followercount

“How the hell do you have 18,000 followers on Instagram???”

I get this question every few weeks when another one of my friends joins Instagram and takes a look at my profile. They often wonder why a freelance designer from Streatham has so many people interested in his sunset and food photos.

I’ve never spilled the beans about how I managed to increase my followers to 18,000 so quickly.

So now it’s time to come clean –

Last year I teamed up with Brendan (@oh_moore) and created an Instagram bot that automated many of the manual interactions I had to go through to get more followers (i.e liking and following).

Essentially, it pretended to be me and by doing so, it generated several thousand downloads for our iPhone app, InstaBAM.

The following post is the story behind it, I tried my best to document everything so it’s a long read. If you want to head straight to the code it’s available on Github.

How it all started

In early 2011 we started a side-project called InstaBAM, a free iPhone app which lets users find photos on Instagram by using the geolocation tags on pictures uploaded to the service.

The app was a fun experiment and was the one of the first iPhone apps to use the Instagram API, if you searched for ‘Instagram’ in the app store, we appeared just two spaces below it in the results.

After some initial buzz from being featured on New and Noteworthy, the downloads quickly dropped to less than 20 a day and InstaBAM tumbled out of the Top 1500 in the Photography section. We let the app rot in the depths of the App Store and moved onto other projects as working life took hold.

Several months later we decided to put a lot more time into InstaBAM and add the features needed to improve it.

After months of hard work we updated the app and foolishly thought the extra effort would get InstaBAM into the Top 50 in the Photography section. It never made it, instead it lingered around 300th place, gradually falling down the charts.

Although the daily downloads went up to around 100-150 a day, it was nothing like the 4 digit downloads a day we were expecting.

To make things worse we lost our search result advantage, there were now over 50 apps that came up when you searched for ‘Instagram’, with several that did the same thing as our app.

I tried to get the ball rolling with marketing and emailed dozens of tech blogs, Instagram blogs and photographer blogs asking if they would cover the app.

Not one responded.

We had no money to spend on banners or any sort of promotion. I used my Twitter account to try and spread the word and I changed the URL on my Instagram profile to point directly to the App Store but it was no use.

InstaBAM looked destined to slide to the bottom of the App Store yet again.

The idea

Around this time I was pretty hooked on Instagram itself and used the app daily whilst trying to think of new ways to use API.

I was spending a huge amount of time going through pictures and reading comments trying to get a better understanding of user behaviour and motivations.

It was pretty apparent that a large majority of users were desperate for likes and follows and would reciprocate any given to them.

If you looked at any photo on the Popular page at the time, you would see hundreds upon hundreds of comments that said stuff like this (Click to view several examples):

desperate

This quickly developed into a huge spam problem for Instagram as users cut and pasted the same comment across thousands of photos. Instagram disabled duplicate comments from users but as you can see above, they simply replace characters with numbers and Emoji.

Why were so many users desperate for more followers and likes? It turns out the motivation was down to the Popular page.

Many (mostly younger) users were trying their hardest to gain more followers with the goal of getting to the popular page in the hope it would provide more attention or some sort of fame.

There are even groups of Instagram users who gather at certain times of the day to like each others pictures at the same time in the effort to make them Popular. It’s insane.

I asked one user why:

despo

I became more and more fascinated by this ‘follow for follow, like for like’ behaviour as well as the popular page and really wanted to pick it all apart. It felt like there was something in there.

But how do photos get onto this mysterious page? What happens once you’re on there? How many people need to like your photo? How many followers do you need to have? What’s the algorithm behind it?

Then I wondered if there was a way to use this trend to our advantage. If so many users were reciprocating follows and likes, it could drive thousands to my profile and the App Store link listed on it. (see image of my profile below).

my-instagram-profile

If I could somehow drive people to my profile they might tap on the link to download InstaBAM

Figuring out Instagram’s like and follow reciprocation

After seeing all these users so desperate to be followed, I decided to start following all the people that left spammy comments and liking a few of their photos. They almost always followed me back and also reciprocate likes on my photos. A great start!

