Hi, folks
When browsing various public sources we recently came across some fascinating research by Raymond R. Panko, Professor of IT Management and Shidler Fellow at the University of Hawaii. His special focus is spreadsheet errors and governance. One of his papers that caught our attention can be found here.
For those of you not wishing to spend time on the original material we will paraphrase a few themes from Professor Panko’s paper:
- Typical cell error rate in ANY spreadsheet is between 2% and 5% and it is a statistically high probability of over 50% that any spreadsheet has material errors no matter what. This is hard evidence backed by years of research
- Unlike software development industry where thorough testing is the norm, spreadsheet coders and users prefer to ignore reality and would be outright hostile to suggestions that their product has errors
- Spreadsheets’ modellers and users are equally bad with identifying errors, a basic check of a spreadsheet would not unearth more than c.15-20% of errors present in any given spreadsheet; more rigorous testing involving expensive and time consuming procedures would improve the error identification rate to 50% only
- However, software development industry which has a lot of similarities to spreadsheet coding and design does a lot batter with error rates of ten time less
How is it relevant to FinRobot? Well, directly.
As a software development company, we follow software development standards and procedures. Our templates are designed and rigorously tested before we release a new model. Once released our automated assembler does the job and no human is involved in the final product. In an unfortunate event, a rogue formula manages to escape our continuous monitoring and clients’ feedback allow feed into the algo of cleaned up code for all clients to benefit instantly. And once caught, an error does not come back.
Hence, we feel our error rate should be vastly superior to most models’ available on the Internet and especially to those where you paid a consultant to produce them. Remember that spreadsheet errors are created by humans.
Food for thought…