“Hegel was right when he said that we learn from history that man can never learn anything from history.”— George Bernard Shaw
“I always avoid prophesying beforehand because it is much better to prophesy after the event has already taken place.”— Winston Churchill
Anyone who has been involved in business planning has been required to develop or acquire forecasts. In my experience, this vital task is often very difficult to do properly – that is, so that the forecasts are sufficiently reliable and credible to provide a basis for initiating costly major projects and new business directions. Best guesstimates are generally unacceptable.
Forecasting in the New Normal, whatever the New Normal may be in practice, is almost impossible to do with adequate reliability. Trends look back at a now-gone past. Speculation about the future can be iffy at best and highly misleading at worst.
Visionary folks like Bill Gates, Steve Jobs, and Jeff Bezos can rely on their internal sense for where things are headed, with credibility being based on the fact that they run the place. They are right even when they are wrong.
The rest of us however are obliged to tackle the typically complex mechanics of analyzing past data and using it to generate various kinds of projections. Then we have to persuade a variety of wise and sensible managers that our projections are indeed sufficiently accurate and reliable to serve as a basis for committing millions or billions of dollars.
This process has worked reasonably well for, well, almost forever.
It requires two critical conditions: a stable business environment and good data relevant to the future. Until very recently, these conditions were generally available. Better yet, data volumes had increased enough to permit artificial intelligence assistance in projecting and guiding actions.
And then, a very nasty tomorrow happened
You may recall that in the distant past of early 2020, things changed. Hugely, and likely forever. We have addressed the “new normal” that is anything but normal in several recent posts.
What we have today is fast-changing, unpredictable for the most part, and highly uncertain for what little we can see about what is happening.
This is definitely not stable in any useful way. But it gets much worse.
There is really no precedent for whatever is happening out there right now. This means of course that there are no data sets on which to base forecasts. Great – no stability and no data.
For a business planner or manager, things do not look good at all. Well, we can always fall back to doing whatever worked in past and hope that the past reappears ASAP or sooner. Are you willing to bet your business on this?
So, we probably have to make some serious changes but on what basis? Crystal ball is busted. But we cannot avoid acting so what to do?
First a bit of terminology: From Wikipedia:
“Forecasting is the process of making predictions of the future based on past and present data and most commonly by analysis of trends. Prediction is a similar, but more general term. Risk and uncertainty are central to forecasting and prediction; it is generally considered good practice to indicate the degree of uncertainty attaching to forecasts. “Forecast” and “forecasting” are sometimes reserved for estimates of values at certain specific future times, while the term “prediction” is used for more general estimates.”
Today, event and value forecasting with useful reliability is extremely difficult, and often impossible. The best we can do in practice falls mostly under the realm of prediction – making general estimates. Guesstimating is probably a better term.
There are however at least a couple of approaches that seem especially well-suited to our present non-stable and no-data world:
Two approaches: Experiment to learn + Develop scenarios
The future is unfolding daily whether we like it or not. That’s just life, or something. We have to deal with this unfolding somehow. Kind of like learning to walk with a blindfold. Fortunately, there is some quite good news here:
- You can act in small steps and see what happens. This is learning but only if you build learning into your action processes.
- You can develop some reasonable scenarios that appear to cover the range of what you think may happen. Then you track the scenario descriptors, such as sales or new orders, to see which scenario seems to best fit what is actually happening.
These can be very powerful management approaches in times of instability and unpredictability. You don’t have to try to predict what might be happening out there but only in terms of how it – whatever is going on – may impact your business. This impact measurement can provide both a situation-tracking metric and a rough means of predicting at least the near future.
How might this work?
Experiment to learn
Using daily business actions as a way to learn about the action-impact relationship can help guide your future actions: Try something, see what happens, and explicitly use what you just learned to guide your next steps.
You take small steps for at least a while – until you have gained confidence that you understand your business’ action-impact linkage. No big bets early on. Escapes and retreats built in as needed.
Note that you have to design the action process to include, explicitly, the learning part. Each action step should generate a description of expected and actual results, along with any further information that might be generated.
