There are three main ways that we measure what is getting attention on Facebook:
A) Interactions with public pages & groups
B) Interactions with links
C) Views/Impressions
There is a huge debate now about problems with these, but IMHO these problems are nuanced & *solvable*. /1
A) Interactions with public pages & groups
B) Interactions with links
C) Views/Impressions
There is a huge debate now about problems with these, but IMHO these problems are nuanced & *solvable*. /1
(A) "Interactions with public pages & groups" doesn't include personal accounts or private groups — but it does include non-link posts such as image memes and Facebook videos.
(B) "Interactions with links" is the only public way to measure engagement with news articles.
(B) "Interactions with links" is the only public way to measure engagement with news articles.
"(C) Views/Impressions" is about what people actually see.
(A) and (B) only includes likes, shares, and comments — so if you see something in your feed, or even click on it, that isn't counted.
Only Facebook sees (C), though it has shared (old) view data with a few researchers.
(A) and (B) only includes likes, shares, and comments — so if you see something in your feed, or even click on it, that isn't counted.
Only Facebook sees (C), though it has shared (old) view data with a few researchers.
Folks like @SolomonMg, @alexstamos, and @johnwhegeman (Head of News Feed at FB) have complained that the @nytimes's @kevinroose is harming discourse.
Kevin uses (A) "Public group/post interactions" instead of (B) "Link interactions" or ideally (C) "Views/Impressions".
Kevin uses (A) "Public group/post interactions" instead of (B) "Link interactions" or ideally (C) "Views/Impressions".
I have real respect for all of these people, and I think they all right. And all wrong. And perhaps missing some solutions.
Kevin, says "I don't have access to (B) or (C)".
Well, it turns out you can can (B) today.
@NewsWhip provides it.
(Other orgs like @ZignalLabs may also?)
Kevin, says "I don't have access to (B) or (C)".
Well, it turns out you can can (B) today.
@NewsWhip provides it.
(Other orgs like @ZignalLabs may also?)
Here's an example of what you could see with @NewsWhip from my research during the 2016 election. (Note the blatant misinformation in the top 5 stories being interacted with in 12 hour period.)
The key point—this seems to be *all* interactions, including private accounts. /6
The key point—this seems to be *all* interactions, including private accounts. /6
I also used the API to track top sources each day and do analysis on interactions with reliable sources over time.
Creating charts like this. (Don't take the numbers literally, lots of room for interpretation there — but the trends are accurate and significant.) /7
Creating charts like this. (Don't take the numbers literally, lots of room for interpretation there — but the trends are accurate and significant.) /7
I also wrote pieces like this: https://medium.com/tow-center/even-in-september-of-2015-hyperpartisan-news-sites-had-massive-engagement-on-facebook-c07ab3016bd1, showing that journalists, academics, and other civil society actors can use tools like @NewsWhip to identify when something fishy is going on in our platform ecosystem.
I appreciate @kevinroose actually trying to do that!
I appreciate @kevinroose actually trying to do that!
So perhaps we have a solution for Kevin; he just needs to use the NewsWhip service or API. It seems to do exactly what @SolomonMg wants:
https://twitter.com/SolomonMg/status/1300098667291049984
It gets Kevin to (B) "Interactions with links". /9
https://twitter.com/SolomonMg/status/1300098667291049984
It gets Kevin to (B) "Interactions with links". /9
That said — there is perhaps something real and important that one is measuring with CrowdTangle. Knowing what is happening on Facebook that doesn't just relate to links is important.
If a page is getting a lot of engagement through sensationalist images memes — that matters too.
If a page is getting a lot of engagement through sensationalist images memes — that matters too.
But leaving that aside, what about @johnwhegeman's (Head of News Feed at FB) critique that you must use reach/impressions/views? https://twitter.com/johnwhegeman/status/1285358835465510912
Well, I agree that this would be ideal.
I think it's bad that we don't have that data. We've been asking for it forever. /11
Well, I agree that this would be ideal.
I think it's bad that we don't have that data. We've been asking for it forever. /11
I even wrote a whole piece about how it's a critical first step to fixing our information ecosystem. Back in Jan 2017: https://medium.com/@aviv/how-to-fix-a-news-ecosystem-4bf623f91d81
But frankly, as far as I can tell, Facebook is the problem here. /12
But frankly, as far as I can tell, Facebook is the problem here. /12
There seem to be easy solutions to this that completely preserve privacy.
To get from (A) to (B), Facebook can simply let users of @CrowdTangle sort by link interactions.
They already provide this data
(via the Chrome extension)
They just need to let people sort by it..
To get from (A) to (B), Facebook can simply let users of @CrowdTangle sort by link interactions.


(via the Chrome extension)
They just need to let people sort by it..
To get from to (C), Facebook can simply provide impressions data in @CrowdTangle as an additional datapoint.
Please tell me what the privacy tradeoffs are here. Please. Maybe I'm missing something? From adding a new number?
/12
Please tell me what the privacy tradeoffs are here. Please. Maybe I'm missing something? From adding a new number?
/12
I should also mention that @SocSciOne has worked with Facebook to create a dataset for analyzing interactions with URLs https://twitter.com/SolomonMg/status/1300156232267976705 . This is great, but if I understand correctly, the data is only historical. We need to understand things in near-realtime. /13
So, to wrap this all up.
1) @kevinroose (and others!) can get some of the URL data they need to hold FB accountable today via NewsWhip.
2) CrowdTangle is useful for understanding Pages/Groups.
3) Facebook can & should allow sorting by URLs in CrowdTangle & add "reach" data.
1) @kevinroose (and others!) can get some of the URL data they need to hold FB accountable today via NewsWhip.
2) CrowdTangle is useful for understanding Pages/Groups.
3) Facebook can & should allow sorting by URLs in CrowdTangle & add "reach" data.