At Kaleida we’ve been expecting to see more declines in engagements for news on Facebook, but, so far, that’s not what’s happening.
Since the announcement in January that Facebook is deprioritising news, engagements for news have remained surprisingly normal. The trend from the previous 6 months was a steady but considerable slope where the median engagements or the 50th percentile for articles that get shared on Facebook dropped between 25% and 50% depending on the publisher.
But the data indicates that everything just kind of stopped going down.
The chart above shows the median engagement patterns for several of the leading news orgs in the US and the UK that we track.
Publishers were informed in early January that Facebook would deprioritise the posts they make via their own brand Pages in their followers’ newsfeeds. But, so far, those promoted posts are showing normal patterns, too.
While the median level has fallen from the overall average last year, the engagement levels are currently holding steady for promoted posts.
Data from analytics provider Parse.ly shows the same pattern.
Some of our assumptions suggested that the trend line might push engagement declines into late Spring before a plateau. That’s not the pattern here, though these low levels aren’t particularly encouraging to news orgs, either. Better some than none, anyhow.
We’ve been working with publishers who want to understand more clearly what their relationship to the platforms looks like through different lenses. If that’s of interest to you, please get in touch. We’ll do what we can to help.
Contact: email@example.com or reach us on Twitter @kaleidnet.
News publishers can expect web traffic from Facebook to drop by about 35%, according to Kaleida estimates. The important number to track when things do eventually stabilise is going to be “Organic” engagements.
After news got demoted on Facebook publishers began wondering what Organic activity looks like for them. Facebook will no longer push the stories that publishers are promoting, but what about the stories people choose to share on their own without being prompted?
We recalibrated our data a bit to find the patterns and excluded all the activity that Facebook plans to block. It’s clear that some publishers will get hit harder than others, but everyone will be affected.
First, let’s clarify what we mean by “Organic” and how we derive those figures.
We collect the total engagement count for every article directly from Facebook’s published figures. That count is the sum of the number of likes, shares and comments for posts that include a link to an article. Posts with the link to the article may come from Facebook users, and they may come from the publisher of the story who posts it to their Facebook brand page.
We separate the engagement counts for the posts made by the publisher to get a Promoted engagement number.
Organic Engagements = Total Engagements – Promoted Engagements
In this case you can see a significant bump in sharing activity after this article was posted by The New York Times to their Facebook brand page. They also posted it to their home page at nytimes.com about an hour prior to this and kept it there for 10 hours. It’s unclear what effect the home page had in promoting the story on Facebook, but it seems pretty obvious their Facebook promotion had an immediate and significant effect.
The article was averaging 2 engagements per minute prior to the brand page promotion and 35 engagements per minute just moments after it was promoted.
We’re not seeing any particular pattern for different types of publishers. The Organic figure varies by region, politics, size and volume without any obvious distinction. There may be differences in subjects of coverage or article types. We will continue exploring those areas, but here is a snapshot of what we can see by publisher from last week, the 18th to the 25th of January.
This chart suggests most publishers can expect to lose about 35% of their engagements on average after the change is complete — more for those publishers who are reliant on Promoted engagements, less for those whose readers are actively engaging with their stories unprompted.
Not all Promoted activity will disappear. Publishers can still pay for placement. But we can expect most of it to be gone soon.
Fox News, for example, will lose about half of their traffic from Facebook. Breitbart may see nearly all of their Facebook traffic disappear.
Managing the transition to Organic
Prior to the announcement Facebook was training publishers to take control of their presence on the platform, to post more content more often. But now that message has changed and news publishers are all rethinking their strategies for dealing with the company.
In this new world news orgs will need a better understanding of the natural activity happening across the social network, because they won’t be able to influence engagement without paying Facebook for placement. And there are still no guarantees that will work, either.
The “Organic” number will be key. The contrast between Organic and Promoted activity will help news orgs understand when their journalism is resonating on its own and how they can influence the market, if at all.
It would be reasonable to expect that this number will be a more reliable long term metric than total engagements. Facebook could continue demoting users’ posts with news article links in them, too, but at some point they will be affecting active users adversely in addition to publishers.
Facebook values their users more than any other contributor to the Facebook ecosystem, and they will give more thought to throwing away investments their users have made in the platform than they did to news orgs.
