How Samba TV measures viewership insights

How does Samba TV get its data?


We gather viewership data from consumers who have opted in to data collection and enabled our software; our software is embedded in Smart TV’s sold by 24 leading brands (the most in the industry). We also license data collected by set-top box providers. Combined, we analyze data from about 46M TV devices globally (around 28M TV devices in the U.S.).  

While we do not estimate viewership on non-TV devices, we do measure content that crosses the TV screen regardless of the device used to play it, whether it be an external streaming device, gaming system, or computer. 

Our data also leverage household-level TV viewership, which means we don’t estimate co-viewing (the number of people watching within each household). 

Our content identification technology identifies what’s on your TV, so if you’re watching on a handheld device, we won’t be able to measure that unless you cast it to your TV screen. For example, if you’re watching HBO Max via mobile and AirPlay it to your Samba-enabled Smart TV, then our technology would capture that. 

Our data set allows us to make statistically significant measurements, and because we're gathering first-party opt-in data from the TV itself and not one content provider in particular, our data sets are agnostic and representative. 

**Updated August 4, 2021

How does content identification work?


Samba’s proprietary content identification technology uses Automatic Content Recognition, a process that involves video fingerprinting and sophisticated machine learning algorithms to to recognize content that is displayed on the TV.

How do you account for an entire country when not everyone has a Samba-enabled TV?


When we want to understand how a population behaves, but we can’t survey everyone in the country, we use a panel that is representative of the general population. This allows us to accurately report on viewership behavior, even when not everyone has a Samba-enabled TV.

We analyze data daily from millions of televisions and set-top boxes. This gives us the size and diversity requirements of a good panel.

Our U.S. panel, for example, is composed of millions of opted-in Samba TV households filtered, weighted and projected to annual U.S. Census data by geography, gender, ethnicity, age, and income to be representative of the whole country – including those without Samba-enabled TVs.

We similarly scale panels internationally, ensuring that each country’s panel is normalized and curated to be accurately representative of the population. Normalizing our panel data to a country’s population makes our panels representative and allows us to measure with statistical significance and accuracy while also remaining 100% privacy focused.

If I have a Samba TV, how do I participate?


It’s easy to opt in (or out) of Samba TV via the set-up screen on your TV. You can always change your mind and opt out (or in) by going to the “settings” menu of your TV. We are committed to protecting your data and giving you complete control over its use. Read more in our Privacy Center

Samba TV's goal is to make the TV viewing experience better for everyone. By opting in, users are also able to take advantage of a number of exclusive features. This includes access to our groundbreaking Picture Perfect℠, which uses Samba technology to enhance your picture by recognizing and optimizing the image quality of the content playing on the screen in real time, as well as TV show recommendations based on your preferences, and more.

What’s the difference between your data and other data providers?


Our data sets are neutral and platform agnostic. Because we're gathering opt-in data from the TV itself and not from one content provider in particular, our first-party data sets are neutral and comprehensive.  

Our data insights are timely. While others have to wait for viewership insights, ours are available quickly, so that neutral insights are available when you want them. 

Our data sets are anonymized. Privacy is central to all we do. All of our data sets are anonymized, and we do not have contact with any opted-in users. We measure at a household level, not individual users, and household information is measured without directly identifiable information (such as names or email addresses) for maximum privacy.

How do you account for subtle changes in your data?


To ensure our data and panels are comprehensive and representative, Samba TV consistently updates our data sets. Because people move, get a new TV, or opt-in or -out of Samba TV, when we pull data at different times, numbers may change slightly to accommodate these changes in our panel.

What does it mean when a city over- or under-indexes?


When we share data insights in the U.S., we often look at DMAs, or designated market areas. There are 210 DMAs that make up the country, many based around big cities. These are sometimes referred to as “metro areas” as well. While we are thoughtful of viewership throughout the country, when we want a high-level look at the areas driving the highest rate of viewership compared to the country overall, we typically focus on the 25 largest of the DMAs. 

Fun fact: the 25 largest DMAs represent almost exactly half (49.7%) of the U.S. population. Outside of the U.S., we similarly break out viewership by region. For example, in the U.K., we look at BBC regions, in Germany we look at German States, and in Australia we look at AU TAM regions. 

When we say a specific location “over-indexes,” it simply means households in that area watched a particular program at a higher rate when compared to the country overall. And “under-index” would be the opposite – watching at a lower rate when compared to the country overall.

What different windows of time do you look at?


For linear TV, we’ll often look at Live +0 Day, also referred to as “Live +Same Day,” which includes the households watching during a program’s initial airing until 3 a.m. local time the following day.  

Now, with viewers having so much more flexibility deciding when to watch, we often look at longer stretches of time. This accounts for time-shifting, people who may watch later, or on demand. We also look at longer time periods to get a wider view on how viewership stacks up over a weekend or within a week or even longer. This can be valuable, especially when considering streaming programming that people watch on their own time at varying paces.