I started to branch out and go beyond the beggars and following tons of random people.  By simply hammering the ‘follow’ buttons that Instagram makes available on list views (below) I was able to quickly follow 30-40 people in a couple of minutes.

It went like this: Go to random persons profile, go to followers, then just keep tapping down the list while I was bored on bus journeys to work.

follow-list

My follower count started to rise, for every 100 people I followed, one or two followed back.

I also noticed that if I took the extra time to Like a couple of photos as well as follow the user, it would increase the follow-back rate by 10-20%. I was onto something!

One of the reasons why bulk following and liking gets you so many back in return is because by default, Instagram sends a notification every time someone interacts with you.

And what’s most peoples first reaction to seeing this? They check out the profile of whoever it was (and potentially my App Store link).

photo

 

I gained several hundred followers but the process was really time-consuming, plus my index finger was about to fall off. I would never get the volume I needed by doing this manually.

Then I had a thought – what if we could make this process automated? A simple script could not only follow people automatically, but also Like their photos too, increasing reciprocation and traffic to my profile.

We needed a bot! It seemed like a perfect way to hack our own marketing.

Building a bot!

I began to write some basic logic for the bot based on what I had been doing manually and sent them to Brendan, who reluctantly agreed to code it up after I convinced him this wasn’t evil, but simply a new marketing trend called ‘Hacker Marketing’ (I completely made that up).

I defined some rules around follow and liking :

  • A follow is worth more than a like. However if a follow is combined with a like, then it’s seen as a bigger gesture and the value increases. The bot needs to do both on every user.
  • Don’t let the bot drift into creepsville – ever had someone like 10 of your photos in a row, a bit creepy right? The sweet spot for amount of photos liked is 2-4 per user (not 3-7 as the graph says)

Here are the actual Powerpoint slides I sent to Brendan which laid out the flow and logic:

Screen Shot 2013-12-03 at 23.58.20

Screen Shot 2013-12-02 at 01.56.05

This covers a couple of things:

1) It specifically follows people who have Liked photos on the Popular page

Why? Because those users are more likely to be active, especially within a recent timeframe (Popular page photos are only popular for 30-60 mins). Also if someone has just liked a picture on the Popular page then there’s a chance they still have Instagram open and will see my interaction instantly.

This means if the bot is running when I post a new photo, the influx of active, reciprocating users could push my photo to the popular page.

2) The ability to follow and like people who post to hashtags

My theory was that if the bot could target people who post their photos to a hashtag, then you could potential get reciprocation from users in a specific location (eg #brazil, #rio #igersbrazil etc).

If they tapped the App Store link on my profile and downloaded InstaBAM, it could increase our ranking in those countries.

3) The bot randomises the amount of likes it gives peoples photos and how often it runs

I figured it would be best if the bot did things randomly and took breaks as not to hammer the API and cause a red flag. We decided that I should run the script in the background via IDLE on my laptop while I worked so I could keep an eye on it.

Within 48hrs, the bot was born! We named it Follow Pie after the delicious pastry and also the language it was coded in (Python).

Running the bot from my own account rather than a branded one

We spoke about creating a branded ‘InstaBAM’ account and running the bot from there. But I decided that it would be better to use my personal account as it would come across as more authentic.

At the time there were tons of spam bots on Instagram with hot girls in the profile picture, 1 photo and some blatantly dodgy URL. You can spot a spammy account from a mile away.

Some people would have probably blocked or reported an ‘InstaBAM’ account liking all their photos, but because it was me and my holiday pics, it looked legit.

I think this approached really helped, the combination of the bot and my own Instagram activity helped me to start building engagement around my photos and InstaBAM.

The risk of running the bot from your own account is of course being banned and wiped off Instagram (along with all your photos).

