The Facebook people talking about metric is made up of the sum of the new likes of the page (more or less the increase in fans) and other kind of  interactions of Facebook users with the page (comments, likes to page posts, shares,…). It is supposed to be a proxy of the capacity of a Facebook page to engage its audience.

The people talking about number should be useful to help users understand whether a page is engaging or not before liking it. If you take the average value of the people talking about this/number of fans ratio of a page (average of six weeks) and you compare it to its maximum and minimum value, the first will be (on average) 1.7 times the average, while the second will be (on average) 0.6 time the average.

Given the large difference between pages with high and and low talking about/total fans ratios, the metric – even taken on a single day – seems to be a good indicator of  how engaging a page is.

This motion chart compares (both in log scale) the people talking about ratio (vertical axis) and the number of fans (horizontal axis) across our six week period. Dots dance up and down a lot and it is not all that true that the smaller a page, the higher the expected people talking about/number of fans ratio.

Another issue that intrigues me is the fact that this metric is the aggregate of new likes and other kinds of user interactions.

New likes are “different” from other types of interactions: two pages with the same number of fans, and the same number of people talking about, but with radically different mix of “new likes/other kinds of user interaction”, tell a different story (BTW this “mix” varies dramatically from day to day).

Lots of pages “buy” fans with Facebook ads and there is absolutely nothing wrong with this. As much as your organic ranking on Google goes up if you invest in Adwords (because Google’s ranking algorithm takes into account also the number of visits that come from sponsored links), your Facebook engagement score will also increase together with the number of new likes and of other interactions with your page that you “bought” via Facebook ads.

This said, in some cases the number of new likes of a page might tell you little about how good a page is at engaging its fans. I separated the talking about ratio in two sub ratios. “New likes” and “engaged” divided by total fans.

“New likes” is calculated as the difference between the number of fans on day t and on day t-7. This value is just an approximation of the value that Facebook plugs into the equation: Facebook computes new likes by actually adding up all the new likes, while my number just is the net between new likes and dislikes.

“Engaged” is just the difference between the talking about number and the variation of number of fans of the previous 7 days. Below is the ranking of the top 25 pages (of the 700-odd we are following), based on the average talking about/number of fans ratio of the last six weeks. The rank  only includes pages with more than 500.000 fans.

I also excluded the pages whose number of new likes is equal or greater to their talking about number.  I do not understand how new likes can be > to number of talking about this. But apparently it happens.

Curious about how your  page is doing vs the rest of the world? This chart shows the 6-week average people talking/total number of fans ratios for 500 pages with more than 500.000 fans. Interestingly, the trend line does not fall all that much up to 5M fans.

What this seems to be saying is that  a priori the people talking about/fans ratio of 5.000.000 page fans is only slightly lower than the people talking about ratio of an 500.000 fan page. Warning: trend line does NOT show average values…

Curious about how your company’s Facebook page portfolio is doing vis a vis other large corporations?

Here I take the total fans of each group and compare them to their total number of “talking about” for the month of November.  Yes and no, “smaller is better”. Many of the big names with relatively low talking about to fans ratio in fact have a lot of pages. For instance Mars has about 20 over 500k like pages (Skittles is the largest with 20M fans).

My favorite Group, Ferrero is also doing pretty well with its official pages, total about 40 million fans (spontaneous Ferrero pages, 11 M fans, post rarely and have much lower “engagement force”). Here you see how this plays out over time.  

The motion chart below separates  two “sub-ratios” (new likes over total likes and other interactions over total fans) and shows how they have moved over the last six weeks for the 700 pages. Guarana’ Antarctica (see previous post) is pretty much the only page that moves in the right side of the chart (ie very high levels of user engagement). To achieve this, Guarana’ might well have run some posts as Facebook sponsored stories, but since they are far from the only Facebook page that uses ads, it would be interesting to know what makes them so much more successful than everybody else. Here is the exact same motion chart, showing only the top 25 pages.

We have been following the “people talking about” metric for more than a month now, on over 700 pages. The current undisputed champion seems to be Guarana’ Antarctica (I am told a great Brasilian beer) with over 2 million fans.

They managed to have up to 80% talking about this to fans ratio.

It is unclear to me what exactly they are doing right and how they are doing it. They seem to get a lot of shares and to post a lot of questions. Other than that difficult to tell.

Apparently (credits to Michael Chiavetta) they launched an ad campaign and have probably ran Facebook ads. The ad was centered on a video that, at least on youtube, seems to have been seen by only a few thousand people.  