Publishers interested in seeing how they stack up against other news orgs can always give us a shout. Kaleida can prepare reports that show whether a trend you are seeing is unique to you or a wider market trend.
Send an email to firstname.lastname@example.org or send us a message via Twitter @kaleidanet.
Facebook announced a considerable change to their news feed that will profoundly affect professional media. We don’t yet know how dramatic the numbers will drop, but there are some indicators that could help us understand what’s going to happen.
What will that curve look like in 2018? If we exclude all the engagements for articles promoted by publishers on their brand pages, because those posts are the ones that will appear less in users’ feeds, then median engagements drops by 50% or more depending on the publisher.
That may not be an apples to apples comparison of the old world to the new one, but we are certainly talking about a big cliff and real impact on publishers.
Engagements for some publishers translate directly into web traffic. Facebook acknowledges that publishers can expect that to drop as a result of the changes:
“We’ll show less public content, including videos and other posts from publishers…Pages making posts that people generally don’t react to or comment on could see the biggest decreases in distribution”
BBC and CNN have huge numbers that surely have a big impact on engagements. When they post a story to their brand page millions of people get that story in their Facebook feed.
A lot of that activity is going to die.
News orgs will still get the pure organic engagements — people sharing their articles without being prompted to do so by the publisher. Though we can’t be sure how much. Facebook will continue to tune their algorithm, and organic sharing by readers may fall, as well.
We’ve taken a snapshot of 2017 performance to use as a benchmark. The table here shows the number of Page Likes for the leading publishers and some engagement figures we’ll be tracking.
So, for example, The New York Times’ mid-point or median engagements prior to this change was 258. That will go down. Most articles will get much less than 258 now, probably closer to 100 engagements.
The leading indicator for future activity will probably be the “Organic” engagements for higher-performing articles, or the 90th percentile. The top 10% of articles earned about 7K engagements or more for The New York Times overall in 2017. That 90th percentile line will probably drop closer to the “Organic” level in this table, or 1,700 engagements, a decrease of 75%.
While articles may still go viral, they will not have the benefit of promotion into the newsfeeds of millions of users who like a publisher’s brand page on Facebook. In the case of The New York Times, that’s over 15M people.
The investments made in Facebook brand pages have been significant. Some publishers worked hard to acquire their followers, and it’s unclear whether they will be able to do anything meaningful with those communities now.
Facebook adds that they will reward content that sparks conversations, “Page posts that generate conversation between people will show higher in News Feed.”
That may help some news orgs, but it will likely be relevant only to niche media or local communities. The change will be frustrating to everyone, exacerbating tensions with Facebook that have been brewing for a long time.
The attempts by Facebook to build bridges with the news ecosystem appear disingenuous now. Though, of course, Facebook may change, as they often do.
There are several positive outcomes publishers may find as a result of this change. Anyone going through a review of their platform strategy will have stronger arguments for investing more in their own developments and working with partners who bring them real value. Perhaps that is long overdue, anyhow.
We’ll be following what’s happening here closely and will report what we find. Please let us know if you want us to look at your content, specifically. We can help publishers understand what the changes might mean for them and how they compare to others who are going through the same thing.
Contact: email@example.com or reach us on Twitter @kaleidnet.
Brexit coverage dominated much of the media this year, as we discovered in Kaleida’s 2017 Year in Review where we looked at news and how attention patterns evolved from month to month. One publisher led most of the activity, but alt media sources are on the rise.
While the vote for Brexit may have surprised many in 2016, the backlash was palpable this year in the form of media coverage and engagement metrics. Nowhere is this more dramatic than with the success of The Independent, a national newspaper in the UK that went digital only in 2016.
Anti-Brexit sentiment was a huge theme this year, and they owned the story in the UK.
Seven of the Top 10 stories about Brexit in 2017 ranked by total Facebook engagements were Independent pieces, all of them anti-Brexit in tone. The first clearly pro-Brexit story in the rankings falls deep on the list at number 21.
Interestingly, the source of that story is not one of the mainstream newspapers typically supporting the government’s views such as The Telegraph, The Times or even the Daily Mail. The story is by alt-media source Westmonster, a blog funded by Arron Banks, the leading funder of the Brexit campaign.
Otherwise, coverage by these leading newspapers is not getting the kind of attention they often get for other subjects, particularly compared to the success of The Independent.