Drip feeding photos

I needed to post photos regularly to get the most out of all the new followers the bot was attracting and try and get to the popular page. The trouble is that I rarely take good photos on my phone (or ever) and quickly ran out of things to post.

I ended up digging out tons of holiday photos (mostly sunsets/beaches) I had taken over the years on my digital camera and started to drip feed them onto my account to keep the traffic flowing to my profile. I tried to post one or two photos every other day.

The results

I ran the bot from Mid-January to September 2012, as you can see from the graph below (sadly Statisgram cuts the graph off before February).

I started off cautiously with the bot. After no sign of being banned, I ramped it up to a few hundred follows and likes each hour for several hours everyday. Between March and April I followed around 20,000 users.

The growth that followed was pretty amazing, the combination of Liking and following people was generating hundreds of reciprocated follows and likes in return. Even the people who didn’t follow back would still like a few of my photos, which was fine by me.

I went from less than 500 followers to over 10,000 in the space of a few months. At my peak, I had 23,000 followers.

When we first created the bot, the API didn’t have too many restrictions, but gradually Instagram began introducing limits to API. The one that really affected the bot is capping every user to only be allowed to follow 7500 users. You can see the green line go flat at this point.

instagram-growth

This is when everything slowed down, although the bot could still like photos, it was no where near as effective when combined with a follow.

On the plus side, I did finish with 23,000 followers who I could now engage with.

Let’s look at App Store downloads. (I was doing stuff manually from Nov to mid-Jan).

Whilst running the bot we were getting around 50-100 extra downloads a day on average, which doesn’t seem much but starts to add up over a few months.

The spike you see is when Apple featured InstaBAM for a week under a section called ‘For Instagram Fans’. I’m not sure if Apple spotted us due to the increased downloads we were getting, but it made a big impact.

It’s a bit annoying as it pollutes the graph, but you can see the correlation between the both charts at the end of March when the bot followed 20,000 users. The downloads more than double to around 500 a day through that period.

chart2

Promoting InstaBAM in my feed

I was quickly building up followers so I started to frequently post photos about InstaBAM and encouraged people to download it via the link on my profile.

It worked great, I was getting hundreds of likes and seeing lots of feedback from people who ended up downloading it.

I was able to start building a little community around users who loved the app.

Screen Shot 2013-12-08 at 17.17.33

Screen Shot 2013-12-08 at 17.02.13

I began to see small increases in downloads when I posted about InstaBAM too, here’s the graph on the day I uploaded the photo above:

Screen Shot 2013-12-08 at 17.13.22

Popular page

Now that I was getting tons of engagement on my profile from thousands of new followers – would posting my generic sunset and beach photos get to the popular page?

The answer is yes!

IMG_3682

It happened again..

IMG_4037and again…

IMG_3653

In fact, my photos have now appeared on the Popular page over 40 times (a selection shown in the image below)

Screen Shot 2013-12-02 at 03.02.10

I began getting comments from people who were liking the photo that was on the popular page, then visiting my profile because of my photos and now downloading the app! It was working!

IMG_3606

App downloads during and after reaching the popular page

Let’s look at a few examples of popular photos and the downloads for that period.

Here’s one that reached the Popular page on the 5th May 2012 and gained 5271 Likes:

Screen Shot 2013-12-06 at 00.47.13
Here’s the App Store downloads for InstaBAM during the same period. You can see downloads increase when the photo hits the popular page:

Screen Shot 2013-12-08 at 17.57.29

Another benefit of being on the Popular page is that even though it only lasts 30-60 minutes, friends of the user who liked the photo will also see it appear in their ‘news’ section, driving more traffic to the photo (and my profile). That can last for several days as more and more people check their Instagram feeds in different timezones.

