At first look, the Facebook Affinity value that is shown in Audience Insights appears to be a simple metric to picking the top targets for your Facebook advertising dollars. But closer inspection of this metric reveals that you need to think a little before slapping down your dollars.
Simplifying it, Affinity is just a statement of how close entities are to each other. The affinity may be a result of a feeling, an established relationship, or shared characteristics.
Affinity is natural, relative, and measured
In the early 2000's there was a set of commercials out of Portland, Oregon that featured Trunk Monkeys, and the commercials were very successful. One of the reasons is that people had a natural liking and connection - an affinity - to the idea of Trunk Monkeys even thought there was no such thing. I still have a great laugh when I see the commercials. This example of an affiliation is based on a feeling between entities, even if the entities are fictional.
Another example of affinity is a relational affinity. Sometimes there are natural and real affinities such as my relationship with my two daughters. Hey, they are blood and therefore they have a very real relational affinity with me. The same could be said of my granddaughter. Those relational affinities are real and natural. Now I also have a relational affinity with my son-in-law. But that relational affinity is unnatural (wait! don't get upset!) What I mean is that the relational affinity that I have with my son-in-law simply exists because of the relationship that exists between my daughter and him and the laws of marriage.
Another example of affinity is one based on real characteristics. If you were to compare a frog with a desert tortoise you might note that there is an affinity between them as they both share the characteristic of having four legs (those are legs on a frog, aren't they?) Simply having that shared characteristic creates an affinity between the frog and the desert tortoise.
One of the affinity concepts that you may (or may not) have picked up on is that affinity is both relative and measured. Let's delve into the affinity between the frog and desert tortoise a bit more to illustrate these concepts.
If we were to consider only the number of legs as our basis for affinity then we could state that there is a strong affinity between the tortoise and frog, and it is a measured affinity. Right? They both have just four legs (sans the occasional mutation). But what if we consider other characteristics? Shells? Reptile? Diet? Habitat? Once you start adding in additional measurements it quickly becomes apparent that frogs and desert tortoise have a weak affinity. But what if we were to add in toads to the affinity measurement? The affinity between toads and the desert tortoise would likely be the same as the frog to tortoise affinity, but the affinity between frogs and toads would be much greater (heck, can you even tell what the difference is between a frog and a toad?) And relative to each other, the relationship between frogs and toads would be a very strong affinity while the relationship between frogs and tortoise would be a very weak affinity.
So what does the affinity between frogs, tortoise, and toads have to do with Facebook?
Can frogs define Facebook Affinity?
There are really two distinct areas that Facebook applies an affinity definition to. The first area is the relationship between a user (you) and something else (be it another user, a business page, a group, or an organization.) Facebook has a secret sauce recipe - closely guarded - that they use to determine the affinity between a user (you) and something else. Basically if there is a strong affinity between you and something else, Facebook believes you want to see more of that something else along with similar things. So if you are always watching funny cat videos, Facebook may determine that there is a strong affinity between you and funny cat videos and show a lot more of them in your news feed. They might also start putting in a few more videos about funny dogs (if they think that dogs are similar to cats which, IMHO, they aren't even close! Cat's rule, Dog's drool.)
The second area of affinity that Facebook uses is in audience insights. In context of audience insights there is no "you". Rather, Facebook is providing a relative measurement of affinity between a group of "you's" and something else, and then morphing that into a comparison with all Facebook "you's" and that something else.
Gag, that sounded confusing to me so let's see if we can simplify it with another example. And for that we well go back to our frogs... and the wetland in which they live.
Our wetland has 1,000 frogs living in it, and there are four ponds in this wetland that the frogs share. There is a very strong affinity between the frogs and the ponds because all of the frogs live in ponds. But peel back the weeds a little bit for a closer look and start doing a detailed count. You will find that while all frogs live in one of the ponds, there is not an even distribution of frogs to the ponds. In this wetland, 500 frogs live in pond #1, 250 frogs live in pond #2, 250 frogs live in pond #3, and zero frogs live in pond #4.
Thus, while there is a very strong affinity between frogs and ponds, there is a different affinity between the frogs and specific individual ponds. Using percentage as our affinity measurement of choice we can see that the affinity of frogs to a pond is 50% for pond #1, 25% for pond #2, 25% for pond #3, and 0% for pond #4. This is the measured affinity of frogs to individual ponds.
Now, I'm not a statistician so I'm not sure I'm getting the following correct, but it's the concept that's important so humor me...
If I wanted to get a sampling of the frogs I could put out some tasty frog fly-bait and gather up a bunch of them. I would have my own little frog control group (let's just say I have gathered up 20 frogs.) I could then determine - via some closely guarded secret which only I know - which of the ponds each of those frogs lives in.
