised when the model it has built is cut apart to reveal how each factor really drives sales in the regions. These business insights are interesting, but 4Thought shows its real value when this information is turned into business benefits. The manufacturer used 4Thought to grow his business in two ways: 1. By monitoring salesperson performance and identifying sales improvement targets, county by county
2. By identifying not only the scope for improvement, but whether to achieve it by opening more accounts, by selling more to the same accounts, or both
You could use the same approach to improve your own Sales Force effectiveness. The factors involved might be different, but that doesn't matter to 4Thought. You provide expert knowledge of what governs your business, 4Thought provides expert analysis, and together you can see the way to build an even more successful sales team. Additional Cases
|
