In the first installment of this series on AI-based predictive analytics, we explored the functionality of this technology, along with its potential to create more effective business strategies.
Predictive analytics can be defined as a form of data mining that uses statistical modeling to analyze historical patterns, and then uses these models to project future outcomes. The deployment of artificial intelligence allows analytics technologies to spot relationships between variables that humans are simply incapable of seeing.
In this article, we want to bring that theory to life with five predictive analytics use cases.
There have been some newsworthy stories in this field, notably the “Target Knows When You’re Pregnant” headlines that garnered so much attention a few years ago.
Things have developed quite a bit since then. The evolution of widely available and accessible analytics platforms has provided access to sophisticated statistical models for companies of all sizes. Beyond the hyperbole of slightly creepy pregnancy predictions, big data is more typically used by small and large businesses to improve their everyday functions.
By defining the problems they want to solve, procuring the right data, hiring people with the skills to make sense of the data, and empowering them with the appropriate technology, any business can get started with the profitable field of predictive analytics today.
There are too many contenders to consider every example within the scope of one article, so we have instead tried to provide a representative sample of effective AI-based analytics across a broad spectrum of case studies.
1. Supply chain optimization: Walmart
We begin with a heavyweight example, but one that provides inspiration for all retailers.
While so many other ‘legacy’ retailers are struggling, Walmart has posted growth figures for the last 11 consecutive quarters. Notably, this has been driven by a 63% year-on-year increase in online sales.
Walmart has received much acclaim for its willingness to adapt in the digital age and is betting on its ability to link the online and offline worlds to compete with Amazon.
Artificial intelligence and predictive analytics are at the heart of this drive. Walmart takes data instantaneously from its point-of-sale systems and incorporates this within its forecasts to assess which products are likely to sell out and which have underperformed.
Combined with online behavior patterns, this provides a huge amount of data points (over 40 petabytes of them) to help Walmart prepare for a rise or fall in product demand.
Photo by chuttersnap on Unsplash
Data is managed in the cloud via Walmart’s “Data Cafe”, which is maintained by the Walmart Labs team in Silicon Valley. It is a sophisticated, large-scale operation in keeping with the number of variables required for a business of this size to make accurate projections from reliable data.
Nonetheless, the benefits it brings can also be sought by smaller businesses too.
For example, Walmart’s use of AI and predictive analytics is invaluable for inventory management, as managers can stock appropriately without running the risk of having to make expensive last-minute adjustments to plug gaps when demand outstrips supply.
These forecasts also allow Walmart to personalize its online presence, showcasing products to specific customers based on their predicted likelihood of making a purchase.
The discipline and rigor that this approach brings mean that Walmart can stick to strict delivery dates, as each step of its supply chain has been optimized through the use of predictive analytics. All of these areas can be improved by any business through accessible technology from the likes of Google and Adobe.
Tellingly, Walmart also offers incentives to customers in the form of price reductions or queue-jump privileges if they collect their purchases from a physical store. Even with all the benefits AI-based analytics can bring to the business, competing with Amazon on shipping costs remains a tall task.
2. Forecasting price trends: Hopper
The travel industry is notoriously competitive, with volatile peaks and troughs in demand and many low-margin routes. This can leave travelers in the dark, unsure of the best time to book. Sometimes it’s better to book ahead, at other times it’s better to wait until closer to the date of departure.
This makes it a field ripe for the power of AI-driven predictive analytics, a fact that has seen the travel app Hopper grow dramatically in popularity since 2015.
Hopper stays one step ahead by predicting future pricing patterns and alerting travelers of the cheapest times to buy flights to their preferred destinations.
It does this by watching billions of prices every day and, based on historical data for each route, anticipating how the trend will develop. Users can then set up notifications to remind them to book when these price drops come to pass.
Although not the only such company to provide this service, Hopper reports a 95% accuracy rate with its predictions and claims to save customers an average of over $50 per flight.
The screenshot below shows how this process functions. Accompanied by a cuddly, bespectacled bunny, I select the New York to Honolulu flight route for that richly-deserved vacation.