Here’s one taken on the 17th June 2012 with 4694:

Screen Shot 2013-12-08 at 18.26.08

Here’s the download chart. There’s a small but welcome bump:

Screen Shot 2013-12-08 at 18.26.02

 

Here’s another on taken on the 26th March 2012 with 2677 likes:

Screen Shot 2013-12-09 at 23.36.13

Here are the downloads for that day:

Screen Shot 2013-12-09 at 23.36.07

Popular page algorithm

I was getting to the Popular page so consistently I began to notice a pattern.

I took screen grabs to document stuff –  once of the photo when it had reached the Popular page to show how many likes it had + timestamp, then another of my profile to show follower count.

It seemed that if my photos could get around 2% of my followers within 22 minutes, it would hit the Popular page.

Expand the image below to see a few examples:

popular

Once Instagram introduced the follow limit, I could no longer reach the popular page so they may now take that into consideration.

The Power of running the bot on hashtags

Here’s one part of the bot that blew me away – the results from targeting hashtags.

The theory was that if Follow Pie run on users and photos that had posted to hashtags for certain countries, we could force ourselves up the rankings for that country.

I ran the bot on everything to do with Brazil #rio #brazil #igersbrazil #igerssaopaulo and other countries like Turkey and Sweden.

Using Majicrank I could actually see InstaBAM rise up the App Store rankings as I was running the bot. It was nuts!

real-tiime

I tried running the bot on USA and UK hashtags but it just simply didn’t have enough power as those App Store require far more downloads to move up spaces. Since the bot was limited by the API to a few hundred follow/likes per hour, it didn’t make much of a dent.

Could it work if you had a dozen bots? We never tried, but it would be interesting to find out.

Real-world implications of boosting your followers and likes

Other than App Store downloads, what were the other implications of running the bot?

Well my Klout score increased quite a bit (68) as Instagram made up 17% of the score. I didn’t pay any attention to this initially but then I read this:

Companies are interested in Klout’s data for two reasons.

  1. Customer support teams use Klout behind the scenes to learn who they’re speaking with. For example, an airline representative might route an influential caller to a manager based on his or her Klout score. A hotel might upgrade a powerful caller’s room for free.

With things like Klout becoming integrated into CRM tools, could the bot improve the level of service I get from companies in real-life?

If I’m getting upgrades, freebies and better treatment then the value of the bot goes waaay beyond app downloads and likes on photos.

With my newly improved score, Klout offered me a perk of 6 packets of Popchips (which I gladly accepted and devoured). Thanks social media!

The End

Once Instagram capped the amount of people you could follow to 7500, the bot became less effective. I tried to keep it going by just using likes, but the results weren’t great and eventually I got bored and moved onto other projects.

The bot was a fun experiment, it was great to get creative around how we marketed InstaBAM with no money.

Overall it helped us generate 1000’s of additional downloads over the course of several months as well as gain hundreds of advocates for our app. It was fun engaging with loads of new people every day and responding to comments.

InstaBAM has now had over 150,000 downloads and is still used frequently by thousands of users. Despite that it’s unlikely we’ll ever update again (if you fancy buying it let me know).

Could the bot be applied to other social networks? Possibly.

It seems like any social network with a ‘popular’ page will always have droves of users desperate to get on it. I recently noticed the exact same ‘follow for follow’ behaviour happening all over the Vine popular page:

Photo 09-12-2013 23 46 06

Grab the code on Github

We’ve released the code on Github for you to play around with.

Here’s a wishlist of things we would have love to have done:

1) Intelligent commenting AI

It would be interesting to see what would happen if you combined clever commenting algorithms that made it look like you were taking part in conversations on photos, similar to the @trackgirl Twitter bot. I’ve got the feeling that it would make a big impact.

2) Ability to automatically unfollow users

It always looks a bit lame when you’re followed by someone who is also following 100,000 people. If the bot could gradually unfollow people after a period of time then the followers/following ratio would look more impressive to users. This may also help get on the popular page again.

Disclaimer: It goes without saying that you use the bot at your own risk.

Say hello on Twitter