Predictive analytics makes forecasts about the future based on the data a company has. We recently published an article about demand forecasting which is a specific type of predictive analytics relating to understanding service demand. But there is a lot more that companies can achieve through predictive analytics than simply predicting demand for products and services in the coming months. In this blog post, we’re going to walk through a few other use cases for predictive analytics.
Our data science consulting team have lots of experience at building predictive analytic models that don't require other third-party products or algorithms to function smoothly and effortlessly for our customers, using either AWS or open-source technology. We mostly work with companies who already have their own in-house data science team and, in our experience, it is this combination of an in-house team with an external team like Dativa that provides the best solutions in terms of both cost-efficiency and time-to-market.
Customer engagement involves getting to know as much as you can about your customers while ensuring that the data you gather about them has a clear and documentable business value. As you get to know your customers' desires and preferences, it becomes easier to acquire new ones, retain the ones you already have (rather than losing them as they switch to your rivals), upsell them your premium services and target them for those particular ads to which they will best respond.
The real goal is to be able to predict the future behavior of the customers. Perhaps your executives have to choose between commissioning a new season for one of two programs while also deciding which two of three new shows to broadcast. The job of the data science team is to present these teams with reports which, by making predictions about future customer behavior, give a helping hand to the executives who will make the actual decisions.
The place to start is by analyzing past consumer behavior while remembering that consumers and their tastes evolve with time, and trying to understand what these changes might be. Someone who loves sport is unlikely to switch to suddenly wanting to watch comedies when their viewing history shows no interest in humor, but they might well be interested in a new car show the company has lined up, or Premier League soccer. Meanwhile, the person who does love anything humorous is worth upselling recently released box office comedies on TVoD. The trick is to find the content that consumers will respond well to, engaging with their viewing desires and preferences, and getting the right, personalized offer to them at the right time.
It sounds like a buzzword (and, well, it is), but predictive analytics in TV advertising can mean lots of things. Some of it is about where to find an audience - what shows are an audience or segment most likely to watch, and where are they likeliest to over-index? We can gain some of this insight by merely analyzing the data, but predictive analytics can unearth patterns that humans may not be able to find.
One of the unintended consequences of GDPR is also about using predictive analytics to audience segments. You may have a small pool of opted-in users who have opted to share their data with you. TV companies are starting to use predictive analytics to understand which other customers in their database that have not chosen to share data share similar properties with the customers that have. We think this kind of probabilistic matching and segment creation will become more prevalent when a more deterministic solution to matching is not available.
There is nothing inherently new in the concept of predictive advertising, albeit the ways to predict how consumers will respond to ads has changed, and is based on analyzing what the data says (and having a wider range of data points to which to match the TV data). So it is worth targeting an ad for a new expensive Ford model at fans of Top Gear America more than at individuals who have never watched any car-based shows. It isn't that the latter group are unlikely to buy a car, but they are unlikely to be as brand sensitive and are less likely to buy an expensive model than the fans of car-themed shows. The team should include any data that can help gain insight into how customers are likely to respond to commercials, as this has clear business value.
Predictive analytics can also help to drive operational efficiency. Sometimes this is very pragmatic. Can we use predictive analytics to understand what bandwidth costs are going to look like in the next five years, and how to provision networks to take advantage of that? Sometimes, a company knows that there are areas that it can improve on operationally, and analytics can most help by identifying what the cost savings are, and how to implement them. The most compelling opportunities are areas that the company don’t know about. Here, a data mining exercise on some of a business’s operational processes can reveal patterns indicating areas of improvement that the company did not even know about.