Based on my selected dates, the surprisingly bossy bunny tells me to book now, as tickets for this route will only get more expensive over time.
Hopper provides a great example of a business that takes machine learning and predictive analytics as the central tenets of their business strategy. Without predictive analytics, there would be no Hopper.
The statistical models it uses to such great effect hold lessons for all business, however. Hopper’s success comes from its reliability as an objective consumer advice platform, essentially. As such, many other companies could assume this role by using statistics to provide forecasts that are in the customer’s best interests, rather than just their own bottom line.
3. Small business growth: Point Defiance Zoo & Aquarium
A survey by SAP in late 2016 found that over 70% of small business leaders felt that they were still only in the “early stages” of deriving insights from their data.
One zoo in Tacoma, Washington bucked that trend by partnering with the National Weather Service to identify the factors that caused attendance figures to rise and fall so unpredictably. This created issues for management, who would always staff the park to cater for a large audience, but often ended up overspending on salaries due to underwhelming attendance.
Intuitively, we could assume that attendance is higher on warm, dry days, but lower when it is cold or wet. However, by incorporating the National Weather Service’s data into IBM’s AI-driven Watson platform, the zoo was able to pinpoint exactly which conditions caused more people to make a visit.
This knowledge was then used to model future visitor patterns, using historical attendance figures and projected weather statistics.
The project was a huge success and is now a central part of the zoo’s business planning. Point Defiance can predict attendance figures with greater than 95% accuracy, allowing managers to staff the park appropriately. This has no negative impact on how visitors experience the park (perhaps even the opposite), and creates some vital business efficiencies.
The applications of this methodology reach far wider than just attendance figures, of course. Port Defiance can monitor how visitors interact with the zoo, helping to provide a better customer experience. Plans are also in place to use AI-driven predictive analytics to monitor health data and diagnose issues with the park’s animals to provide pre-emptive treatment.
4. Staff retention: IBM
The fundamental attraction of predictive analytics is the potential to deliver better outcomes against organizational goals. These are often overtly profit-based, but predictive analytics can also help identify staff retention issues and suggest solutions.
By uploading a structured data file (as in the screenshot below), Watson can spot the common contributing factors in staff attrition. This then feeds into the generation of a ‘quality score’ for each employee, based on their projected likelihood of leaving the company soon.
Where this really comes into its own is in its ability to respond to natural language requests from users. In a similar fashion to Google’s new Analytics feature, which will fetch data in response to user questions, Watson can respond to specific queries and build data visualizations based on the user’s preferences.
This is a great example of a platform that moves quickly from exploratory and diagnostic analysis, into the realm of predictive analytics. Any business owner or manager can make use of these tools to identify with precision what exactly causes staff to leave, but they can also see what lies behind those factors and put in place preventative measures to appease any potential departures. Given the cost of recruiting new staff versus retaining current high-performers, this leads directly to decreased operational costs.
5. Audience extension: Under Armour
Audience extension is another area of marketing that benefits significantly from the use of AI and predictive analytics. By understanding the quantitative characteristics of existing high-value customers, it is possible to identify similar individuals and target them with personalized messaging that is likely to resonate.
Knowing where to spend your advertising budget is essential, but so is knowing where not to spend it. Predictive analytics allows companies like Under Armour to hone in on the areas that will deliver the greatest returns, and reinvest budget that would otherwise have been spent inaccurately.
Artificial intelligence is used by Under Armour to perform tasks such as sentiment analysis and social listening to understand what customers think of the brand, and where the gaps in the market are. This has led the company to focus on becoming a digital fitness brand, an initiative that has seen it carve it a new niche in a saturated market.
Under Armour produces physical fitness products, but also apps and wearable devices to tie the offline and digital worlds together. The more people use the products, the more data Under Armour can gather to improve its offering. And with over 200 million registered users and more than 10 billion digital interactions per annum, there is no shortage of data.
Read on to the final installment in this series: AI and predictive analytics: What does the future hold?