Social Media » Half-Life of Social Media Posts
Well, essentially, in every context it’s a length of time taken for something to halve in size. Having not played the game which tops Google’s results for ‘Half-Life’ I can’t comment on how the term becomes relevant when shooting baddies, but in physical terms it’s often used as a constant, predictable, measurable amount of time it takes for a material to decay until it’s half as massive. Given that time again, it’ll halve again. And then again. It’s used in radiocarbon-dating which measures beta decay of carbon-14 as it’s a handy and very accurate way of calculating a long length of time – especially when mammoths didn’t have a stopwatch on them when they died.
It sounds a bit negative, right? If you’re getting your word out there and advertising to the right people your business will grow – not decay over time. But what does reduce is how many people are exposed to each of your messages – or act on your message.
Let’s imagine a scenario where all the figures work out very nicely for this example: imagine sending out a voucher for your product or service to everybody in your address book, and they all landed on the recipients’ doormats with an almighty thump of voucher-on-coir and high expectation. That first day you’ll get a great response, and maybe the next day slightly less, and then less… and in a few weeks maybe just the odd one or two people would still be acting on it.
That rate of change could follow a similar pattern to the radioactive decay of a material. Maybe on the first day a thousand people get in touch. On day two you get 500 responses. Day three brings 250. If day four then brings 125 you can be pretty sure the half life of that campaign is one day. By day five, you’ll be looking around 60-65 responses – just one sixteenth of the figure on that first day. Probably time to get another campaign out soon.
If, however, day one brought one thousand inquiries, then day two earned 707, day three earned 500, day four 354 and day five saw 250 then your half life is looking more like two days – and still with a quarter of the amount of response after five days you can probably sit back a while longer before the next campaign rolls out. Indeed at a two-day half life, you’d have to wait a whole nine days to drop to one-sixteenth of the response. Knowing that your half-life is two days has just saved you sending out a campaign too early and getting a big overlap.
Aha, yes, well you know what else doesn’t last forever? Social media posts. Therefore – in theory – we can also use the same calculations to judge the rate of decay in impressions or reach of your posts on social media, and because of the speedy nature of social we can surely see the results coming in a lot quicker than the direct-mail example above.
If it was, this would be a shorter blog post, and I’d have fewer grey hairs. Let’s look at the issues.
There are two trains of thought in how you could measure a half-life. One way is to wait until the number of new impressions reduces to zero, then look at what amount of time had passed when the number of impressions was exactly half the final total. We’re not keen on that for a couple of reasons – firstly, how do you know it has definitely reached the final total number of impressions? At what point do you say “right that’s it, experiment concluded” – we’re pretty sure if we ever did that a new impression would pop up and make us wait again. A second reason we’re not keen on that way is that, to be completely accurate, it would need a record of the time lapsed whenever an impression occurred which, unless you are only reaching a dozen people, is not going to be recordable. You could instead look at Shares/RTs and Likes instead of impressions but, if nobody Shares/RTs or Likes a post then how are you going to measure it? And what if the 999th person of 1000 happens to like it – that could seriously warp your figures.
Again, for the purposes of this blog let’s transport ourselves to a world where everything works really well and just as the mathematicians promised – even though that’s hardly likely (I’ll get onto that later).
Imagine you already knew a half-life on your Twitter account, you clever soul. Let’s say it’s four minutes (yeah, much much shorter than a direct mail campaign). You’d post your Tweet with a great image, relevant hashtags, at the right time and everything else we recommend doing, and in the first minute you get 100 impressions. By minute number four you’ll be looking at 50 impressions per minute. By minute number eight you’ll be around 25. Minute 12 brings around 12, and by minute 16 it’ll be down to one impression every ten seconds. By minute 20 you’re probably ready to post again.
That’s why it would be handy to know.
>However… we’ve conducted our own experiments of course, and here’s what we found. Let’s stick with Twitter for now – it’s short, sharp and a great way to experiment and gain data for half-life fans.
