You’re a shark. Sharks are winners, and they don’t look back because they have no necks. Necks are for sheep.

That’s right, a week after Shark Week, we’re putting out a post about sharks. TIMELY! But this isn’t really about sharks – no, it’s a combination case study, admonition, and guidebook for not being a sheep.Because you don’t want to be a sheep. You want to be a shark. Because sharks have no necks!

At this point, you’re probably wondering wtf I’m talking about. There is a tendency in marketing to succumb to group-think – the sheep mentality. Someone writes a great blog post, it makes the round of the internet, and by next Thursday every single person in social media has seen it and is following it to the letter. And there’s nothing really wrong with that, per se. If you’re new to the field (and given that “the field” has existed for all of a decade now, who isn’t?), or you don’t have a lot of time or resources, I can see the appeal of reading something from HubSpot and thinking “Wow! That chart is really pretty, and I don’t REALLY want to be awesome – I just want to be good enough to not get fired!” If that’s you, this isn’t your post. This is for people who DO want to be awesome. The people willing to take big risks, and reap big rewards (and sometimes eat a BIG helping of humble pie when things go south). This is the story of how we stopped being sheep and started being sharks.

So What’s The Big News?

  Take a look at this graph. If you work with Facebook, you know exactly what that is, and you probably spend a good deal of your time staring at these things trying to benchmark how you’re doing compared to your competitors. If you work for a really large agency, you might have some tools that stare at this for you and give you something prettier, but the point remains the same. There’s a nice little bump at the end, but there’s nothing special about that, right? You often see that shape when you’ve just started running a contest, or when you pump a few bucks into advertising, or just from general usage patterns. This is the “People Talking About This” graph for MIMOBOT. It’s a happy graph, but not something to blow your top for. Until you see this:   That’s the other part of that happy graph, and it’s what we use to calculate a rough estimate of how well a page is performing that we like to call the Engagement Rate. Engagement Rate is calculated by dividing People Talking About This by Total Likes. It’s not the best metric, for a number of reasons, but it’s stable and can be seen publicly, making it a great little number for comparing Facebook pages. We generally like to see it between about 3 and 6%. That’s a good baseline, and is slightly better than most Facebook pages pull off. It tells you that you’re doing something right, and people are responding to your posted content. Some larger pages with millions of likes, like Coca-Cola, struggle to keep it around 1-3% (it’s about 1.4% for coke right now) due to the sheer difficulty of engaging THAT MANY FANS. Really social and popular brands, like Black Milk Clothing, manage to run numbers of 6-8% regularly (it’s at 6.5% there now). That’s considered pretty damn spiffy. If you’ve brought out your calculators and done some quick math, you already see why this is an exciting happy picture. That’s a big number. A big happy shiny awesome number. And it’s been going up every day, with nary a drop in sight. In fact, if you want to look at the Engagement Rate across two weeks ago, it’s absolutely phenomenal. Let’s do that, shall we?   From the minute the new strategy started, we had an almost 100% growth in engagement, overnight. It was cool to see it in action, but these kinds of spikes happen, and often dissipate, so there was a lot of caution to the optimism. Fortunately, the caution seemed to be no end in sight for growth. It has been almost a full month since the strategy began, and we’re still seeing strong growth, occasionally hitting an engagement rate of 20% before new likes catch up and drop it down again.

Enough Bragging: How Did You Do It?

The short and simple answer is we did it by ignoring all the common sense and a good bit of the much-vaunted research that so often gets thrown around in the social media scene. We looked at the information we had in front of us. We ran tests, and compared the results we got to what we expected to get. And in the end, we threw away the preconceived notions that didn’t fit the data, instead of molding the data to fit an existing narrative – something too few agencies do these days. The long answer starts with a list of what many social agencies consider to be the golden rules of social. The preconceptions and common-sense guidelines that I mentioned earlier:

1. Facebook users are fickle, and don’t like being exposed to too much from one company
2. Facebook’s algorithm changes make it impossible to reach a sizable chunk of your potential audience without paying
3. The half-life of a Facebook post is short, and barely worth tracking any significant amount of time after posting

These axioms have been around for a while, and there is lots of very good research out there backing them up. I am not a dedicated researcher, and my sample size is too small to make any hard statements, so maybe this is a fluke. I don’t think so, though. I think in the rush to embrace the mantle of science and big data, a lot of people are out there conducting poor research and shoehorning evidence into whatever theory they happen to embrace.