The Independent produced 3,694 articles about Brexit in 2017 which accounts for about 5% of their total output. Only The Guardian and The Times produced as many stories. The other leading newspapers produced half that amount. The digital only news sources published less.
On average The Independent’s stories were about 560 words in length, and most stories earned around 50 engagements each. But when they took off they really took off.
The activity pattern across the year for all Brexit coverage was relatively steady with a dramatic dip in the month of May. It was then that the UK election was in full swing, and media attention was clearly elsewhere.
However, immediately following the election in early June and the chaos that ensued a wave of Brexit activity arrived driven increasingly by alt media sources. Westmonster led the pack, but The Canary, Breitbart, Your News Wire, Brexit Central, Guido Fawkes and Another Angry Voice brought engagement growth for Brexit coverage in the months that followed.
Between them these smaller media orgs published about 1,500 Brexit stories. The most successful was Westmonster’s article, “Morrisons vows to only sell British meat — Massive Brexit boost for farmers” with nearly 50,000 engagements. While this is an order of magnitude smaller than The Independent’s top story the trend line is showing increasing relevance both for Westmonster and Breitbart on Brexit issues.
Interestingly, a small dip in the number of reactions (likes, smiley faces, angry faces, etc.) occurred in the late Summer across all Brexit coverage compared with all other coverage, mainstream and alt media alike. Comments and Shares remained steady or even climbed toward the end of the year.
The increases in attention toward the end of the year point to an interesting change on the horizon. While common sense might suggest Brexit is becoming less interesting over time, the numbers indicate that it is in fact more important to people than ever. Perhaps Brexit is a catalyst for other political discussions happening on Facebook. Perhaps the dramatic nature of the issues and the personalities involved is Facebook fodder.
Regardless, the data is clear that attention for Brexit coverage is still driven by anti-Brexit voices, while a pro-Brexit minority is gaining momentum.
Half of the 1.5M articles tracked by Kaleida in 2017 earned less than 36 engagements. By the end of the year the bottom half accounted for merely 0.03% of the engagements that we counted. Download the full report PDF (11MB).
A change in Facebook’s reporting methodology masks the change, as the highest-performing stories, the ones that really go viral, bring the total number of engagements to new heights while the rest of the market gradually fails to reach readers on the platform.
The implications for news are important to understand.
First, this news wasn’t exactly buried, but it wasn’t given much air time either. Facebook dropped a lot of new stuff on the market in 2017, and making a fundamental change to the core metric of the wider Facebook ecosystem seems like it might have been worth more than a footnote on a blog post.
They have redefined what engagements are several times, and now Facebook has revalued its own currency. The volatility of the system will surely affect Facebook’s relationships with the companies they wish to serve.
Second, it means the growth most large news organisations noticed in the Summer wasn’t of their own doing. That’s not to say their efforts were worthless. It just means publishers may not have much influence on the numbers they are seeing.
Similarly, agencies seeing campaigns taking off in the Summer may have thought they succeeded when in fact it was just numbers changing on reporting tools.
Third, we may be witnessing a decline in Facebook’s influence on news. The new numbers are hiding it in plain sight.
The median average engagements number, the 50th percentile or half of all articles, has been declining most of the year. In Spring the median engagements figure was 36. And now in December that number is down to 23.
Total month-to-month engagements may look encouraging, but the highest performers are the only ones to benefit. The bottom 90% of articles are all in a steady decline.
While it may appear as if the company is obscuring an overall decline by introducing a topline increase, we don’t know what Facebook’s intentions are with this change. It’s conceivable this pattern started well before the change and that we are only now seeing truth with more accurate figures than what we had access to before.
Sampling would surely skew toward flatter growth in a viral system, and now that they report ‘real’ engagements, as they claim, we might be seeing patterns that have been there for years.
Regardless, the larger trend is not good news for news. If most of the news is getting shared less and less on Facebook then publishers will likely also see a reduction in an important source of customer visits, both new customers and loyal customers.
The true test is the amount of referral traffic to publishers from Facebook. A decline in both referrals and engagements would have serious implications for the industry. Recent analysis from Parse.ly suggests this is already happening.
The data may be getting clearer, but the reality behind what that data is showing us may not be what news media wants to see.