What's important to note here is that this is not a sampling of frogs in the ponds, rather it is a sampling of frogs that happen to like my tasty frog fly-bait. Out of the 20 frogs that took the bait, let's say that this is the breakdown of pond life:
- 10 from pond 1
- 2 from pond 2
- 8 from pond 3
- 0 from pond 4
Now is when we get into the complexity of relative affinity... I have my frog control group; the frogs that liked my fly-bait. In terms of relative affinity the Facebook way we now need to compare that relative to the frogs that like each of the individual ponds. Doing so results in the below ranking in order of highest affinity of the selected group (the 20 frogs that took the fly-bait) compared to the total number of frogs in the wetland. This affinity order is because - relatively speaking - more frogs from my sample group liked pond 3 over any of the other ponds considering how many total frogs liked each pond. Notice how pond #1 is actually second on the affinity list even though it have the most representative frogs.
- 8 from pond #3
- 10 from pond #1
- 2 from pond #2
- 0 from pond #4
So even though pond #3 has fewer total frogs in it, there is a higher affinity of fly-bait takers in that pond than in pond #1.
Ok, I think it is time to move on to Facebook because you now have an idea how frogs can define Facebook Affinity.
What Exactly do we Mean by Facebook Affinity?
Typically when Facebook Affinity is discussed it is within the context of an individual and some second Facebook entity. For example, from a Facebook user to another user, a business page, group, or organization.
“One of the factors the Facebook News Feed algorithm considers is affinity–that is, how much of a connection you have with each fan. If you can get them liking, commenting on, and sharing more content, you demonstrate a greater affinity and they will see more of your future content.”
(source: BufferApp, November 2014)
But that's not the affinity I am going to discuss. Specifically, I'm going to talk about the affinity shown in Facebook Audience Insights for audience targeting, advertising, and segmentation purposes.
First, I have to let you in on a little bit of inside information... Facebook is quite a bit more complicated than four ponds and a bunch of frogs. So I'm not going to try and explain it all. Partly because it would take a lot more writing and partly because Facebook keeps many of the secrets, well... secret.
But what I can do is look at a real example so you can see the effect of what Facebook is doing and therefore have an understanding of how to interpret and use the information.
For this example I'm going to use a real client and some of the data from that account. This Facebook account has been active for some time and has enough followers to be able to use it as a control group. Our objective is to find pages that we want to use for audience targeting.
In Facebook Audience Insights the location is set to United States and the client's page is put into the Connections > Pages > People connected to dialog box. Those are the only two values that are defined in the audience insights setup page. From here we can go to the Page Likes tab and scroll down to take a look at the Page Likes panel.
The chart displayed for Page Likes is quite easy to understand just from a quick look. First, the Relevance and Affinity columns say pretty much the same thing. They are just a rating of how likely people are to like the referenced page compared to all Facebook users. The difference in the columns is that the Relevance column is purely a ranking from 1 to whatever and no relative strength is shown. The Affinity column is also in order but it provides a relative strength value and bar. To me, the only value that the Relevance column provides is an indication of where the row falls in the entire list (e.g. #35 out of 100), but other than that it is of questionable value.
The Affinity column on the other hand deserves a little bit of scrutiny. One might think that the best thing to do is just sort in order and pick the top ten sites... but don't be so quick. You better do a little analysis and thinking about this.
Facebook Audience Insights Affinity may not have that much affinity
The conventional teaching is that the Facebook Page Affinity score is a closely held secret by Facebook, and I don't have any reason to dispute that as I haven't looked at it closely - and it makes sense. I'm sure that they have some development minds that are a lot bigger than mine working on cooking up a secret sauce - Facebook has stated there are over 100 factors that go into calculating Affinity. But while I do think there is a Page Affinity between a user and a specific page, I'm dubious about the same level of sophistication when calculating the Audience Insights Page Affinity. In fact, I think Facebook is being a little deceptive...
Let's go back to our real-life example and take a closer look. I've sorted the results table by the Affinity column and taken a snippet of the results. You can see that there is a significant affinity difference between the first row results (Marketing for Health Coaches) and the second row results (Center for Nutrition Advocacy). Just a quick glance at this chart would infer that people who are in my control set are much more likely to like Marketing for Health Coaches than any of the other pages.
But does this really mean that my control group has a higher affinity for Marketing for Health Coaches over Center for Nutrition Advocacy? Let's look a little closer by looking at a different snippet.
For this snippet I have used the same control group but have sorted by Audience so I could find audiences that are the same size. I selected all the results (4) that have the audience size of 67. Take a look at the affinity scores for those four results. There is a huge difference between the result in the first row (Marketing for Health Coaches) than all the other results. So why is this? Why is the affinity so much different for the results when the control group audience size is the same?