For one client in particular we’ve looked at three of their different Twitter accounts – each posting regularly but to different audiences and different numbers of people. For each account we’ve done at least 10 tweets, recording the number of impressions each minute. We do at least 10 for each account as there will be variations, but none of these, however obscure, should be discounted. They did happen – but to average things out, the extremely effective against the less effective, those that got RT’d by celebs against those that got no RTs at all – we need as many examples as possible.
This is how it looked.
Although they are different sized audiences, being rated against the number of total impressions after the first minute levels the playing field.
This graph shows two definite trends – the red and blue accounts clearly following the same kind of decay. The orange account, though, shows a different behaviour.
We find this reassuring. The orange account has a different raison d’etre from the other two. The blue and red are corporate affairs which, although generate great content with a fantastic amount of interaction and reach, are not necessarily as free to be pleasing to the consumer – whereas the orange account has content with barely any direct sense of ‘selling you something’ – indeed the orange account can’t even be called up on the phone. It has no call centre or products to buy. However, by having such great content on it, it can direct people to the other brands, and gives those brands’ customers a greater sense of value by being there.
Remember I said “I’ll get onto that later” when talking about it being unlikely to work as predicted? Well, this is late enough so let’s take a look.
If we study the rates of decay you’ll notice it is not a simple half-life. The red and blue show an initial half-life at 3 minutes (roughly where they cross the 50% line), therefore at 6 minutes we’d expect 25% and, you know what, it’s not far off… however, at 9 minutes we’re certainly not down to 12.5%. Nowhere near.
This is due to the nature of Twitter and that not everyone sees it or interacts with it as soon as you post. At any one moment in time less than 10% of your followers will be watching. More like 3% nowadays, in our experience. Unless a follower happens to have selected to turn your notifications on (which would mean their device gives them a prod if you tweet), they probably won’t spot it until they next go onto Twitter. Even then, if they follow lots of people it will have plummeted out of the bottom of their timeline. A keen fan may search for it – something the orange account may utilise to maintain its shallower decline curve – but your content has to be worth searching for.
Side note – Another tool in the orange account’s arsenal (or indeed any account with highly-interesting content) is Twitter’s “What you may have missed” feature. The effects of this won’t show in the graph above as it’s only showing the first 15 minutes, but if your tweet generated enough attention, Twitter will choose to pop it at the top of your followers’ timelines when they next open Twitter hours later and may not have seen what Twitter believed to be a good tweet – further boosting the number of impressions in the long term.
Ok sorry I’m waffling – all this just means we can’t blindly whip up a figure for a half-life on social media – we have to be cleverer than that. Just because the decay is not linear (well, logarithmic, if we’re being pedantic) doesn’t mean the decay and half life is not predictable. To put it another way, although the rate of decay may not be a constant, the rate of change in the rate of decay could be – especially if we also factor in a predicted rate of followers visiting their timelines. This opens a whole new can of differential equations and, to be quite honest, trying to dig out a simple way of plotting this on a graph is not really worth the effort unless you get your kicks out of logarithms and differentiation.
I prefer creating and sharing cool content myself, so – if you’re like me – the best thing to do is run an analysis over many posts, collate the data together, stick it on a graph and rather than figure out a ‘half-life’ as such, just find the point when your posts’ impressions have dropped to less than, say, 20% or 10% of what a new post would get. That figure’s up to you to decide.
Knowing this length of time is just as powerful as being able to say what your half-life for a more traditional marketing campaign is. Find that point when it’s worth creating a new post, and experiment with times of day too. You could spend your time determining some kind of equation to figure this out but when you can look at a graph and see it anyway – why bother?
So there you are – if you’re concerned about finding out the magic ‘half-life’ of your social media posts then you’ve perhaps missed the point of social, and the point of sharing your content. If you’re concerned with simply knowing when you can safely post new content then you’re on the right tracks – that’s what really matters – and if a graph can show you that quickly and accurately then you don’t need to get your logarithms in a twist, and you can spend your time playing Half Life, not calculating it.