The site
Who is behind the project?
The project is led by film data analyst Stephen Follows and includes data scientists and film experts. You can see more detail on the core team here.
Are you affiliated or connected with any VOD platform?
No. We are a wholly independent project, free from the control of any VOD platform. Although our data is solely from Netflix users, we see this analysis as relevant to the whole sector.
Why did you remove my comment / forum post?
We will do our best to leave all contributions available and unedited. However, we want conversations and observations to be on-topic, professional and pleasant. If your contribution was removed it was likely to have been either not relevant to the conversation or wasn’t in the spirit of respectful debate. We do not remove comments or ideas we don’t like (feel free to disagree with or object to anything), just do so with respect for everyone.
The analysis
Can I use or re-print some of the content?
Drop us a line. In theory, we’re open to the sharing of ideas. In practice, it’s a little more complicated as some of the data and analysis comes with particular terms and we are not completely free to give access. We’ll always try to allow use where possible.
Can I have more detail about [x]?
Members have access to forums, in which the content is being discussed. If you have a comment or question about a particular piece of analysis then you can discuss it in the appropriate forum. If you have an analysis request, then add it to the Analysis Requests forum.
The data
How do you know what people are watching on SVOD?
We have gained access to a unique, large and powerful dataset. This clickstream shows 610 million actions taken by opted-in, anonymised Netflix viewers around the world over a 42-month period. These clicks reveal what they watched and how they arrived at the content.
Do you have viewing figures?
Yes and no. We do have raw numbers of viewers within our panels of users, but this is tricky to scale up to provide exact viewing figures. As users were acquired via the use of services, tools and browser extensions, it would be hard to disentangle an increase in viewing figures with an increase in panel size. However, we are able to reliably track relative interest in VOD content, when compared to other such content on the same day in the same country.
What type of activity is being measured?
The dataset measures desktop and laptop users of Netflix. Netflix has stated that around 25% of global users watch via a browser, although this fluctuates between countries. This is another reason why we’re only tracking relative interest in titles, rather than raw viewing figures. We can measure the clicks people made, and from that infer views, viewing time and a whole host of other audience behaviours, such as their progress through a series, the interconnection of genres, etc.
Is the data live?
No. Due to changes in the clickstream industry during the summer of 2019, our dataset is static and historical. We have data for a 42-month period for the US and Canada, ending in June 2019. As time goes on, the insights become less useful, although the audiences and their taste in content are unlikely to change dramatically. We suggest readers apply the same discounting they do to historical box office data (i.e. lessons learned from patterns of cinema attendance between January 2016 and early August 2019).
How does your data compare with other SVOD viewing data?
Outside of our clickstream dataset, there are two other methods of tracking SVOD performance:
- Data released by the platforms
- Data collected by third parties
Each type of data has its strengths and weakness, and each uses a different methodology to measure the same underlying trends.
1. The SVOD platforms have the largest, cleanest and most comprehensive data. Not only do they have a complete log of all views, but they can also track other audience behaviours (such as how they navigate around the site), demographic data and subscriber history. However, very little of this data is ever released, and the majority of it comes edited by the PR department. This means that while we can be sure that the stated figures are correct, we cannot know exactly how they were created, what was left out and how it compares to other titles.
2. Third-party ‘Ratings’ companies apply the same technology used to track viewership for traditional broadcast and cable television channels to SVOD (typically via bespoke hardware in the homes of chosen families). This means the data is live, independent and can be combined with demographic data on the participants (allowing for weighing to accurately represent the general population). The necessity to install physical hardware means that only a limited number of participants can be included and makes it hard to take a global perspective.
Our clickstream data is a vast dataset which spans the whole world over a meaningful period of time. Our analysis is independent and created in the best interests of the wider global film/ TV communities (not just those with deep pockets). However, we cannot track any demographic information about our users, the data can be fuzzy in some places and it is for a fixed, historical period of time.
If you have access to internal platform data, definitely use that! If you can afford third party ratings data then it is certainly worth including in your calculations. For the rest of, us, there’s the clickstream.
How reliable is the data?
The data supplier is an experienced and well-respected player in the sector. When we first received the data, we performed a number of tests to see how it compared to statements made by Netflix around the performance of their content. This is harder than it sounds as Netflix rarely make such statements and do so without providing full methodology. Nonetheless, we were able to reach equivalent results using your data and methods.
It should be noted that using clickstream data will be less reliable than having access to Netflix’s internal systems. We are using a foggy lens to observe reality. We have taken this into account in our analysis and have not published anything we do not regard to be reliable.
There is more detail on the data, including some caveats and limitations here.