For example, suppose that your business sells primarily by means of workshops and large-group product demonstrations that are no longer allowed in an effective format. You might first try a Zoom-based version but learn that customers want to experience your products hands-on. This may lead to trying a product demo vehicle that can be easily set up for small customer groups in parking lots. Here, you may learn that this works but it is very expensive and slow compared to factory workshops with large customer groups. Your next step might be to build self-contained demo units that can be shipped to customers and presented via remote product experts. More learning, more experimenting, until you discover something that both works well and has acceptable costs.
The goal here is two-fold: actual selling and, now, learning how to sell in changing and uncertain business conditions.
Develop (and test) scenarios
Action planning still requires some sense for where things are heading in your business environment. It focuses on direction – as defined by some impact tracking metrics rather than background situation causes or driving events.
The goal here is to predict direction, at least roughly, in terms of how whatever-is-happening-out-there is actually affecting your business. In this respect, your daily business actions serve as experiments on the world that can provide you with useful information on future directions.
Your necessary business activities and their impacts reflect what is happening in at least general terms. But only if you have a way to extract this information.
This approach requires the use of a set of situation scenarios based on a few primary tracking metrics. A regional retailer, for example, might choose weekly sales, store traffic, and store cash flow. As we detail elsewhere, this might be elaborated to compare strong store groups against weak store groups.
Your situation tracking metrics will quickly begin to favor one or an adjacent pair of your scenarios if these scenarios are defined in terms of your tracking metrics.
Neither experimenting to learn nor scenario planning is difficult or complex. These can be quickly implemented to help you deal with whatever may be happening. You may over time gain some knowledge of your business situation’s primary drivers but such knowledge is not essential.
This approach is heuristic in nature:
“A heuristic technique, or a heuristic, is any approach to problem solving or self-discovery that employs a practical method that is not guaranteed to be optimal, perfect, or rational, but is nevertheless sufficient for reaching an immediate, short-term goal or approximation.”
Okay, so should we give up on forecasting and prediction entirely? Nope …
Even today, with its great changes and huge uncertainty, there are no shortages of predictions and speculation about what the future may hold. It is hard to resist simplistic trend extrapolation because it is easy and logical. But despite their appeal, trends:
- are a reflection of the past because all data is inherently historical
- don’t account for wildcards and unpredictable events (black swans)
- accuracy declines the further out we look
- extrapolate the past into the future, rather than creating new visions
- encourage an economic/tech-driven mindset
- tend to be simple and linear, and not holistic
- can be vague, misleading, or just plain wrong
- often capture short-lived and unique situations
If the past ever comes back, then by all means rely on trend extrapolations.
These deficiencies are shared by most traditional forecasting methods, such as theory, quantitative models, and trend extrapolation. All require good data but data reflects the past, not the future, except under stable conditions. This situation led to the development of several methodologies that can be applied effectively to unstable, unpredictable, and uncertain situations such as we have today.
The Delphi Method
Delphi, sometimes called “Estimate-Talk-Estimate (ETE)” is a systematic, interactive forecasting method that relies on a panel of “experts” and has been widely used for business forecasting. It is based on the principle that forecasts from a structured group of individuals are more accurate than those from unstructured groups, more recently framed as the “wisdom of crowds”.
The name Delphi derives from the ancient (ca. 1400 BC) Greek shrine for the god Apollo at Delphi, located in central Greece, on the side of Mount Parnassus. Delphi predictions, not surprisingly quite expensive, were made by the priestess Pythia rather than a panel of experts. This oracle was in business until the fourth century AD when Greek prophesying was abolished by the powers-that-were.
The non-priestess, panel-based, Delphi method was developed at the beginning of the Cold War to forecast the impact of technology on warfare. In a typical Delphi process, experts answer questionnaires in two or more rounds. After each round, a facilitator prepares a summary of the forecasts from the previous round as well as reasons provided for their choices. Experts can then revise their previous forecasts for the next round based on replies of other panel members. Hopefully, the range of forecasts will decrease and converge towards a forecast more accurate than any single expert might generate. Hopefully. And absent group-think.