In my previous post I said that – for large pages – the people talking about metric was mainly driven by new fans (a number that for large pages is often higher than the number of comments, likes, answers to questions eccetera). In the case of the Guarana’ page, this does not seem to work.

The upper blue line represents our talking about metric. Remember this represents the total of the last seven days. The yellow line represents the increase in number of fans (or likes, whatever you want to call them) relative to seven days before. The read line (unless I have got something horribly wrong) represents the difference).

In other words the read line represents the number of talking about the page that were not becoming new fans, and therefore evidently “talked about” the page in some other way. Wow.

Of course, since the yellow line indicates the net variation of fans (we have no info of “unlikes”) the real number of new fans might be bigger and the difference between the two lines smaller. No idea of what is going on here.

In any case in one month Antarctica’s number of fans went up from 1.3 to 2.0 million. On some days (rather “groups of 7 successive days”) the number people talking about Guarana’ was as high as 1.4 million. For the whole of November the talking about this to fans ratio has never gone under 30%.

Did I write something in a previous post about the fact that “once pages grow over 2 million fans, the 5% ratio becomes virtually unattainable”?

Saude!

Facebook page impressions are driven by the Facebook edgerank algorithm that “decides” whether or not to show a given post on the wall of a given fan. This algorithm works by putting together “affinity” (the history of the interactions of a given fan with the page), “weight” (the number and importance of the interactions of other users with the post) and “decay” (how old the post is).

For fan pages – where the vast majority of fans have never interacted (ie commented/liked a post/answered to a question/checked in,…) with the page after the first like – what really drives the number of impressions is “weight” and “decay”.

“Decay” depends on the page’s publishing policy, and on the page admin’s desire not to get un-liked by fans that might get annoyed if they see too many posts from the page. There are no rules here. What works for you might not work for me.

It is completely wrong to say, in abstraction, that a page should not post more that X times per day and never less than than X times per week. Or that a post should not be longer than X letters, or pretty much any other predefined “this works for everybody” rule (beyond commons sense, of course, but then you already know).

It would be too simple.

What works for you depends on your audience and on how good you are at coming up with engaging content. In order to understand how engaging your content is, it is not necessarily enough just to count likes and comments. You should relate likes and comments to the “type” of post (both in terms of whether it is just an update,  link, or whatever and – more importantly – in terms of its content relative to your brand affinity goals) and then measure comments not only in terms of positive/negative sentiment but in terms of “how” you fans are reacting to the post. Brandbook.sm can help you here.

The edgerank algorithm works on a “post by post” basis – impressions of a given post are largely driven by the number and type of interactions of that specific post. Fans do not get to see stuff that is “engaging on average”, they get to see the most engaging posts – often as defined by the number and quality of the interactions of other fans.

Brands want Facebook pages that engage fans, that have lots of impressions and that get more and more likes every day.

The “number of people talking about this” metric measures exactly this – it sums up stuff that has to do with engagement and the edgerank “weight” parameter (liking, answering a question posted, commenting on or sharing a page post, or other content on a page, like photos, videos or albums) with stuff that has to do with the page’s attractiveness (like the number of new likes, and the number of fans posting on the wall).

Precisely because it is an aggregate measure, the  “number of people talking about this” metric can and does move erratically over short periods of times and can reflect completely different “mixes” of “engagement-driven” and “attractiveness-driven” elements. Therefore insights derived from the observation of this metric – taken in isolation or over short periods of time – often are not meaningful.

This chart shows a 10 day “view” of how the “number of people talking” metric of 700 mostly large pages has moved. You can see the effect of a campaign by Volkswagen USA, and a large jump by Samsung mobile.

If you are talking about a relatively small page, even a temporary change in the number of people talking about this/number of fans ratio probably signals that something is happening – engaging posts, an ad campaign, whatever.

However, in a large page, a large % change of the index (relative to it’s previous value) might not mean anything – unless of course this new value of the ratio remains more or less stable over time.

Analyzing the “mix” in the ratio is also important. If a large part of the “people talking about” total comes from interactions (comments, post likes,…) obviously something is working  in the page’s editorial policy. Does this engagement transform itself in new fans, and in further growth? A page with lots of new fans and a limited number of interactions might indicate a successful ad-driven fan recruiting campain, or simply point to the attractiveness of the brand (while the page might engage poorly or not at all).

By gathering this data every day, and connecting it with other info, brandbook should be able to give marketers useful info to benchmark their pages in a more meaningful way.

One more word of caution is required here.

Brandbook can pull data from different pages, help “distill” the elements that drive engagement, and hopefully give marketers insights on how to optimize the way they manage their page.

