Forecasting is the third rail of business. Few companies are really good at it, and there can be big penalties for being wrong. In fact, a survey of more than 500 senior executives showed that only 1% of companies hit their financial forecast over three years, and only one out of five are within 5%. Overall, companies were off by 13%, which impacted shareholder value by 6%.
Forecasting, as analytically challenging as it is, is a lot like politics, in that there are multiple agendas. Those responsible for delivering a revenue forecast typically want to lower it. Those seeking more resources want to increase it. This manipulation makes sense: If a forecast is unlikely to be accurate, then you may as well align it with your agenda.
The clear solution is to improve the quality of forecasts, especially in further-out years. With the growth of big data, it is tempting to hope that forecasting will get better. But some forecasts focus too much on macro data, like GDP, urbanization, and population growth. Those are certainly important, but often things get lost in translation by the time they get to your category.
Other forecasts focus too much on micro data that is very specific to your category. They can become too narrow, and it becomes easy to lose sight of bigger disruptions that might occur. Some even believe the answers will emerge if we just keep crunching bigger and bigger data with better computers.
Our view is that “middle data” is the key to improving forecasts. By middle data, we mean information that’s somewhere between the big, country-level data and the category-specific microdata. Sometimes this information indicates life events and triggers at the individual level — getting married, moving to a new home, or a lifestyle change such as finding a new job. All of these things can dramatically influence your need for products and services. Middle data is closer to actual consumers than far-flung data like GDP, but it elevates the frame of reference, as most companies mistakenly believe consumers spend more time thinking about their categories and brands than they really do.
Consider a few examples:
Trends in religious growth are one of the best predictors of fashion-related categories. Certain religions strongly influence the type of clothing you wear during religious services. They can also influence your attitudes to traditional dress versus Western clothing outside of religious services. In areas where Christianity is growing, sales of Western-style formal fashions are likely to grow too, since Christianity usually carries a strong acceptance and influence of Western culture and clothing. As an example, an area of the world that could be strongly impacted is sub-Saharan Africa. The share of the world’s Christians living in sub-Saharan Africa is expected to grow from 24% in 2010 to 38% by 2050 (according to Pew Research). Growth of religion is a stable, steady, and sustainable trend, which is the ideal type of data for forecasting.
Your climate and the type of house you live in are great predictors of entertainment (e.g., toys) and education (books) spending per capita on babies and toddlers. Contrast the available living space of a large suburban home with a full-size basement versus a small condo in a large city. Compound this with your local climate, which dictates how much time you spend in or outside your home in harsher weather. You will quickly see how one family in a larger house and harsher climate has much more capacity and demand to buy entertainment products and educational toys and books for their growing children.
The fundamental driver of demand isn’t the choice to live in a certain house and climate. The parents’ demand for the type of home and climate to raise their children in drove the choice of the house and climate, which in turn drove demand for entertainment and education spend. But the choice of house and climate is easy to measure, whereas a set of big data on parental motivations does not exist.
The UN human development index (an amalgam metric of education, economy, and infrastructure) is another great predictor of consumer behavior, specifically as it relates to consumer spending on beverages. This index can help explain whether consumer demand for beverages is more “drink to live,” buying basic functional beverages such as dairy products and big bottles of water to cook, wash, and drink, or “live to drink,” in which people buy more enjoyment-related beverages such as wine, coffee, and other premium or functional health-related beverages.
China was a great example of this as it shifted from “medium” to “high” on the human development index during the 2000s. Specifically, from 2000 until 2012 its development index rose from 59 to 72, putting it into the “high” development standard. During that same period its grape wine consumption increased more than fivefold, from just over 0.25 liters per capita to 1.5 liters per capita, according to a study by Australian National University. And again, this is publicly available data that is also slow and steady in development, making it great for forecasting.
In the end, like many analytics exercises, forecasting is a “garbage in, garbage out” process — you get only what you put in. Using big data to incorporate “middle data,” — ideally from slower, steady, and sustainable data sources, and often using data adjacent to your business and category, can be a great way to improve forecast accuracy, growth, and shareholder value.
Jeremy Bartlow is a project manager with The Cambridge Group.
Tim Joyce is a principal with The Cambridge Group.
IMAGE CREDITS: http://qvs.nl/