As the 21st century speeds on like an unstoppable train of technological progress, an increasing number of companies are leveraging the power of digital data. As a result, business analytics teams are growing as commonplace as marketing, sales, and finance functions. Yet, while the ubiquity of data plays an increasingly vital role in access to real-time insights, the analytical techniques involved are not so new.
In this explainer, we explore the difference between two established types of data analytics: predictive vs. prescriptive analytics. While these approaches are not new, they’re finding novel applications in the digital age. We’ll cover:
- What is predictive analytics?
- What is prescriptive analytics?
- Predictive vs. prescriptive analytics: What’s the difference?
- How do companies use predictive vs. prescriptive analytics?
- Key takeaways
Ready to learn more? Let’s dive in.
1. What is predictive analytics?
In short, predictive analytics relies on past data to predict what might happen in the future. These predictions usually forecast when something might occur, along with the corresponding uncertainty of this outcome. A simple example is a company using past financial accounts to forecast how much they are likely to earn in the upcoming financial year. Based on factors like supply and demand, customer behavior, and other metrics, they can determine the likelihood of different possible profit margins.
In general, predictive analytics involves building statistical models that can identify and interpret patterns and trends within large datasets. Using complex math, they also determine the likelihood of any given outcome.
Because predictive analytics works best with big data, it often utilizes machine learning algorithms. These can predict output values based on input training data. While building machine learning algorithms is a complex task, they ultimately save time and can spot patterns that humans might miss.
Finally, while the forecasts of predictive analytics may imply certain courses of action, they don’t specifically make recommendations. Instead, predictive analytics attempts to state facts as clearly as possible, e.g. ‘Since X happened, Y is likely to occur.’ The word ‘likely’ is critical here—predictive analytics does not guarantee a future outcome, instead offering a measure of the probability that it will occur. In this respect, predictive analytics is really nothing more than informed guesswork (albeit very sophisticated guesswork!) but it is not a crystal ball.
2. What is prescriptive analytics?
As we’ve established, predictive analytics attempts to forecast what will happen in the future. Prescriptive analytics, meanwhile, takes this one step further by identifying one or more actions that an individual or organization can take in response to a given forecast. Prescriptive analytics also attempts to determine what outcomes these actions may lead to in their turn.
For instance, using predictive analytics, a company might determine that its revenue in the upcoming financial year is set to stagnate. As a result, they may use prescriptive analytics to inform future growth strategies. By modeling the potential revenue growth of different approaches (such as ‘product A versus product B’) they can decide which has the highest potential for success (in this case, the product most likely to increase profit).
In many ways, prescriptive analytics is more complex than predictive analytics (although the two are not always easy to separate). While the former focuses on the likelihood of a given outcome, prescriptive analytics carefully dissects these outcomes. Using an even broader range of measures and data, prescriptive analytics suggests actionable steps and attempts to measure the interrelated impact of these actions.
Because it uses such a wide range of data, prescriptive analytics also commonly applies machine learning techniques. It can tackle various problems, too, from risk management to business optimization. As new data becomes available, prescriptive analysts will constantly update their models to improve recommendations. This makes prescriptive analytics one of the more useful yet complex tasks for any data analytics team.
3. Predictive vs. prescriptive analytics: What’s the difference?
Now that we understand how predictive and prescriptive analytics are used, what are their main differences? You’ll find many resources trying to draw a clean line between predictive and prescriptive analytics. In reality, they’re more like segments in the same orange. Namely, they are not completely separate processes in which one is more powerful or ‘better’ than the other, but both parts of a unified whole. Nevertheless, predictive and prescriptive analytics do have some differences. And in this section, we’ll summarize these.
- Predictive analytics forecasts potential future outcomes based on past data.
- Prescriptive analytics uses a wide range of data to create specific, actionable recommendations for these predictions.
- Predictive analytics often uses structured historical data (e.g. credit histories, transactional data, customer data).
- Prescriptive analytics commonly uses hybrid data; a combination of structured data (as already described) and unstructured data (e.g. videos, pictures, and documents).
- Statistically speaking, predictive analytics involves predicting the value of an unknown variable using the values of known independent variables.
- Meanwhile, prescriptive analytics involves determining the optimum value for a decision variable (within certain constraints) to optimize one or more performance metrics.
- The same predictive analytics model will always provide identical predictions based on the same data.
- Prescriptive analytics models constantly require updating with new data to keep recommendations relevant.
Despite these differences, predictive and prescriptive analytics have more in common than that which separates them. As such, it can be difficult (and at times unhelpful) to try and categorize them too precisely.
At this point, it’s also worth mentioning that these two techniques are usually accompanied by two other forms of analytics. These are descriptive analytics (determining what has happened) and diagnostic analytics (determining why something has happened). We won’t cover these in detail here but you can read more about descriptive analytics and diagnostic analytics in this guide.
4. How do companies use predictive vs. prescriptive analytics?
Despite their differences, predictive analytics and prescriptive analytics are really most potent when they’re used to complement each other. Without examples, it’s hard to grasp how they work in practice. So in this section, we provide some real-world use cases.