However, this does not necessarily mean that it is a good idea to increase impressions by simply studying and selectively replicating whatever it is that makes other brands’ pages more engaging. Page admins should not “just” want impressions and interactions. They want to create impressions and interactions that lead to a greater brand affinity.

“Replicating” another’s page approach might not be consistent with this objective.

While  one would like to post content that is both engaging and relevant to the brand, in practice this is not always easy.

On the other hand, posting engaging content that has nothing to do with the brand will get you impressions and engagement, but little brand affinity, while posting non engaging content that talks about the brand will drive brand affinity, but result in little engagement and few impressions.

Since the edgerank algorithm works on a “post by post” basis, it is unclear whether mixing “engaging but not relevant” and “relevant but non engaging” comments will help in driving impressions and ultimately getting the message you care about in front of the eyes of your fans.

Brandbook can help understand this better. It allows you to tag page posts and fan comments according to any criteria, and will soon be able to link this info to Facebook Insights data about impressions.

If you are wondering, in brandbook, tagging is done manually, both for mapping sentiment (positive, negative, neutral…) and comment content.

Negative/positive are not always “objective” concepts (is the comment “I really love these candies even though they make me fat” positive or negative?) and, while valuable, the positive/negative metric is often not a key info for marketers,  that might be interested in keeping track of how fans react to the page’s post and have little need to be continually reassured that their fans continue to love the brand.

Automatic sentiment tagging will get better but at the moment it has a disturbingly high error rate, and – in my experience – at the moment it is simply not capable of extracting more “subtle” information (what are fans talking about, are they engaged?,…) from short and “noisy” data “bites” such as Facebook fan comments.

Therefore, we have built an efficient and flexible manual tagging interface into brandbook that allows marketers and agencies to outsource tagging at a low cost and with good quality results (as long as the posts are in English and the tagging brief is reasonably simple). To give a number, tagging 10.000 comments (that corresponds to six months for a fairly large, multi million fan page) would costs around 300 dollars.

Now lets dive into our numbers.

I tried analyzing the talking about metric for FMCG pages (mainly confectionery, soft drinks, beauty and home care), ranking the top FMCG groups with the largest numbers of fans across their various pages. My objective was to make myself an idea of how brandbook can help use this metric to help brands and agency optimize the way they manage fan pages.

We are missing some important pages (some of the top Unilever pages are not yet in brandbook’s DB) and we just mapped the people talking about number of the 19th of October (which, as mentioned, can deliver misleading results).

This first chart shows the total number of fans and the people talking about/number of fans ratio (that refers to the seven day 13-19 October period) for the largest FMCG groups on Facebook.

This is the same chart, but it excludes the pages with more than 10M fans (since these big pages might skew our results). The 10M “cut” (totally arbitrary) has no effect on the ranking.

Excluding the over 10M fan pages, Kraft’s average ratio seems to be about four times Ferrero’s.

Does this mean that  Ferrero (or any other brand in a similar position)  has the opportunity of multiplying by four one (key) parameter that drives the impressions of its pages by simply studying and selectively replicating whatever it is that makes Krafts pages engaging?

As mentioned, no.

However, the long answer is that there is information in that number that can be useful, if interpreted correctly, to benchmark one’s page performance with what the rest of the world is doing.

The six FMCG over 10M  fan pages that we have “excluded” are detailed here.  Interestingly, the  larger pages have higher ratios than the smaller ones .

How stable is this difference over time?

It isn’t.

I tried plotting the people talking about this/number of fans ratio over the 2 following days, and, boy, does it move! Avoid drawing quick conclusions from punctual values of the people talking about/number of fans ratio.

While the ratios of Coke and of the two Ferrero pages don’t move much, Skittles drops from 0.8 to 0.6% while Oreo and Starbust both gain around 0.1% (which is a lot, in terms of the absolute variation of the people talking about number)

What is happening?

For instance in the Starbust page went up from 47.000 to 58.000 people talking about it in one day. Over the previous seven days, Starbust added 45.000 likes (average 6.000 per day) net of ( an non known number of) un-likes. Over that same period, the page’s posts got a bit less than 10.000 likes, comments and shares (on average, 1400 per day). The page wall seems to get a few dozens posts per day.

What is moving our metric is a change in the number of new likes to the page.  Measured over only three days, this variation is most probably random and meaningless.

The seven day moving sums for the three days we are considering (19, 20, 21st October) corresponds to the three continuous lines below. The daily “talking about”  value of Starbust for the 19th of October (which we do not know, Facebook publishes only the 7 day moving sum value) must have been  higher than the daily value for the 12 October.