The Initial Research

We began with as scientific an approach to Facebook testing as is possible. Obviously, since FB has yet to release, or even make feasible, a true testing platform (we’re working on that, don’t worry!), we had to make do with the closest thing we could: a rigorously designed testing matrix that would run for about a month and a half and would cycle through every possible permutation of day-of-week, time-of-day, posts-per-day, content-of-post, etc. Essentially, we kicked around every variable we could think of that would influence posts, and tried to find a way to isolate as many as possible so that we could later go back and build some sort of regression model to help us predict the factors that actually mattered. What we got was roughly 50 days of posts planned out in what can only be described as the spreadsheet from hell. It was horrible. Truly horrible. But it was as close to real science as we could get. And it came back with some awesome results. I’m not going to share the spreadsheet, or the data, mainly because a) I think that it’s important to do it on your own, b) I think there is a shocking amount of variance from page to page, so you need to run one for yourself to see what happens, and mainly c) We haven’t gotten sign-of from our client to release it. Suffice it to say there were a lot of numbers and it gave me a headache. Running through a regression analysis is pretty easy stuff, and has been explained over and over again, so I won’t go into details. Instead, lets get to the good part: one of the very interesting things we found was that while post reach, page and post likes, and engagement was correlated positively with post frequency fairly strongly, page and post unlikes and other negative actions were only very weakly correlated. In layman’s terms:

The more you post, the more people engage with you AND the more they tend to unlike your page or post. HOWEVER, the engagement gains are MUCH higher than the engagement losses.

 Do Facebook Users Dislike Oversharing?

In a word, no. Not according to what we found. In fact, based on our preliminary results, which went up to four posts per day tops, there seemed not to be an absolute limit to how much a brand could share and still keep net engagement and net likes positive. Contrast this with the tons of articles that talk about being as cautious as possible. Here’s one with comically bad methodology that says you should post three times a week, and no more than once per day (here are some more). Why is their methodology bad? Well, for starters, they preface the section with data from 2011. That’s two human years, and about 5,000 social years, out of date. More importantly, they don’t define shit. Let’s look at a quote:

Companies who posted once every other day had 32% higher like rates and 73% higher comment rates compared to those posting 3 or more times a day.

Higher like rates per post? Higher like rates on the page in general? Higher like rates on the sum of posts on a given day? And are they accounting for page types, brand, content types, context, and other critical factors in these rates? My guess is a broad no to the latter. As far as the former, I suspect that they used publicly available information rather than asking the companies in the study to provide data. As such, the “rates” they are most likely talking about is “likes per post divided by total fans” and “comments per post divided by total fans”. Sum up these like and comment rates, then average them out per day (or week? or month?? We can’t say, because they DON’T TELL US! This is how NOT to report on a study). So now you get some great little percentages that tell you…well, absolutely nothing. Because at the end of the day, if I put up 20 post in a week, and each post gets 2 unique likes, I have 40 unique likes at the end of the week. Whereas if I put up 3 posts in a week, and each one gets 10 unique likes, I still only have 30. 40 is more than 30. More is better. Better wins! Huzzah!

Once we decided to throw caution to the wind and follow our research to its ultimate logical conclusion, we set an ambitious schedule of posting 12 status updates EVERY. SINGLE. DAY. Starting at about 10AM Eastern, we posted one update every hour, on the hour. Ballsy, I know, and we were a little terrified that things would go poorly. My hand was on the plug, ready to cut things short the minute people got pissed off and started leaving in droves. Except they didn’t. They loved it. On our first day, posting 12 updates in one day, we had 254 total likes, 65 comments, and 41 shares. Total. Compared to 73 likes, 10 shares, and 4 comments the day before. The next day was even better, with 656 likes, 54 comments, and 130 shares. People obviously liked what we were doing. The numbers continued to stay high. We’ve been at it for 20 days now. Here’s how it compares to the previous 20-day period:

	07.16.13-08.05.13	08.06.13-08.28.13
Likes	74	402
Comments	11	55
Shares	12	85

 

Those are the average numbers per day for the total sum of every post put up that day. Some pretty good gains, no? Best of all, they’ve been consistent. Bestest of all, thanks to Facebook’s new handy fan/non-fan reach breakdown, we can see that our reach among non-fans has increased dramatically. We haven’t fully parsed those numbers yet, but we’re getting to it, and might add them in when they’re done.

Conclusion: Your Facebook fans are not opposed to seeing a lot of messages from you. Actually, that’s part of the story, we’ll get to the rest later. A better conclusion is that posting more often will increase the number of unique likes, comments, and shares that you will receive within a given period. Each individual post will likely get slightly less than you were averaging before, but on the whole the result will be a HUGE net positive.

One more note before moving on: there is a whole school of social media gurus and experts that like to say “post only when you have something interesting, relevant, or current to say.” To which I reply, you’re content creators, damnit. You should ALWAYS have something interesting, relevant, or current to say. If you don’t, come up with something. If you can’t, get out of the business.

