Certainty in forecasting is the holy grail, being certain of the future means success. However, as we know the only thing we know for certain about the future, is that it will not be the same as the past, or present.
Quantifying uncertainty appears to be an oxymoron, but reducing the degree of uncertainty would be a really useful competitive outcome.
When you explicitly set about quantifying the degree of uncertainty, risk, in a decision, you create a culture where people look for numbers not just supporting their position, but those that may lead to an alternative conclusion. This transparency of forecasts that underpin resource allocation decisions is enormously valuable.
How do you go about this?
- Start at the top. Like everything, behaviour in an enterprise is modelled on behaviour at the top. If you want those in an enterprise to take data seriously, those at the top need to not just take it seriously, but be seen to be doing just that.
- Make data widely available, and subject to detailed examination and analysis. In other words, ‘Democratise’ it, and ensure that all voices with a view based on the numbers are heard.
- Ensure data is used to show all sides of a question. In the absence of data showing every side of a proposition, the presence of data that emphasises one part of a debate at the expense of another will lead to bias. Data is not biased, but people usually are. In the absence of an explicit determination to find data and opinion that runs counter to an existing position, bias will intrude.
- Educate stakeholders in their understanding of the sources and relative value of data.
- Build models with care, and ensure they are tested against outcomes forecast, and continuously improved.
- Choose performance measures with care, make sure there are no vanity or one sided measures included, and that they reflect outcomes rather than activities.
- Explicit review of the causes of variances between a forecast and the actual outcomes is essential. This review process, and the understanding that will evolve will lead to improvement in the accuracy of forecasts over time.
Data is agnostic, the process of turning it into knowledge is not. Ensure that the knowledge that you use to inform the forecasts of the future are based on agnostic analysis, uninfluenced by biases of any sort. This is a really tough cultural objective, as human beings are inherently biased; it is a cognitive tool that enables us to function by freeing up ‘head space’ reducing the risk of being overwhelmed.
Consistent forecast accuracy is virtually impossible, but being consistently more accurate than your competition, while very tough, is not. Forecast accuracy is therefore a source of significant competitive advantage.
Header cartoon courtesy Scott Adams and his side-kick, Dilbert.
Forecast in cartoons