An important weakness of the Delphi method is that future developments are not always predicted correctly by a consensus of experts. If they are misinformed or have limited knowledge about the target topic, Delphi may serve only to add unwarranted confidence to their flawed predictions.
Another problem is its common inability to make complex forecasts with multiple factors. Potential future outcomes were usually considered as if they had no effect on each other. Extensions to the Delphi method, such as cross impact analysis, address this problem by taking into consideration the possibility that the occurrence of one event may change probabilities of other potential outcomes. But, the Delphi method is used most successfully in forecasting single scalar indicators.
“It’s tough to make predictions, especially about the future.”― Yogi Berra (and maybe Niels Bohr)
Although not really relevant to this article, mention should be made of a specialized, recently developed, mass-participant prediction mechanism – prediction markets. In a prediction market you buy shares in a specified event outcome, such as the result of an election. As market-maker Augur explains:
“There are two types of shares in a prediction market: YES (long) shares and NO (short) shares. The ultimate value (payout) of each share depends on whether an event occurs or not. In a simple prediction market, each YES share pays out a dollar if the event in question occurs and pays out nothing if it does not occur. Each NO share pays out a dollar if the event does not occur and nothing if it does.”
Anyone can participate in these markets. Expertise not required, very different from Delphi forecasting. Data not required, which seems to fit situations where data is either unavailable or unreliable (as in the past predicting the future).
Wisdom of Crowds
James Surowiecki’s 2004 book The Wisdom of Crowds has become a bit of a classic in crowd-based forecasting and prediction. It cleverly refutes some of what a real classic describes – Charles Mackay’s Extraordinary Popular Delusions and the Madness of Crowds, published in 1841.
Surowiecki states that not all crowds are wise but only those that meet these five conditions:
- Diversity of opinion —Each person should have private information even if it’s just an eccentric interpretation of the known facts.
- Independence —People’s opinions aren’t determined by the opinions of those around them.
- Decentralization — People are able to specialize and draw on local knowledge.
- Aggregation — Some mechanism exists for turning private judgements into a collective decision.
- Trust — Each person trusts the collective group to be fair.
Diversity among participants is needed to ensure enough variance in approach, thought process, and private information to avoid group-think. This requirement extends to organizational diversity so that participants come from a variety of business units and levels. There should also be minimal or no communication among participants to avoid imitation and group-think. This avoids peer pressure, herd behavior, and collective hysteria.
In the absence of stable, predictable business conditions such as we have today, new approaches to business planning are vital. Fortunately, several of these are available even to smaller businesses. They can help prevent managing by flying blindly or doing what worked in the forever-gone past.
CPO Innovation, an online publication for top-level procurement & supply chain executives, looks at forecasting briefly from the critical supply chain perspective in its “Forecasting for the “New Normal” Post Covid-19“. It makes our point that backward-looking data is no longer of much use and may well be misleading.
FutureCFO, a website that provides strategic insights for financial executives, makes our current point in a somewhat different manner:
“The world has changed so much and so quickly that it has vastly impacted our ability to forecast in the current environment. I was shocked to learn recently that one of my most cutting-edge clients, one who has actually been successful in incorporating Artificial Intelligence (AI) into their robust forecasting processes, had basically disconnected their AI capabilities in the present situation. The reason was simple, the model and forecast were no longer providing useful insight and foresight to the organization. This is just another example of a quote that I often reference by the famous British statistician George Box, “All models/forecasts are wrong, some are useful.”
Wharton, U Penn’s business school, looks at a particular aspect of dealing with the current unpredictable business environment: “How Firms Can Become More Resilient in the New Normal“:
“The study offers a close-up of the “resilience” of businesses, which Deloitte classifies under three broad phases as organizations cope with the fallout of the pandemic: Respond, Recover and Thrive. “Almost a third of the businesses were negatively impacted during initial stages of the pandemic,” said Omar Aguilar, principal and global strategic cost transformation leader at Deloitte Consulting LLP. “Most companies are trying to power up or transition into the ‘thrive’ mode in a rapid way.”