Generic Whining About EdgeRank Filtering

This is our second sacred cow, slaughtered on the altar of science. Since Facebook introduced some pretty strict filtering last year, the general complaint is that it has become impossible to reach your total possible demographic without paying out the ass for the privilege. And in a very real sense, this is true. You will never see the kind of reach numbers that were common in the heady days of late 2011. Gone are the posts that were regularly seen by 65% to 75% of your audience, plus a good chunk of their friends. Contrary to popular belief, though, this is a good thing. It culls the chaff from the wheat, and allows really exceptional Facebook marketers to stand out. How? By getting strong engagement! The more people that like, comment on, and share your posts, the more people will see each individual post. By posting more frequently, we increased the total number of likes, comments, and shares (as we mentioned above). At the same time, we’ve slightly decreased the per-post engagement rate. So in theory, we should see that our total page reach has increased dramatically, and our post reach should have decreased slightly. In fact, what we saw was that our average post reach doubled, while the average page reach increased by about 400%. The latter number fits in perfectly with our predictions (it’s slightly lower than the increases in likes, comments, and shares, but still significantly improved. The discrepancy between how much likes, comments, and shares went up is likely a combination of filtering and duplication (if me and you have the same group of friends, and we both share the same post, reach isn’t going to double, since we’d be reaching the same people, only twice). The former number, however, was a bit of a shocker. We expected it to go down slightly, or possibly remain the same. Part of that discrepancy might be due to a particularly phenomenal post halfway through the period that broke all records for the page. The other part, however, is that once your fans interact with any post, they are much more likely to see other updates you’ve posted. It becomes a kind of virtuous cycle, though one with an attainable ceiling. Currently, daily engagement for the page sits at just under 100% of fans (though that’s not entirely accurate, since a decent portion of that reach is spread out to non-fans). Either way, it’s a good number, and probably about the limit you can get through pure mechanics. Over the next couple of months, the goal is to get it to roughly twice that using optimized content, but that’s a post for another day.

FB Post Shelf-Life: How Long Should A Post Engage?

One of the goals behind this experiment was to try to overcome some of the shortness of a Facebook post’s shelf-life. Estimates put post shelf life at about 3 hours. This is hardly a huge window of opportunity to try and grab someone’s attention, right? Our thought was that by introducing more posts, we would artificially expand the window of who we could target with our posts. One thing that has played hell with this whole endeavor is Facebook’s incessant tweaking of the algorithm that powers who sees what. There have been at least 3 major changes since we started doing this, and god knows how many minor ones. A lot of the changes have to do with how FB determines who sees what, how often they see it, and when and where it pops up in their feed. All that aside, what we found was that as soon as we started posting more frequently, our post window expanded greatly. Instead of getting the lion’s share of impressions in the first hour, followed by two hours of stragglers, our posts actually tend to under-perform reach expectations (though slightly over-perform engagement expectations) in the first hour, then significantly over-perform on reach AND engagement over the next 4-5 hours. In fact, many of the posts get 50%+ of their total reach more than 2 HOURS after posting. So much for a 3-hour shelf-life. We haven’t fully explained this to our own internal satisfaction, so we’re going on some speculation and educated guessing at this point. My thoughts are that FB is smart enough not to display a post from MIMOBOT every hour to an engaged fan, but instead picks a handful of posts that it has determined they will be more likely to interact with,  and then spreads them out and shows them at times you are most likely to interact with them. If so, this is pretty awesome (and we have some evidence that this is so, not least of which is FB’s own admission on where their algorithm is going). Essentially, FB has taken the work of apps like Buffer or PostRocket and integrated it into their own functionality. Either way, we found that saying FB posts only last 3 hours is kind of a load of crap, at least in this very specific case. We’ll keep looking into it, and report what we find.

Conclusion

In less then a month, we increased engagement rate by over 400% and reach by almost 700% from our baseline, all without spending a single penny on FB advertising (that’s a bit of a lie – we spent some money, but on specific product promotions that were excluded from this post). We did it by following the trail of data without trying to guess where we would wind up at the end. Most importantly, we weren’t afraid to throw out established axioms and take a chance. That’s social marketing like a shark – you don’t look back on what the experts and the guru’s tell you SHOULD happen. You look at what IS happening, and figure out strategies to shape those existing trends to your advantage. Instead of forcing the direction of your strategy, you pick an existing thread, and give it a good (but controlled, and always cautious) tug to see what happens. Sometimes, the whole sweater may unravel. That’s a risk, but it’s not a huge risk so long as you take precautions and stay involved. More often than not, though, you’ll find out that you can get much better results than by relying on conventional wisdom and common sense. So, if you want to social market like a shark, look at the data. Break it down to the components you can control, and see what happens when you mess with them. Don’t rely on the wisdom of sages or social superstars. Things that worked for them might not work for you, and what works for one brand might be suicide to another. Observe, experiment, confirm, repeat. Done.