The answer lies in the Facebook column. This column tells how many total Facebook users like the page (give or take whatever Facebook decides to give or take.) This Facebook column is Facebook's control group. Facebook uses this number to set up their control group as a relative comparison with my control group.
Looking at our snippet again we can see that the Facebook size for Marketing for Health Coaches is 3.7K. They compare the overall Facebook size of 3.7K with my Audience size of 67, make it relative to the total number of Facebook users and then assign it an Affinity value. They do the same thing with all the other rows. Looking at the second row result for Aviva Romm, MD, we can see a Facebook size of 71.3K which Facebook compares to my Audience size of 67 and gives that an Affinity value. In this case, the relative affinity value of Marketing for Health Coaches is much higher than Aviva Romm, MD because the size of the Facebook audience for Aviva Romm, MD is much larger than the Facebook audience for Marketing for Health Coaches.
So let me be very clear about this...
The Facebook Audience Insights Affinity value is very significantly dependent on the size of the total Facebook audience for a given page compared to the size of your control group for that same page. For a given control group size, the larger the Facebook audience size becomes the lower the Affinity score goes. Actual characteristics of the control group do not play a significant role.
Say what???? Yep, you read that right. It is my opinion that - despite all the talk - Facebook algorithms and affinity don't make very much difference to the score that Facebook assigns to the Audience Insights affinity. By far, the greatest impact is simply a numbers game... How many people overall like a page compared to how many people in your control group like that page.
So in looking at the affinity scores, does this mean that Marketing for Health Coaches is a much better group for me to target than all the others? After all, the affinity score is so much higher so more people in my control group like Marketing for Health Coaches. Don't be so quick to jump to a conclusion...
Really understanding affinity requires you to know your history.
I'll admit it, if I were to be handed the affinity report in our example it would be very easy to just say "hey, go after the Marketing with Health Coaches crowd. They really like what you have to offer."
But stop a second and ask why -- why is the affinity for Marketing with Health Coaches so much higher than other groups?
Part of the answer lies in what we have previously discussed; the number of people in the control group that like the page compared to the total number of Facebook users that like the page and relative to the total number of users on Facebook.
But that is only part of the answer for affinity; for the other part we have to look at history. And this history is something you will have to research because it will be unique to each situation.
We know that the affinity score for my control group is higher because the number of people in my control group that like Marketing for Health Coaches is high relative to the Facebook audience size for that page. So the next logical question to ask is "why does my control group tend to like Marketing for Health Coaches?"
History... know your history... and if you don't know it, a little research could prove beneficial
And in this case I have first hand knowledge of the history. I know that a marketing campaign was run specifically to Marketing for Health Coaches which resulted in people in my control group liking Marketing for Health Coaches (or vise-versa). A similar marketing campaign was not conducted for other pages in the affinity list. What this means is that the Audience Insights Affinity score for Marketing for Health Coaches is likely artificially high simply because a campaign was run to get more likes to the page.
Accepting the affinity score as is - and acting on it - could result in amplifying the misconception. By always running campaigns to the highest affinity pages, those pages are solidified in the top spots and greater opportunities are very likely to be overlooked.
In this case, by knowing the history I was able to apply some judgement to the affinity score and decide if the Facebook Audience Insights affinity was naturally high, or a result of some induced cause.
How to take action on Facebook Audience Insights Affinity
You might be thinking after all of this that the Audience Insights Affinity score is somewhat worthless. Well, that really isn't my intent for this article. Rather, I want to change the way you interpret the affinity score. It really does have value, so let's discuss how you should interpret it and come up with an interpretation action guide.
And thirdly, the code is more what you'd call "guidelines" than actual rules. Welcome aboard the Black Pearl, Miss Turner. -Barbossa from Pirates of the Caribbean
How data is interpreted is usually dependent on the answer you want to provide. And when looking at Facebook Audience Insights affinity scores, we want the data to give us audiences that are most likely to be similar to our control group. There will not be a definitive yes/no column that you have access to, but you can still setup your own guidelines (or spreadsheet) to help you interpret the data. Here are my guidelines to help you...
- If Facebook is showing you results for affinity, it is highly likely that there are some shared characteristics between the Facebook results pages and your control group.
- Remember that the affinity scores are relative. The absolute number provided by Facebook has meaning, but it's not absolute meaning.
- Higher affinity scores does not mean it is more likely that members of the result page will also like the same things that your control group likes.
- Magnitude of the affinity score has more meaning than the actual ranking. That is, there is a huge difference between an affinity score of 20 and 2000, but little difference between an affinity score of 2000 and 4000.
- If your affinity scores are all less than 100, keep searching. You need to find better affinity.
- Experiment with your control group to see the affect on affinity.
- Look for opportunities with large Facebook audiences that may not be the highest affinity scores.