Amaury Anciaux identifies an unrecognized problem that occurs in all organizations and provides steps to mitigate risk.
Why quantifying modeling risk matters
When testing the ideas that eventually led to River, I spoke with many analysts who build decision-support models in Excel. Almost everyone told me the same thing:
They know there’s a significant risk of errors.
They wish there were better tools.
But they assume that with enough late nights debugging, they can reduce the risk to an acceptable level.
The question is: is that assumption actually true?
As a scientist, I believe the only way to answer that is to define what the risk is, and then find a reasonable way to quantify it.
What do we mean by modeling risk?
For decision-support models, risk means:
The probability that the model leads us to make the wrong decision
A model may produce wrong insights if:
The input data is inaccurate.
The conceptual logic is not a good representation of reality.
The model implementation is incorrect.
The outputs are misinterpreted (e.g., scenario analysis mistakes).
In this article, we focus only on #3: errors introduced during model construction, assuming the other steps are sound.
Sources of modeling risk
Errors tend to come from three closely related factors:
Key points include:
- Scale of real world models
- Safe risk thresholds
- Reducing risk in Excel
Read the article, Quantifying the spreadsheet modeling risk, on river-solutions.com.