Download the full End of Year 2017 report
Kaleida is a data services and media research company. The company provides data, tools and analysis about the attention economy to companies who do business in the media ecosystem.
News publishers can learn about platform referral traffic and get performance benchmarks to compare against the market by joining Kaleida’s news referral research project: https://survey.kaleida.com/
The idea for the Internet’s news desk is based on things some of us have been thinking about while trying to solve problems for the journalism trade over the years. Publish.org is off to a good start, but it needs writers to write, editors to edit, and members to help fund the journalism that gets made via this new platform.
A few years ago I spent some time sitting with the news desk at The Guardian. I was trying to understand the process a little better so I could recommend ways we could apply some of the fantastic technology resources in the building to improve things.
The most eye-opening thing for me was seeing how much goes into producing the stories every day. As readers that process is totally invisible to us.
We don’t know about all the stories that weren’t good enough to get published. We don’t know about the interrogation Guardian editors give reporters and each other with every piece. We don’t know about the context brought through from deep knowledge of a beat, the history of a story and the framing for it amongst all the other things being published. We don’t know about the checks and balances, the policies and routines, and the standards that professional journalists have built up over the years. And we don’t know about all the people involved in a story ensuring it meets those standards.
It’s so easy for anyone to put words on a page on the Internet and to make it look the same as any other news web site that the casual observer just doesn’t know the difference anymore.
Adding to the problem for us as observers is the fact that the process is closed and rigid. Editors protect the process because it works. Great journalism is very hard. And the process I saw delivers great results. But in today’s connected world this kind of process could be opened up in ways that could make it even better and certainly more accessible to people who are hungry to be part of the journalism community.
What would an open news desk for the Internet look like?
We tried a version of the idea at The Guardian with crowdfunding cooked into the model. We called it Contributoria. It was mostly successful, actually, but having a corporate owner looking for short term ROI was an impossible position for a community platform. Plus, they already had one of the best news desks in the world.
So, we closed it down, left The Guardian and started over completely, this time as an independent nonprofit with a much longer term view on this challenge.
Now we can genuinely put the community first. We have to. It’s codified in the structure of the organization. And we can make it better even though the core intent is the same.
What we’re building at Publish.org is the Internet’s news desk.
The whole point is to make quality independent journalism possible at a global scale by creating a process that is both open to a community that wants journalism to succeed and also organized to maintain high standards. It’s not about left or right or any political agenda. It’s about enabling journalism for the benefit of everyone.
Of course, being nonprofit doesn’t mean we don’t have to make money. We can’t operate without money. Being nonprofit means that our profits, or rather “surplus funds”, get reinvested back into the community we serve.
The more we collect, the better journalism will be served.
We raised €160,000 so far this year which we have used to build the core platform and to start paying journalists.
Next year we are going to need €300,000. I’ve posted about that over here if you want to know more about the financial plan.
There are lots of ways to help us reach that goal. We are actively looking for foundation grants, coverage sponsors, and paying members. Let us know if you are interested in any of those things.
The other thing you can do, whether you’ve joined or not, is to introduce us to your friends and peers. Send an email. Post on Facebook. Tell us who we should reach out to.
Every contribution is critical to our ability to operate.
We are aware that we aren’t the only journalism outlet looking for your support. We genuinely hope that you are supporting others, too. Journalism needs as much support as it can get.
We hope you support Publish.org because you value professional journalism and that you want to make that process accessible to more people. Help us to distribute some of the best things about the traditional news room into the wild open Internet for the benefit of everyone.
Articles that go viral on Facebook show increases in Reactions in greater numbers than either Comments or Shares. Football stories are the category leader. But do those engagements convert to traffic to the publishers’ web sites? Our new study tackles this and other related questions.
While digging through all the news articles published last month to produce the October rankings report we found an interesting insight about football coverage.
The engagement number we get from Facebook for each article is the sum of the comments, the number of shares of the URL, and the reactions associated with the link. Reactions include likes, angry faces, laughing faces, etc. CTO Graham Tackley wrote up an explainer about all this here.
The median number of engagements for news articles in October was 26. That’s half of all news articles get fewer than 26 engagements. But there are many articles that do much better than that, and it turns out that the split of comments, shares and reactions looks different for articles that perform below average vs those that perform better than average.
