“It’s better to be roughly right than precisely wrong” is a saying that is often attributed to John Maynard Keynes, but it was coined by Carveth Read in the 4th edition of his book “Logic: Deductive and Inductive” published in 1920 where he wrote:
“It is better to be vaguely right than exactly wrong.”
Roughly right versus precisely wrong has since become a best practice call-to-arms in forecasting. It encourages fresh activity over static analysis. The message is that it is better to guess, review and adjust quickly, and to start again immediately than to take time and refine over a longer period.
Being roughly right means getting close enough to a valid forecast, and being precisely wrong means spending too much time trying to make the forecast reach an accuracy it can never achieve. But are these terms correct?
Semantics and Principles
If we lay the options out in a line where wrong is on the left and right is on the er… right, it will look like this:
Precisely Wrong | Wrong | Roughly Wrong | Neutral | Roughly Right | Right | Precisely Right
Let’s assume we start from a neutral position in the center and are skilled enough to enrich the forecast with insight and adjustments. If we do this, the first stage of improvement will be roughly right. If we continue enriching, it means that the next stage of accuracy is right, and if we keep going, then ultimately, it becomes precisely right. This outcome is unlikely in demand forecasting, but that’s the logical progression.
If we are not so skilled at enrichment (and our starting point is neutral again), then adjusting poorly means that the first stage is roughly wrong. If we continue to adjust poorly, then it becomes wrong, and then after that, it becomes precisely wrong.
Of course, this is just semantics. First, it is entirely feasible to make a massive blunder and zoom straight from one end of the accuracy scale to the other. Second, the starting point of neutral is too neatly unbiased for the real world. Third, it’s unfairly weighted in that demand planning predictions will never be right, let alone precisely right. And this is the point of the saying; at the point of consumption, every categorization to the left or precisely right will be precisely wrong.
‘It is better to be roughly right than precisely wrong” brings to mind another famous planning statement: “the one certainty about your forecast is that it will be wrong.” This saying is somewhat trite, but we should look at these sayings as principles rather than failings. Appreciating that it is better to be roughly right (since, ultimately, everything will be precisely wrong) and accepting that it’s okay to stop adjusting can open opportunities to improve your forecast output and frequency.
The Effort of Forecast Enrichment
It is quite common for analysis and adjustment to be deeply ingrained in a forecast cycle. Each planning contributor believes that they are the critical factor holding the forecast together and that their insight is essential. But are you being “busy fools” or are you adding value?
The critical lesson here is to measure the impact of your adjustments. You need to understand when:
- You are no longer enriching or applying value add to the forecast.
- The effort required to enrich is greater than the improvement likely to be obtained.
Your forecast cycle should allow enough time to make adjustments that have greater value than the effort to apply them. Identifying this sweet spot where enrichment is maximized is a powerful insight that will make your forecasts better since you know it is time to stop adjusting.
The Efficiency of the Forecast Cycle
Critical data can change while deep analysis takes place, and the risk is that over time a forecast becomes less valuable and potentially even damaging. Addressing this crucial change of focus from accuracy to re-assessment essentially means changing the forecast cycle time.
Forecasting more frequently is not so easy to do, of course. Dirty data that needs cleansing, statistical forecast engine run times, exception management analysis, consensus agreement meetings, and approval procedures are a few examples of why faster forecasting can be a very difficult thing to achieve.
However, consider your forecast cycle in light of enrichment. When you know that a planning process step is no longer adding value (or the value is exceeded by the effort), then the step can be made shorter or even removed altogether. Consider that maybe the forecast cycle should be longer for some combinations. If value is not being added at all with repeated forecasts, then perhaps those intersections should be left alone for a longer period.
In summary, using the phrase “It’s better to be roughly right than precisely wrong” as a principle and understanding that your forecast will be wrong elicit these two key questions:
- Is the effort of forecast enrichment worth it?
- Can the forecast cycle become more efficient?
Of course, the answers depend on business type, culture, data, systems, resources, cycle time, levels, allocation, forecast type, measurement, and a whole host of other things. As planning maturity increases, the answers to these questions will change.
With the o9 Digital Brain, we can help you improve accuracy and reduce the time spent in forecasting. This is achieved through a combination of data, process, and technology. With the right data drivers, machine learning, collaborative planning, and cloud functionality o9 will transform your planning.