One of the two* most common questions we encounter with clients and prospective clients is about Facebook. Specifically, what data is actually available for searching and monitoring, and from whom? That’s because all social media monitoring platforms have revealed, perhaps reluctantly, that useful specific data which was previously available has gone. Users are in a state of slight confusion, but in fact what has gone is quite small and what is left now is quite useful. How it all works is quite simple, once you understand it.

What Facebook data is no longer available?

Facebook previously allowed monitoring services to access a sample of content from personal public posts and Facebook Pages (through the Public API). After a review aimed at upgrading privacy and ensuring that members could feel very confident about the privacy of their data, Facebook removed access to the sample of public posts from individuals. Facebook announced this last year and closed down access in April 2015.

The reason this had an adverse effect on social media monitoring software vendors was that customers were using the software to monitor for brand and topic conversations from the stream of personal public posts. When a relevant mention was identified the user could also be identified, and some action taken, e.g. customer service actions.

Facebook has deliberately and specifically removed the ability to monitor personal posts, and to identify individuals, as a product of its Privacy-By-Design approach to services.

Social media monitoring companies are not headlining this, as it removes a feature which many users found beneficial — especially for customer service and online reputation monitoring. For many social media monitoring solutions it now means, in practice, that users are left with only Twitter for real-time monitoring.

In addition, some social media monitoring companies are offering services which apparently do harvest and provide public posts from individuals. This is achieved by hacking Facebook and collecting data in violation of the Facebook Terms of Service, although some of their brand-name customers seem oblivious to this illegality.

Thus, given that (a) it is not a winner to promote the loss of Facebook personal post monitoring, and (b) false and misleading claims are being made by some social media monitoring vendors, customers are often confused.

However, what customers thought they were losing is often a lot less than they imagined.

The previous access which Facebook provided to personal public posts was only a sample. Even with those solutions that had access to Facebook’s Public API, end users were only receiving a fraction of what was being posted and nowhere near a large enough percentage to perform full customer service or reputation monitoring effectively on Facebook.

Facebook still allows access to Facebook Pages, and monitoring systems can connect those pages to their monitoring analysis.

That means that you can monitor your own Pages or the Pages of others — your competitors — in the same way as before with the same access to and volume of data available to you.

Here is the simple summary, so far:

  • You can no longer collect or see any information from personal posts;
  • What you saw before was a small fraction of public personal posts;
  • Brand Pages — yours and anyone else’s — are still available for monitoring;
  • There are no legal exceptions.

What is Facebook Topic Data, again?

Although you can still connect to and monitor Pages the activity on Pages is dwarfed by activity by people on their personal newsfeeds. In fact Datasift analysed the latter to be 70x the former.

The question facing Facebook was how to provide insights into its global treasure trove of personal activity while adhering to its Privacy-By-Design guidelines?

The answer is Facebook Topic Data, which is made available exclusively through Datasift to third-party social media monitoring and analytics vendors. This latter point is important to know: that the only source of authorised Facebook Topic Data is Datasift, and Datasift approves each and every third party seeking access to that data to ensure they meet Facebook’s privacy and security conditions.

Facebook Topic Data is easy to understand. It is everything you see when you log in to your Facebook account, with everything which could personally identify you or your friends stripped away.

What is left is the actual content, the actions you took (liking, sharing, saving etc.) and all of the demographic data about you (where you live, what you do, age, etc.). Plus, all the information which Facebook knows about you (the Facebook Graph still lives) which is an enormous amount, given that Facebook knows every click you’ve ever made.

Facebook uses the latter information as part of the filtering for content, which is a key part of the power of Topic Data.

When Datasift on behalf of clients (third-party social media monitoring platforms) requests searches on topics from Facebook, Facebook strips all personal identity from the relevant posts it finds and stores the post text and its metadata within Facebook’s infrastructure. It then allows Datasift access to this aggregated and anonymised data under certain conditions.

There are three key conditions:

  1. No post content can be associated with any individual;
  2. There have had to have been at least 1000 accumulated mentions by unique individuals of a specific topic which is the subject of the search; and,
  3. For any segmentation or filtering within the recorded minimum collection of 1000 mentions there have to be contributions by at least 100 unique individuals.