What we found was that the number of reactions increases disproportionately to comments and link shares. Articles that succeed on Facebook and have particularly high engagement counts are mostly driven by reactions.
We looked more closely to find out what kinds of articles demonstrated this behavior most clearly. That turned up the second interesting data point.
Football was the answer. “Real” football, as my British friends remind me every time we talk about the sport also known as soccer.
Sports, in general, performed well by this measure, but football demonstrated it more than any other type of coverage.
If you download the spreadsheet of October news articles and calculate which stories have the highest reaction counts as a percentage of total engagements you’ll find rows and rows of stories about football — everything from comments from coaches and stories about top players to trades and rumors to scores and previews covering all the European leagues, the Champions League and the World Cup.
The list is dominated by the Daily Mail, The Telegraph, Mirror Online, and HuffPost.
The key question is how well this activity converts to referral traffic back to the publishers’ web sites. More specifically, publishers need to understand whether reactions for their stories have a higher or lower correlation with referral traffic. That result will indicate whether this pattern is worth encouraging or if it only helps Facebook.
If reactions on Facebook are easier to achieve off the back of football coverage than other stories, and if those reactions covert to clicks to the publishers’ web sites then it would make sense to beef up the editorial resources in this area.
It’s clear from some referral analysis we’ve done that coverage of sports is an important part of any news publisher’s portfolio. It accounts for over 10% of referral traffic for some publishers.
Not all publishers will show the same patterns, but now that we know what is working for some news orgs we can help identify which publishing strategies work best for others.
About the study
If you would like Kaleida to look at your data and see how we can help please do get in touch. By joining our upcoming referral research study you can see how you compare with the market.
The Referral Impact Study will look at traffic patterns driving readers to news. It will size the market and put numbers behind key questions news publishers need to understand.
Participating publishers will get a custom referral report highlighting their performance and benchmarking with the wider news ecosystem.
It will be produced independently with support from our partners at Google.
First, we’re seeing an increasingly steep power law curve or Long Tail for article engagements. Half of all the articles published in September that we tracked earned 26 engagements or less.
My first thought was that we must be tracking more articles. While that happens to be true (152,658 articles in September vs 148,391 in August) the performance across the board was down vs the previous month.
The top 1% of articles in September earned 24,236 engagements or higher vs 27,998 in August, a decline of 13% month-to-month. And when you look closer to the mean average or the mid-range at the 90th percentile, the September figure was 1,717 engagements vs August’s 2,085 engagements.
In other words, the top 10% of articles declined nearly 18% month-to-month.
The declining engagement trend may or may not affect all subjects and content types the same way, as we found in a mini case study we conducted recently. And the trend does not necessarily reflect declines in overall referral traffic, though it would obviously suggest Facebook is driving fewer visits to news media from month-to-month.
Publishers knew this already — they have to work harder to generate their own traffic.
That then leads to the question of efficiency.
Publishing efficiency is something everyone is looking at across the news media.
There are many dials for determining what publishing efficiency means. It might mean contributor vs staff output. It might mean optimising for average visits per article per desk.
Sometimes publishers struggle to know what their own internal activity looks like, so we can use Kaleida’s data to look at efficiency in terms of output vs engagements on Facebook.
Looking at efficiency through this lens shows that CNN leads the pack. Though ProPublica is remarkably efficient given their output volume.
Clearly, more is not always better. And web sites employing some sort of paid content strategy are going to see different engagement behaviours than those that publish content openly and for free.
Regardless, it is surprising to see the difference in output volume from publisher to publisher. The Daily Mail is publishing over 20,000 articles every month. The other UK newspapers are publishing between 6,000 and 10,000 articles.
Finding differences between publishers with similar levels of output can tease out some interesting strategies.
The Guardian published about 8,500 articles in September. The Telegraph published 7,400 articles. Both are producing between 250 and 300 stories per day, on average, so their output is at similar levels. But The Guardian is performing above average on Facebook. They had six stories earning over 100,000 engagements, while The Telegraph had just two. The Guardian’s median engagement level is strong at over 100. The Telegraph’s is mediocre at 26.
Both publishers had strong engagement performance on breaking news such as the Mexico earthquake, but The Guardian appears to get a lot of traction from stories about scientific or academic reports. Seven of their stories that made it to the top 5% of all articles in September were about mental health research, climate studies or something similar.
