What is Predictive Analytics?
Predictive analytics uses historical data to find patterns and then use those patterns to forecast future events. Machine learning algorithms, a fancy alchemy of statistical math and computer code, reviews data from past conditions such as marketing spend, competitor activity, pricing, and even the overall economy. And the algorithm finds the relationship to the KPIs you care about, like units sold or gross revenue, with all the precision of Sherlock Holmes piecing together clues.
While a person can look at charts to see these relationships one by one (e.g., a graph of marketing spend vs. units sold), the machine comprehends all of it at once, like some all-seeing oracle communing with the powers beyond. But this isn’t tarot cards or tea leaves or Wall Street’s technical analysis; this is based on solid and proven mathematics.
In the 1970s some companies adopted just-in-time manufacturing and some didn’t. In the 80s some companies adopted Total Quality Management and some didn’t. In the 90s and 2000s some companies embraced the Internet and e-commerce and some were laggards. Predictive analytics is a tool for strategic advantage; neglect its potential and watch your competitors thrive.
Electric Vehicle Launch
Let’s imagine you live in a world where everyone gets around town in vehicles powered by burning liquid made from dead dinosaurs extracted from deep under the earth. Sounds crazy, right? So you want to sell battery-powered vehicles that can be charged by plugging them into the electrical grid just like all our other technology, from toasters to TikTok.
Your competitors have dinosaur goo distribution stations all over the place, and you know that the range of an electric vehicle is a big concern for consumers. You need to quantify just how much more consumers are willing to pay for extra range. Predictive analytics can work out that relationship with the precision of an artisan crafting a delicate masterpiece. Then it’s a straightforward spreadsheet exercise to look at how “willingness to pay” changes with range and how cost to manufacture also changes with range. Somewhere there’s a sweet spot.
A lot of predictive analytics comes down to finding sweet spots! And you may have heard about an EV company that’s been fairly successful over the last few years.
Predictive analytics is the engine underneath almost all forecasting. And accurate forecasting is not only a way to maximize revenues, but perhaps more importantly is a way to mitigate risk. Do you have any idea how many TV shows, already filmed and ready to air, because forecasts based on the completed show indicate that the audience is going to be too small? This coming Tuesday night at 8:00 PM only happens once; no network wants to waste it on a product that is expected to fail.
Years back I worked on a product launch where PR had scored a big win; the product was going to be included, along with a handful of others, on a talk show hosted by a major celebrity. Unbeknownst to me or most of the team, the forecasting had anticipated only the “average” response to being featured on such a talk show. The team could have modeled the demand spike of a best case scenario in order to have some slack in the production system in case production needed to quickly ramp up.
But they didn’t. The celeb, on her own accord, commented that this product was her favorite among the group. It took only minutes for online orders to surge beyond the company’s ability to produce the product in a reasonable time. Huge opportunity… vanished like a whisper in the wind.
An accurate forecast of demand, or even of potential scenarios, supports not only production levels and backup capacity, but also pricing strategy, marketing budget allocation, call center staffing, and perhaps a dozen or so other important factors for a business. Predictive modeling helps anticipate what’s going to happen as well as how those future predictions react to changes in your plan. A business with a good predictive model can look to maximize returns, or minimize costs, or limit risk, or optimize to whatever metric suits its needs.
Predictive modeling can also be used for product design. A subset of predictive models is preference modeling (often referred to by the term “conjoint analysis” because conjoint modeling is the most common type of preference model). Preference models are often built by proposing different combinations of services, features, and pricing to consumers via a survey and getting feedback about demand and willingness to pay.
The model can then look at each of the features and assign a relative value with the precision of a master tailor measuring fabrics. Remember when Apple took away the headphone jack from iPhones, forcing us all to move quickly to wireless? The headphone jack was the limiting factor in making the phone thinner, among other technical requirements. While Apple is known for its secrecy, a Harvard Business School Online blog post speculated (correctly IMHO) that a preference model likely played a role in weighing the potential benefits of a thinner phone without a headphone jack vs. a thicker phone with the jack.
The Brookings Institution used preference modeling to look at decades of sales in the automobile industry, seeking to explain the large loss of market share that US firms ceded to Japanese and European manufacturers. Their model showed overwhelmingly that the problem came down to the basic attributes of the cars, and not as many media pundits had speculated to issues around brand loyalty, distribution, styling, or other “soft” factors.
And because price was an input into the model, Brooking simulated what price would have been needed in 2000 for US auto manufacturers to regain their 1990 market share without improvements in the functional attributes of their vehicles; the model indicated that a 50% price cut would be needed! Predictive modeling clearly showed where companies like GM and Ford needed to focus to restore their global competitiveness.
Minimizing Risks and Maximizing ROI
Predictive models are weather forecasts. In meteorology, input data like wind, humidity, and such goes into the machine, someone turns a crank (is that how computers work?), and out comes a percentage chance of rain. Substitute business metrics like marketing spend for weather data, and sales for rainfall, and it’s the same thing. Just as knowing the chance of rain can guide your decisions when you are headed out for the day, predictive modeling can anticipate your business needs the foresight of a seasoned navigator mapping the uncharted waters.
And predictive modeling can help prioritize different scenarios that might arise… their likelihood, their results, and how those scenarios play out given different decisions by the leadership team.
- Does the model predict likely good results, but with high downside risk? What factors can you adjust to minimize those risks?
- Is the model forecasting likely good results, but with some possibility of much higher upside? How can you maximize that possibility? How can you mitigate potential opportunity costs of unmet demand?
- Uh oh, the model says the outlook isn’t so good? Do you need to surrender to the omens of doom, or are there variables within your control that can move the KPI needle where you need it to be?
Predictive modeling isn’t just about getting a forecast. This is a dynamic tool that supports data-driven decisions with the acumen of a master chess player, foreseeing the consequences of each move.
Tools to Consider for Predictive Analytics
If you’re looking to incorporate more predictive analytics at your business, there are many tools that can help. You may already have business intelligence tools like Tableau, or advanced statistical software like SAS or SPSS, or versatile spreadsheet tools like Microsoft Excel. But if not, predictive modeling, despite its high value, doesn’t have to be expensive.
The best things in life are free, and so are the best predictive modeling tools! The Python and R programming languages are open source and easy to download from the Internet. Tools with easy-to-use interfaces also have free versions: for example, RapidMiner, KNIME, and Orange. All of these tools are potential game changers that can be acquired without expensive software licensing and can usually run on your laptop without the need for a supercomputer or specialized hardware.
Now that you know, you’ve got no excuse not to use these cutting-edge tools. The time is nigh!
Shape the Future
Predictive analytics is not just a tool but a strategic asset. By understanding its capabilities and applications, you can take the proactive step of incorporating it into your next product launch, thereby gaining a competitive edge. It’s not about predicting the future; it’s about making informed decisions that shape that future.
As you reflect on your business strategy, can you envision a scenario where predictive analytics could have transformed a past decision into a greater success? Share your insights in the comments below and let’s start a discussion on the potential of data-driven foresight. After all, hindsight is 20/20, but foresight is priceless.