How to hack Twitter’s ‘who to follow’ suggestions. Mass follow the right way.

Disclaimer: This is obviously a strategy that is not without risks. If you follow too many people too quickly (especially with a newish account) you will get banned. Don’t say you weren’t warned.

Back in March, Brent Halliburton wrote a fascinating piece recounting his own experiences experimenting with “mass following” i.e. following a large number of people on Twitter in a relatively short period of time.

What Brent did and his results

Brent followed 9 or 10 people a day in the first few days before ramping things up to a few hundred a day. Unsurprisingly, Brent had a large boost in followers but what was surprisingly was that a lot of these new followers were not the people he had followed

Brent’s hypothesis

Brent’s explanation for this involves Twitter’s ‘who to follow’ tool which looks like this.

Twitter's who to follow tool This is, as the name suggests, just suggestions of people Twitter thinks you might want to follow. If you follow all three it suggests another three. Brent suggests that “following lots of people makes you a better fit for Twitter to recommend you to users” on this tool. Specifically, Brent thinks that if, for example, I follow a bunch of people who talk about “vegan” stuff then Twitter will recommend me to people who are interested in vegan stuff.

My hypothesis

So I think Brent is partly right. My own hypothesis is that the ‘who to follow’ tool is the reason for the new followers but I think the “who to follow” tool recommends people based on who follows them rather than who they follow. So mass following turns the ‘who to follow’ tool to your advantage because of the reciprocal follow backs … assuming these are from people who Twitter thinks are interested in “analytics” stuff for example

My reason for diverging from Brent on this is pretty simple. If I worked for Twitter I’d design the tool my way rather than Brents because users would rather see popular niche accounts in the tool than just accounts that follow a ton of people in their niche … though of course they will get some of the latter when people like myself and Brent mass follow in a niche …

Show me the data

So Brent presents us with some data and helpful charts showing that his follower count did indeed increase and that this correlated with him following large numbers of other people. What we don’t have, though, is data showing that these new followers were not people that had been followed. Until now!

I copied Brent’s methods with one of my own Twitter accounts but this time used data from SocialBro to see how many of my new followers were people that I had followed and how valuable these new followers were. I also compared this to a control period two weeks later when I tweeted and followed as normal

Getting this data was actually pretty simple. After the first week of experimenting (following 600-700 over about a week in a particular niche) I exported a list of all new followers for that period to Excel where I then filtered out those I had followed.

Some graphs showing how mass following works
To the graphs!

Okay so for that week I gained 352 followers with 119 of these being people I hadn’t followed. Compare this to the control week where I gained 57 new followers with 47 of these being followers that I did not follow. So far so good for Brent’s method I thought but then something occurred to me …

What if these new not-followed-followers are themselves bulk following and are just sitting clicking follow-follow-follow in the “who to follow” tool?

To test whether this was the case I went back to my 119 not-followed-followers and manually removed any that I considered spam accounts. My criteria here was extremely salesy profile descriptions and irrelevance (i.e. my account is about vegan stuff and they are a “Mobile Client Management Solution” … Some other examples of spammy descriptions I filtered out include “finally i got more 1000 followerzz after visit this website” and “HAVE BETTER SEX. NATURALLY”). You get the idea. So what did I find?

Well thankfully for my and Brent’s hypothesis, after removing the obvious spammers I was still left with 101 new unfollowed followers. Further supporting mine and Brent’s hypotheses, a lot of these people were interested in things like nutrition, veganism, fitness etc which is what my Twitter account (it’s not my personal one) is about.

Another way I tested for spammers was looking at the engagement levels for the bulk following period versus the engagement levels for my control period. So let’s look at that.

engagementstats18-26mar2014
Here’s the bulk follow week. Click to enlarge. Worth noting I did in fact use retweets but these were old style and done through Buffer so the tool isn’t registering them.

 

And here's the control week
And here’s the control week. Click to enlarge.

 

Well most obviously the conversation rate is a lot higher but this doesn’t really mean anything in light of the number of people who will send a quick tweet saying “thanks for the follow”? The increase in amplification rate and applause rate, however, really suggests that the mass following did generate engagement. Of course, it’s worth noting that the engagement may have come from new followed-followers or not-followed-followers or both. It would be strange though if the followed-followers engaged at a higher rate than the unfollowed-followers though right?

So there it is, mass following can indeed work in a meaningful way but (in addition to my original disclaimer) I would still suggest caution when employing this approach. It’s no secret that following a ton of people can get you banned from Twitter so tread carefully. Yes I got away (and not for the first time) with doing 100 a day but it’s worth noting that the rules don’t seem to be the same for everyone. Do some digging on the blackhatworld forums and you will find people complaining about getting banded for following a lot less people than I or Brent did. These people getting banned seemed to be using newly created or at least very young Twitter accounts for which I think the criteria for getting banned is lower.

So yeah tread carefully and don’t say you weren’t warned.

Have you had any experience (negative or positive) with trying something like this? Let me know in the comments.

3 thoughts on “How to hack Twitter’s ‘who to follow’ suggestions. Mass follow the right way.

  1. Hey Peter! Thanks for a great write-up.

    Previously I’ve viewed irrelevant accounts as at least a +1 to boost my followers number, making my account(s) seem a little bit more authoritative. But if I were to force them to unfollow, it’d make me seem more niched in the eyes of Twitter and thus increase the chances I’m presented as a “Who to follow” to a more relevant target audience.

    Is that a valid take-away?

    1. Hi Tobias,

      glad it was useful! No, afraid I’d say that wasn’t the right takeaway as I think it’s just about the sheer number of people with certain things in their profile descriptions. I don’t think it’d be worth taking the time to tidy up and force the irrelevant people to unfollow.

      Also I think it’s time sensitive. Like the effect drops away after a while. So i think it’s more about who followed you from a niche within last say da or week.

      I’d recommend a mass follow of about a 100 in a niche once a week or at least once a month. It’s hard to find the time for sure.

  2. Hello Peter,

    Good to see a thought-out experiment on this — and thank you for the write-up.

    > The increase in amplification rate and applause rate, however, really suggests that the mass following did generate engagement.

    My impression is the same. Inspired by your article, I’ll try to quantify that (some rainy afternoon).

    Re. getting banned — as you note, it’s not just following activity that matters. I think the key test is to objectively ask yourself: “To an outsider, does my Twitter activity look active-but-reasonable-and-legitimate, or does it look spammy?” Then err on the side of caution. Which is what we do at SoGrow (http://sogrow.sodash.com).

    I believe Twitter use fuzzy spam/quality detection, plus crowd-sourced (do people block your profile), combined with a human check. If you run a useful profile that behaves itself and adds to the Twitter ecosystem, then you should be fine.

Leave a Reply

Your email address will not be published. Required fields are marked *