Starbust jumps on the 19th, but Starbust only posted the 11th and the 15th. Therefore a daily spike on the 19th is not tied with variations in the level of interaction with posts (due to aging of old posts and/or to new posts coming up) since no new Starbust post is “missing” on the blue line versus the red one.

Same thing with Oreo, with a small variation. Oreo had tons of interactions with its posts (about 100k in the week preceding the 19th) with an 14k increase in the number of people talking about the page. The difference, once again, is due to an increase in the number of fans since, as seen in the table below, all post “fit” under all three “lines”. The table shows all the interactions of each post directly under the day the post was posted.  Of course this is not how things works, but it is a reasonably accurate simplification.  As an example I took the post of the 9th of October, with 13020 comments. Only 3oo have been posted after the 9th.

In the case of Skittles the fall in the people talking about/number of fans ratio may well – finally – have to do with a (probably non relevant) decrease in the level of engagement of the posts. The posts “under” the orange line got about 30.000 interactions less than the posts “under” the blue line.

It is really important to think of the number of people talking metric as the sum of the fans’ feedback to the page’s “prodding” (comments to posts, likes,…) and the interactions that have do to with the general attractiveness of the page/brand (new likes, comments on the wall,..) – even though of course new likes can be driven by the pages’ posts.

The point I am making is that you can have a decent “people talking about” number even if the page is not posting anything, as long as the brand is powerful and the fan page has many fans and therefore is visible on Facebook. An interesting case to the point is the Ferrero Rocher page, that has not posted in the period we are looking into.

While Rocher is less performing than its “sister” nutella page, it is getting a significant number of new likes per day, which make up most (>90%) of the people talking about this total.

The next chart plots all the pages of each top FMCG group, by number of fans and people talking about ratio (excluding those with more than 10 million fans).

It is a funny distribution.

Smallish pages (between 500k and a million fans) seem to attain relatively high (5 to 10%) values of people talking about/number of fans ratio with relative ease (about 11% of the 200 pages in our sample). Once pages grow over 2 million fans, the 5% ratio becomes virtually unattainable. Samsumg Mobile (2.5 fans and 5.5% ratio) is the only case I found out of 157 pages with over 2 million fans. Most over 2M fan pages lie in the 0.5 to 3% range with significant dispersion also among fairly large pages.

It is clear that – for the part of the metric that is tied to new likes – the bigger the page gets, the more difficult it is to obtain the same growth (measured in added % increment of likes). I would also assume that, the smaller the page, the higher the weight of new likes on the number of people talking about this metric. Small pages with very high ratios are not necessarily posting more engaging stuff and therefore, do not necessarily have higher edgerank and higher number of impressions per 100 fans.

The scatter chart is color coded by FMCG group. All other monitored pages are shown in the background.

Mars and Procter (but this may well be an issue of product line, or even simply a fault of our page list) seem to have lots of relatively small pages (often dedicated to specific national markets) while the other four FMCG groups we are considering seem to be more “concentrated” in fewer bigger “international” pages focused on many markets.

Pages that get 30%-40% people talking about/number of fans ratio attract the attention but it is likely that the “relevant range” in which large and small pages “live” in the medium long term is below 5%. Over 5% you are probably looking at some kind of campaign.

Here is a 10 day chart of car pages. There are a couple of outliers (like Volkswagen as mentioned engaged in a very successful Times Square Facebook campaign) but otherwise all pages are neatly spread horizontally in the 0.5% to 5% band. It is here that the real, medium term, game is played.

So back to our Ferrero Kraft example. Look at the two red Kraft dots up there!

Ferrero here has six pages (average 2.4 million fans) while Kraft has 11 (average 1.3 million fans – two hidden under other dots). Kraft’s best performing (in terms of ratio) pages are two pages (800 and 700k fans respectively) dedicated to Brasilian and Indian fans.

While Brazilian pages seem to enjoy a very healthy people talking about to fans ratio, Kraft’s Hall’s page seems to be a real out-performer. It would be interesting to analyze both the page’s posts (subject, type, frequency,..) the fans’ feedback, and the number of new likes, and to understand how stable this exploit is over time.

Oreo India also seems to be doing an outstanding job, even for India’s Facebook page standards. It might be worth really understanding what they are doing right, to replicate it in existing pages and/or to help establish new page focused on the Indian market (again assuming whatever works in Oreo India is time-enduring and replicable).

In reality it is far from easy to understand what it is exactly that makes Oreo India work so well. It is certainly not enough to look at their wall.

Here is their fan growth. Something punctual has happened at the end of last month. Most probably a very effective Facebook ad campaign.