Now, not all articles are alike. And that makes it hard to count them sometimes.
In a recent blog post by Newswhip the company reported that Fox News published 50,000 articles in September! We’re not seeing numbers like that at Kaleida. It seems inconceivable any publisher could produce that many pieces even if they published a full news feed from Reuters and AP. The Washington Post publishes entire wire feeds and releases about 25,000 articles a month in total.
To be fair, publishers may not even know how many articles they released last month. And sometimes a story will have duplicate URLs, or a component of a story such as a video will appear in multiple places making it difficult to track.
But publishers do need to know these things. The whole market needs to know what efficiency looks like. Those kinds of benchmarks are going to help everyone in the business, as the news media may not always be able to rely on distribution channels like Facebook to find readers.
If the decline in engagements continues at this rate it won’t be long before half of all output gets no engagements at all.
But it was HuffPost and The New York Times who were able to contextualize the news in ways that really resonated with people on Facebook. The collection of material published by these two sources outpaced everyone else in the market.
HuffPost’s highest performer was an article with a long stream of photos and video of the day’s events: “Look At What’s Happening In America In 2017 — The Charlottesville rally is just another example of the deep-rooted, bone-chilling hatred in this country.”
It was featured on their home page all day Sunday, and it was promoted on their Facebook page where they have 9M followers saying, “If you’re thinking, ‘It’s not that bad,’ then you’re not paying attention.”
The New York Times had 3 of the top 10 most popular stories. The first was an article published prior to the rally on Friday the 12th:
“A month after a Ku Klux Klan rally here ended with the police using tear gas on protesters, Charlottesville is bracing for a weekend of white nationalist demonstrations and counterprotests, and suddenly this tranquil college town feels like a city under siege.”
Fox News made the top 10 for it story on ESPN’s decision to pull one of its college football announcers from broadcasting the University of Virginia game that week because his name is Robert Lee.
One interesting insight worth considering is how HuffPost clocks significantly higher commenting volume compared to all the others. Most publishers in the top 10 see about 60% of their total Facebook engagements attributed to commenting. HuffPost often sees well over 80% of their Facebook engagements from commenting.
While most of the others see similar patterns for reactions and URL sharing, around 25% and 10% respectively, Fox News does stand out for considerably higher reactions to their ESPN story. They earned over 40% of their Facebook engagements from reactions.
Whether performance of coverage was good or not good enough depends on if you were a subject of one of the many big stories coming out of Washington that week, too.
Everyone saw the response across the media to the announcement that transgender military personnel would no longer be permitted to serve. It was a robust response on both the left and right sides of the political spectrum. It was particularly strong given the other activity surrounding the White House that week.
Not only were news orgs covering the announcement in depth, but they were promoting the coverage heavily. And not surprisingly, readers were very actively engaged, spreading that coverage across Facebook in phenomenal numbers.
It was by all accounts a high impact news event getting traction with publishers and with readers.
Engagement volume was much higher than other coverage. The median number of engagements across all articles published in July was 31, while the median for coverage of transgender people in the military was 381.
Engagement activity was split 26% reactions (including likes, frowns, smiley faces, etc.), 66% comments, and 8% shares. This breakdown of engagement activities was normal compared to all articles in July.
This strong performance happened despite competition from several big stories that week. At the time of the announcement about transgender people in the military on the 26th of July Trump was facing criticism for changes in his staff.
He announced the resignation of Sean Spicer and his replacement Anthony Scaramucci a week before the announcement. He was two days away from replacing his Chief of Staff. And Trump was losing support from Republican senators as he publicly attacked his Attorney General Jeff Sessions.
Whether or not those are coincidental strings of events it’s worth noting that the New Yorker interview with Anthony Scaramucci that ultimately led to his resignation was published on the 27th. The transgender policy was announced the day before that interview was published. Of course we all heard about it, because the Scaramucci story flew across the Internet earning 870,000 Facebook engagements for The New Yorker and big numbers for all the publishers who covered the interview, too.
The data can’t tell us if the events are related, but the timing of them and the scale of the response suggests an interesting change. Perhaps we’re seeing media orgs and readers learn how to spread attention across several big stories in parallel.
Was it the level of drama that made this possible or is there something new about the way publishing and reading habits are evolving?