Customer Lifetime Value (CLV) is a data analytics and marketing concept that estimates the monetary value of what an average single customer is going to be worth to a company, during the entire period that they are customers of that company.
CLV is useful because it gives a company insights into its long-term performance. Short-term measurements which companies typically use to measure their success on a quarterly basis (profitability, share price, revenue, sales) can be affected by seasonal variations. For example – a spike in the number of visitors to your OTT app last month might get your TV analytics team excited - but if that spike is driven by the latest popular drama also starting last month, the company cannot judge how well they are doing based on one show.
CLV can be a tremendous long-term planning tool, that helps to reveal light on the long-term performance of the business. It can also help to understand overall performance that cuts across multiple business units, revenue lines, and so on.
There are some obvious structural reasons within the TV industry as to why CLV is essential. Ad-created content is free and increasingly shareable; most subscription services have generous free trial periods and opt-in/opt-out arrangements that encourage a fluid sign-up and drop-out period. Just because someone is watching your service now, it doesn’t automatically mean they will continue to do so; there is no lock-in as with more traditional TV services. This fluidity of customer movement between services means providers must work extra hard not just to keep those customers, but to understand how valuable those customers are
Focusing TV analytics on the customer
By focusing on the customer, and measuring their behavior over the time they are likely to spend with the business, it becomes easier to avoid vanity metrics and short-term trends while instead focusing on the long-term health and profitability of the company.
There are broadly two ways of approaching the challenge of calculating a CLV:
Static/exploratory approach: This is an excellent first step towards creating a customer lifetime value model. The static approach takes specific static data inputs and calculates a historic score by making certain basic assumptions: based on how much revenue a customer has generated, how long they have stayed a subscriber to the company’s services, pricing trends, and so on.
Predictive approach: This takes CLV a step further by using machine learning and predictive analytics to create a more detailed approach which doesn't just create a historical value but also tries to predict future values, by understanding how variables will change over time. Including these forecasted values in reporting and modeling can reveal what is coming around the corner.
Our data science consulting team have identified four factors which can make a real difference when calculating a CLV score:
Getting the right data and the right metrics, which includes avoiding vanity metrics and broad metrics, which can be deceptive and produce a CLV model which could hinder rather than help the business. Data science teams need to work at understanding what accurately define success and failure. It probably isn’t, for example, the number of people who download your app, which is a classic vanity metric that doesn’t really tell a business much except how successful its launch strategy was. It’s more likely to be something like users it acquires each month, cost of acquisition, loyalty scores, and so on.
Maximize relevant data feeds: We advise our clients to maximize the number of relevant data feeds in when developing a CLV model. This doesn't contradict our basic rule that data storage should be kept to a minimum but it does mean that the data science team needs to justify every input, as undoubtedly more pertinent data will help create a more accurate CLV model. As long as the data is relevant and clean, we have found that more tends to be better.
Keep it fluid: A CLV model should not be a one-time deployment, but an evolving project. Metrics can change swiftly. Assumptions can be both proved and disproved. It is imperative to revisit your model and check that the inputs and assumptions you made are still relevant. How often a team should make these checks varies from business to business as it is a matter of balancing the cost of re-assessing the CLV compared to the financial gain from an updated CLV score. Checking the CLV every day is unlikely to be worth doing but leaving it a year is likely too long. We recommend re-doing the CLV every three months as a starting point.
Picking the right formula: There are lots of ways to calculate CLV, with no real “right” answer. Most formulae will use a combination of subscription revenue, churn rate and rate of new customer acquisition. Including profit margin and cost of acquisition provides a more rigorous picture, as does applying a discount rate to future revenues. Other CLV metrics might include the rate at which people buy content; seasonal factors; content costs, and so on. We usually recommend a hybrid approach to CLV modeling, using a composite of several methods to come up with the most rigorous final value.
Talk to our data science consulting team to talk about how we have helped media companies define their CLV models.