It sounds a little confusing until you realise that the social media monitoring software will set up a topic search and then request Datasift to begin “listening” to Facebook data for that information. Until Facebook finds the first 1000 mentions of that topic by unique individuals it will block Datasift’s access to the pool of topic records it is creating. The key concept to know is that topic searches are not retrospective analyses. Rather, they are monitoring from the time of initiation.

The social media monitoring software will provide ways to visualise and report on the accumulated data — providing more than 1000 mentions from unique individuals has been reached — and each of these filters or requests will be subject to the requirement that the results comprise results from at least 100 unique individuals. For example, you have the 1000 mentions and you request a gender breakdown but the 1000 mentions only has mentions from 89 women. In this case Facebook will refuse the request and no information on gender breakdown will be returned.

Here is the simple summary of this part:

Provided more than 1000 people mention something you are searching for you will be given access to the context of those posts.

If you then want to examine that context in further detail you will only be able to see that detail if there are more than 100 people in the smallest cell of that detail, e.g. age breakdown.

Why is Topic Data so valuable?

Having lost the ability to monitor non-Pages for customer service or real-time interaction you may be wondering why Topic Data is so valuable.

The first thing to get clear is that it is not valuable for real-time monitoring. That is history and that is all over except for Pages.

Topic Data is for analysis, and it is incredibly valuable for one simple reason: Facebook knows so much about so many people.

Twitter is useful, but Twitter knows very little about each user. Facebook knows a lot about you, stretching from the first-party data you have provided (self-declared demographic data including location) through to the graph it has built about your interests and actions.

Not only that, but Facebook has one of the most advanced Artificial Intelligence laboratories in the world and some of the most advanced applications of AI in business. It uses this to understand if a topic — say, a brand or an event or an activity — is the main focus of a post or only an ancillary artefact. In this way it uses “human-like” intelligence to understand stories (every post is a “story”). It also uses similar applications of AI to determine very accurate sentiment analysis, which is accessible to Topic Data users.


Image: Facebook Topic Data Sysomos Scout

Topic Data covers 139 countries. It is not meant to be used to view posts, and cannot be used in that way. It is used to understand trends and data from a macro view.

Topic Data can provide the following types of insights, accurately, and in a way never available before:

  • Which image of our product is most popular in Chicago, New York, London, and Paris among users with Pinterest demographics?
  • Which image of our new game is being shared most globally?
  • Which link to which web site is being shared most about Donald Trump among middle-aged women?
    Find females talking about Hockey in Toronto;
  • Find people talking about sleep apnea devices in North America;
  • What type of content is most popular about banks in Australia — links, photos, videos?
  • What are people saying about Qantas and safety?
  • What are people saying about Obama and healthcare?

These questions sound kind of innocuous — we have been able to search those types of questions in social media before, right?

Well not quite right, actually. Facebook Topic Data allows you to search those across every post on Facebook. That is across 100 per cent of private and public posts. This is a firehose of data like none other that has ever existed before, because of both the volume and the quality of the descriptive data which is associated with each of those posts.

So those searches have the potential to reveal incredible insights. Of course you can track trends over time— for example to find out how the popularity of a TV series is tracking by age or location or gender, or how a link to an external site is tracking across time, or sentiment about a detailed topic.

The difference between this and social media monitoring is that there is no content available, only the analysis of the context within which that content appears.

If you are thinking ahead you might ask “how can we fine tune a topic if we cannot see if the data being found is relevant?”

Facebook provides for this by using Super Public Posts. These are posts are completely anonymised and their text is made available for inspection for the sole purpose of improving the topic formulation and filtering.

Facebook Topic Data allows for unprecedented data access for such purposes as brand research, brand health analysis, product development, share of voice, audience analysis, demand forecasting and a multitude of other research objectives.

The data which was previously available from Facebook Pages to social media monitoring platforms remains essentially the same.

The data available for Topic Data comprises 100 per cent of data on Facebook, whereas before it was a very small subset of public posts.

New reporting products such as Sysomos Scout act as a reporting and analysis layer on top of Facebook Data and add value through innovative user interfaces, insightful graphics and potentially data science applied to the topic data.

Deriving data-led marketing and social media strategies using Topic Data is the newest and one of the most powerful games in town for all businesses.

Previous post

Tech sector has poured over $31B into IoT investments in four years, says Ovum

Next post

Digital spurs an overhaul of approaches to travel, and marketing reflects this