We could  do a couple of easy things here to get a better idea: (i) measure the underlying “mix” of the people talking about/number of fans ratio (how many comments, how many likes, how many shares, how many everything else), by using brandbook to capture the likes and the comments of the page, (ii) monitor the evolution of the ratio over time,  (iii) understand how fan comments map with page posts.

Brandbook offers a number of tools to help marketers understand what is happening. For instance you can analyse and dig into a tag cloud of the most frequent terms used in comments. It is possible to compare different pages, compare the same page over time, and to build tag clouds only out of the comments tagged in a specific way (for instance positive comments). You can “dig into” each word and create sub-tag cloud with the comments that contain that word.

Brandbook also offers different visualizations to understand the trend in the number of comments and likes a page is getting. For instance you can compare the total number of comments (or likes) across different pages, and analyse it by gender.

Another visualization lets you to build a chart made up of 20.000 tiles, each one of which represents a comment and is color coded according to the tag and to the gender of the fan. You can then filter out comments and zoom into unexpected trends and insights.

We are just doing a simulation here, so I did not do any further analysis on the Indian and Brazilian pages, and just took them away them from our numbers.

Once we exclude all non USA and non European pages, our top FMCG groups on Facebook look like this: the difference between Kraft’s and Ferrero’s ratios has halved.

The remaining gap can be probably explained by the different size (2x) of Ferrero’s and Kraft’s pages – as mentioned the people talking about/number of fans ratio tends to fall rather sharply when pages grow beyond 2 million fans.

But now why exactly is Coca Cola doing so much better than everyone else?

We collect “talking about this” data for most brand pages with more than 500k fans plus some smaller ones we find interesting, about 700 in total.

You can play with the graph to change its settings. The chart “moves” showing the data of the first three days.

The colors represent the category of the pages’ brand (hover with the mouse to see the category and select the corresponding bubbles). You can change the graph and see the pages by target market (we grouped together US pages and international pages, since it is often difficult to tell the difference), or by group (10 groups with more fans, led by Mars, and others together).

It would be interesting to see if the pages with a high people talking about to fans ratio manage to keep it high over time.

For the moment it is clear that smaller pages have an easier time doing engaging. It also seems that BRIC + pages (Brasil, Russia, India, China, Malaysia, Turkey) have on average a higher level of engagement. Maybe simply because they are on average smaller than the USA/international pages.

BTW what did Volkswagen USA do to zoom up that way?

Was it the thrill of seeing your pic on a billboard in Times Square?

This time we took 64 pages, with between 1 and 34 million fans, and calculated the ratio between people talking about this and the number of fans (the ratio is relative to the interactions of the last week). It seems that the bigger you get, the harder it is.

This conclusion is maybe not unexpected (would you want to comment on a post that has already been commented 15.000 times?) but to me it is not easy to explain. Assuming there is a real causality, of course.

The “people talking about this” metric has just been launched. It is important because it introduces some easy bench-marking capability to Facebook pages.  Comparing ten (almost) randomly chosen pages on this metric does not make much sense, especially on a single data point (that relates just to the last seven days). Anyway here are some ratios. Would be nice to know how much they correlate with impressions and edgerank.

Just discovered a (for me) very interesting concept – lean startups. Core ideas are as follows:

1. Most startups fail not because they can’t build the product they set out to build, but because they build the wrong product, take too long to do that, waste a lot of money doing that, and waste a lot of money on sales and marketing trying to sell that wrong product,
2. Startups are more likely to succeed when they rapidly and iteratively test assumptions about a new venture’s business model based on customer feedback, then quickly refine promising concepts and ruthlessly cull the flops
3. Lean startups don’t try to scale up the business until they have product market fit, a magical event-more easily recognized in retrospect than in the moment-when they finally have a solution that matches the problem
4. Of course, in carving a path to the PMF, startups may find that they have to shift the company in a completely new direction. In lean startup lingo, it’s a process known as “pivoting.”

Could not agree more, but how are we doing relative to these ideas ?

1)      Is Brandbook the “right” or the “wrong” product? Unknown at the moment. If Facebook page monitoring is useful and important then Brandbook can become the right product. If there is nothing to distill in fan comments, then it cannot. Considering what Brandbook does, we did not waste any money. However, it took a relatively long time to develop.

2)      We are adding new functionalities to test what works. Some have little to do with the core of the product, to test there is interest. But no, we can’t really say we are “quick” and we release new features overnight. We do not have the resources to be quick.

3)      Our marketing and sales effort at the moment is minimal. And will stay so until point one is solved.

4)      I think we question what we do all the time. And I would not put good money after bad. However, the issue is about inertia more than anything else. How much time do you take to pivot?  Small changes are not pivots, they are bau.