I don't hate spreadsheets. Let me make that clear. I've used them for over 20 years and found them to be very useful for all kinds of things, and they always will be. But what I want to talk about here is a mindset that I think we have slipped into that is characterised by an unjustified confidence in our ability to model a situation and predict a future outcome.
The epitome of this is “goal seek” or “solve for X” – In other words, build a model that represents your "problem" and manipulate the variables to maximise your favoured outcome.
How many business cases have been justified based on a spreadsheet that “proves” that if we do this thing, then we will get this much value (at some point in the future). But do things ever turn out as such in reality? I’ll bet you a beer that they don’t. The value might be better or worse but, more importantly, I guarantee there were a great deal of other factors that were missed out.
There is an equivalent of this process in the natural world which serves as a good comparison on the dangers of spreadsheet thinking. It’s called Trophic Cascades – or, powerful indirect interactions that can control entire ecosystems.
A great example of this phenomenon includes the reintroduction of wolves in Yellowstone National Park in the early turn of the millennium that had the knock-on effect of dramatically increasing the beaver population.
If you were a cattle rancher or livestock farmer in the early 1900's you wanted to maximise the amount of meat you could sell (this is your X). Wolves (W) eat meat, therefore driving W towards zero positive influences X. Simple right? The last wolf was killed in 1926 and sure enough it worked – or at least it appeared to have worked in the short term.
- Can AI play a disruptive role in the spreadsheet ecosystem? (opens in new tab)
A lesson learned
In this case, Elk populations increased but they also caused all sorts of other problems, including a reduction in the number of beavers because of the loss of willow trees lining the rivers. Coyote numbers also shot up, so numbers of Coyote prey went down (rabbits etc.). It was anything but a targeted intervention and it completely changed the ecosystem. Wolves were reintroduced in 1995 and there are now nine beaver colonies instead of just one. It seems obvious now, but at the time the cattle ranchers made a convincing argument and the government listened to their case and subsequently supported wolf eradication.
There’s another lesson to be learned from an example which is closer to home; that of arable farming in the UK. Farming deemed the most ‘efficient’ here is clearly unsustainable nowadays. If you are an arable farmer your goal is to maximise the yield from the land you have. This has typically been achieved by:
- Is spreadsheet risk management a better fix for excel user woes than AI? (opens in new tab)
- Removing hedgerows and making fields as big as possible so they can be worked easily by machines
- Using fertiliser to increase the growth rates of crops
- Using pesticides to kill all the things that eat your crops
Over the long term, unexpected effects have started to appear as a consequence of these actions. These include increased soil erosion, the pollution of rivers and the decline in the bee population. We now have warnings that the decline in bees could wipe out the British apple industry altogether, for example. The problem, of course, is that these interactions were not incorporated into the original model.
We can't keep on making these catastrophic environmental cock-ups, and planning the next phase of growth will have to take into account the wider impact on the environment.
Which brings us back to spreadsheets!
These and other similar problems in the natural environment arose from a “Solve for X” mentality of “spreadsheet thinking” - where we assume that we have correctly modelled the system and X is the outcome we want, but we have no way of knowing all the side effects of this solution.
The defence of the spreadsheet author is to say “well, how could I ever have predicted that”. And they’re right, it is almost impossible to predict these things by just sitting at a computer and writing a spreadsheet. But it doesn’t escape the fact that they happened and it would be much better to know about them than not.
We build spreadsheets to answer the questions we want but they will never tell us all the other questions we should have asked. In practice, it is actually very difficult, if not impossible, to properly model a complex system (at least cost effectively).
Rather than pretend we have correctly modelled the system we want to change, we need to find a way of creating a small disturbance (a perturbation), measure the outcomes and check that they are actually the outcomes we actually want.
Inspired by nature
- How can work management tools help your journey to digital transformation? (opens in new tab)
The trick is to make small reversible changes first before we commit to a big irreversible one and validate our models with real world information as soon as possible.
Making a business case for an investment based purely on a spreadsheet with a financial value oversimplifies the problem. It doesn’t have a way of valuing other things like quality of life and staff retention, as two of many examples.
We should take our inspiration from Nature - organisations are very complex systems just like ecosystems. If we want to transform the way people work, we need to go step by step and understand all the interactions in the ecosystem. The lesson here? Make the small changes and measure the outcomes before we make the big ones.
Murray Callander, Founder, Eigen (opens in new tab)