A common concern with market research is the belief that too many of the findings focus on the past and “growth forecasts” often are based on gut feeling, not data. It’s common for researchers to describe “where the hockey puck has been,” but as one of our clients routinely says, “I know where the puck has been, I want to know where it’s going to be.”
Forecast modeling is not a new technique, but going “beyond the numbers” is key to creating a forecast model that is accurate and truly actionable. An effective forecast model will combine traditional data-driven insights with in-depth qualitative triangulation. This serves to qualify the data forecast and create a more realistic, industry-based assessment of future trends. This is particularly helpful in B2B markets with few published long-term forecasts.
We believe there are nine key factors to build a useable, long-lasting forecast model that won’t “collect dust” on its electronic shelf:
- Keep the math simple – The more the customer can understand, the better. Simplicity leads to a model that is easier to refine and identify problems.
- Pressure testing – Run the numbers by people who know the industry. This will help to ensure realistic projections. A model that projects 4x the growth experts expect may be flawed. Explain the factors which most influence the results to see if it matches their experience. They’re “experts” for a reason and likely know the field better than you.
- Test it on historical data – If the model is completely wrong about the past, it’s probably not going to be right on the future.
- Make it malleable – Make inputs easy to tweak. Allow the customer to input new market data when it becomes available, and set up the model to update automatically if possible. This creates more value with less effort.
- Training, Training, Training – Make sure the customer knows how to interpret the results, and segment/slice results if applicable. Also teach them what not to touch, to make sure they don’t unintentionally change the results.
- Use the “Rubber Duck Technique” – This is a technique borrowed from computer programming (Rubber duck debugging) where you explain the model, in detail, to a rubber duck (or any other inanimate object nearby). In doing so, it helps bring up issues, logical jumps, and gaps in the model that might have otherwise gone unnoticed.
- Get creative – A model really can’t be any better than the data it’s based on. Where data is scarce, look for adjacent markets or markets that grow and contract with the subject but have better data available. Look for any possible mathematical relationship in historical data between the two areas.
- Have a clear, explainable path from input to output – Be able to explain the main points of the model in just a couple of sentences. If it takes longer than this to tell what data you’re using, and how it relates to the output, it’s going to be a hard sell to the customer.
- Every number has a meaning – Minimize constants and multipliers that aren’t connected to real-world data and be able to explain the constants and multipliers used. If you have to decrease the value of an input to 75% to get a tight fit on your model, find out why that input is only having 75% of the expected impact.
Whether you are interested in power tools, automotive parts, chemicals, housing starts, or almost any other product or market, growth forecasts enhanced with qualitative research can provide insights and add confidence to the “go/no go” decisions for your products or services.