The Worldwide Leader in… Forecasting?
How ESPN Leveraged Data to Predict its Demise, and Begin its Reinvention
ESPN (and its parent company, Disney) faced an existential threat beginning in 2011. Cable subscriptions were on the decline, and this presented a huge threat to the ~80% of ESPN’s revenue that came from TV distributors such as Comcast and DIRECTV (the balance coming from advertising)[i]. How much worse would the situation get? Should ESPN, and Disney as a whole, take drastic action, or was this merely a minor blemish on an otherwise stellar business run? Enter data science.
From Time Series… to Advanced Regressions
Disney, similar to other major entertainment companies (including Viacom, who spoke at the Digital Summit at HBS) has an in-house Data Science team. The team is made up of statisticians, mostly at the Master and PhD level. It is well-respected, especially within the Parks division of Disney where they helped the hotels create an algorithm for dynamic pricing for hotel rooms.
In 2011, ESPN began a partnership with this Data Science team to help forecast how many households in the United States would subscribe to pay-TV (through cable, satellite, or telecom companies) in the next 10 years. In the early days, the data scientists mostly looked at historical data – how much had subscriptions declined over the last 12 months, and how much were they like to decline in the next 12 months? Given the changing patterns in the industry, the lines usually came out looking slightly down for the next 12 months, and because there was no history of stabilization given this was the first drop in subscriber numbers, for the rest of the forecast as well. These predictions were not super helpful – the corporate teams refused to believe that subscribers would decline slowly and steadily into the future, and instead believed that perhaps declines would stabilize in roughly two years when economic models suggested full employment would be reached. These early forecasts lacked corporate buy in given their limited rigor and scope.
Over the next few years, the data scientists looked not only at basic data (such as the last 12 months of subscriber data), but also transformed and missing data.
Transformed data included the net of median household income vs. the average house payment, and income vs. the average cable bill. This data showed that income was barely increasing, while rents and the average cable bill were increasing mid to high single digits each year[ii]. These data sets suggested a more structural threat, given squeezed household budgets.
Missing data sets were the final piece of the puzzle, and came from an annual customer survey ESPN and Disney conducted of ~5,000 households. The results suggested that customers were (1) continuously less satisfied with the television experience given the high cost and outdated interface, (2) continuously more satisfied with the experience of alternatives of Netflix, Hulu, and others, and (3) increasingly at risk of not paying for television in the coming 12 months.
The data scientists incorporated this transformed and missing data into the forecasts, and came up with a far more robust and believable forecast which clearly showed significant threats to the business if Disney and ESPN did not take action.
“That’s Neat, but Can You Sell It?”
As noted in the Hoffenheim and Flashion cases, internal uptake is much harder to obtain than a robust model. This was no different at ESPN, where revenues and subscribers had been moving up and to the right for 30 straight years. Armed with data, ESPN’s strategy team attempted to convince executives of the rigor of the model using complex analyses and charts which displayed trends down to the second derivative level. This approach failed miserably – eyes glazed over quickly, and reasonably so. In subsequent meetings, the team pivoted, instead positioning data as a risk mitigator. Successful phrases included “it may not be as grim as the data suggests, but if it’s even halfway there, we will have to make significant changes”, and “the whole industry is talking up leveraging big data – this is one of those opportunities!”
Executives warmed to the idea quickly, and ultimately supported new digital TV products such as Sling TV and Hulu to try to pick up some of the subscribers that were leaving the cable bundle. Disney even adjusted its guidance to the Street, triggering an (overblown) selloff that reset media company multiples in a more realistic world.
By combining existing, transformed, and missing data, Disney and ESPN were able to paint a more accurate future of the TV world, which has indeed come true as legacy subscriptions continue to decline, offset by increases in digital alternatives.
[i] SNL Kagan
[ii] Bureau of Labor Statistics, FCC
Great post Ross! As a sports obsessed kid growing up in a cable subscribing household, I would have found it difficult to imagine a scenario in which ESPN was at risk of falling from its perch as the primary source of sports content. It’s impressive that they had the foresight to see this coming and took action. Your final section about communicating the data to company leaders really resonated with me. The employees who are doing this sort of analysis are always going to use data to convince themselves of what’s going on, but mounds of data alone will do little to sway the company’s decision-makers. The data must instead be translated into a relatively simple and compelling story. However, it may be tough to expect that the same people who are great at doing the data analysis will also be proficient at crafting the stories that will convince the leaders to buy-in. My hope is that our generation of business leaders will at least be data savvy enough to meet the data scientists halfway rather than entirely relying on them to cater to our needs for simplicity.