World would have been so good for marketers and decision makers if they could safely predict the future, the consumer behavior, market dynamics and uncertainties. Well, the good news is we are taking a nice step in this direction through the use of advanced analytics, predictive analytics, simulations, big data etc. The bad news is our predictions through various decision models may still come out far off from reality many a times. Some examples include retailers using big data to capture shopper history and their behaviors while they roam around in stores and using that to prioritize the product arrangements. Or travel companies gathering customer data from various channels such as kiosks, web and stores and consolidating it to provide useful insights around customer behavior.
I have built some of these models in the past. Through my research and understanding, I can suggest a few points to always remember when trying to use these models:
- Data is the foundation of these models. It should be clean, better if real time, standardized and unbiased. Better devote time to get the right data in place than fine tuning the model to improve the accuracy levels of your model.
- As this article points out, decision models can’t be used in just every scenario. Many times these models predict outcomes which the decision makers can’t influence, for example, weather forecast or quality of wine but executives are supposed to execute as well. Decision models are increasingly powerful for tasks requiring the impartial analysis of vast amounts of data. But when we can and must shape outcomes they may not suffice. We may be wise in relying models when estimating consumer reactions to a promotion. But motivating a team to achieve high performance is a different matter. Here, a combination of skills can be a probable answer.
- Prediction is just a prediction – an indication or pointer towards the possible outcome. Nothing can surpass your ability to influence the outcome through direct influence.
- The decision makers can even look at influencing the behavior of their teams through the prediction of these models to drive towards the desired outcome. For example, if the model predicts that growth rate might miss the target by 1% point, you may very well start planning around what can be done to achieve the target. The model’s output plays a different role here: shaping outcome.
- Wherever possible/applicable, use feedback to improve models. Dynamic improvement depends on two features: a) the observation of results should not make any future occurrence either more or less likely and b) the feedback cycle of observation and adjustment should happen rapidly.