Transport, travel, and logistics
Predictive and prescriptive analytics are also big players in the financial industry. They have myriad applications in areas from fraud detection to algorithmic trading. However, a fascinating application we don’t often consider is cash management. Where does your money go when you spend it? How quickly does it move around? (In this case, we mean notes and coins rather than electronic payments). Answering these questions (predictive analytics) is critical for helping banks determine when and how to put cash into circulation (prescriptive analytics).
Meanwhile, the travel industry uses prescriptive analytics to manage delays. For instance, should airlines hold an aircraft for passengers on delayed connecting flights or allow the planes to take off, forcing passengers to rebook? Both options come with costs to the airline: direct financial costs and impact on customer retention (disgruntled passengers aren’t likely to use that airline again in a hurry!) Using a combination of predictive and prescriptive analytics, airlines can improve route optimization, accounting for variables from customer loyalty and lifetime value to the cost of delaying individual flights, and so on.
Finance and cash management
Predictive and prescriptive analytics are also popular in the financial industry. In this sector alone they have myriad applications from fraud detection to algorithmic trading. However, one interesting application we don’t often consider is that of cash management. Where does your money go when you spend it? How quickly does it move? In this case, we mean notes and coins, rather than electronic payments. But answering these questions (predictive analytics) is important for helping banks determine when and how to put more cash into circulation (prescriptive analytics).
Many factors are involved in making these kinds of decisions. For example, a lot of income is generated in large cities but spent in rural areas. This means we need a system for moving cash back to economic hubs. To ensure the appropriate supply of cash to different locations, banks use sophisticated data platforms that balance factors like transport costs, cash processing, physical storage, and transaction charges.
Prescriptive analytics informs them of the consequences of various courses of action, such as depositing cash in a particular economic hub or bank vault. All this is fundamental for keeping the economy running. This is also a prime example of how predictive and prescriptive analytics go hand in hand. If a bank incorrectly forecasts where money will be spent, its cash management decisions will have costly real-world impacts. If you’re interested in reading further, Syntelli has a great article going further into prescriptive analytics in finance.
Improving the healthcare industry
Like finance, data analytics is commonly used in healthcare, with applications from disease diagnosis to setting insurance premiums. For example, within the pharmaceutical industry, companies use predictive analytics to determine which drugs are most likely to be successful. Next, prescriptive analytics can help them increase the speed of drug development. They can identify which demographics or patient groups are most suitable for a particular clinical trial, for example, based on data ranging from age to location and medical history.
Using predictive and prescriptive analytics in drug development is not just a sensible precaution for keeping people safe. It also minimizes the risk of a pharmaceutical company spending large sums of money developing the wrong drug. When used well, prescriptive analytics can speed up development and approval, reducing overall costs and increasing speed to market.
Prescriptive analytics also plays a critical role in the emerging green economy, where it’s used to optimize renewable energy plants, such as solar arrays and wind farms. While machine maintenance is not exclusive to this sector, when things go wrong with power infrastructure, it can have devastating costs beyond the repair of a mere broken component. That’s because homes and businesses facing power outages also suffer economically.
Predictive analytics can help renewable energy plants forecast when expensive machinery is expected to break down. Based on these predictions, prescriptive analytics can help firms allocate appropriate resources for maintaining machinery before it breaks. While this comes at a cost in itself (labor, replacement parts, etc.) it ultimately reduces costs in the long term.
Predictive and prescriptive strategies in the energy sector utilize metrics ranging from weather data to energy demand and real-time component data. This approach not only improves short-term efficiency but increases the overall lifespan of the technologies used, due to better care. And as renewable energy companies demonstrate the continued benefits of predictive and prescriptive analytics, the more funding they will secure. This is hugely important in a sector that is vital for helping us tackle the climate crisis.
Retail product optimization
Realize it or not, retail businesses use predictive and prescriptive analytics to optimize their product assortments. Products in this sense can be anything from laptops to clothing lines or even ice cream. In fact, let’s say a national ice cream chain wants the optimal range of flavors available to boost profit. What does that look like? Their branch in a city with a large demographic of vegetarians might demand vegan ice cream, allowing them to charge more per scoop. However, a branch elsewhere might prefer dairy, making it a costly mistake to roll out their vegan range in this location.
Predictive analytics can solve these issues, helping our ice cream chain determine which of their products is most likely to be in demand where, at what time of the year, and so on. Meanwhile, prescriptive analytics can help them decide how much to charge in different locations, the impact of associated costs of hiring and training new sales staff, what might happen if they switch chocolate chip for rum raisin, and so on. All these factors will play into optimizing product ranges to maximize profit.
In this post, we’ve explored the similarities and differences while looking at predictive vs. prescriptive analytics. We’ve learned that:
- Predictive analytics forecasts potential future outcomes based on past data.
- Prescriptive analytics involves making specific, actionable recommendations based on these forecasts.
- Predictive analytics models always produce the same outcomes when using the same data.
- Prescriptive analytics models constantly use new data to update their recommendations.
- Both predictive and prescriptive analytics are powerful business tools and are most effective when used together.
- Prescriptive and predictive analytics have applications in many sectors ranging from transport and logistics, finance, healthcare, the green sector, and retail.
To learn more about data analytics, check out this free, 5-day data analytics short course, or read the following introductory guides: