Finding influencer metrics - brainstorm ideas

rootjazz

rootjazz

Elite Member
Joined
Dec 21, 2012
Messages
3,166
Reaction score
1,190
Have seen a few threads about finding influencers recently and has got me thinking.

What formulae would be best for deciding if an account is an influencer.


Metrics to consider:
#followers
Probably need a minimum of (x) to be considered. And this number is used to find ratios with other metrics


Last 10 days images
#likes
#comments

Stories
#watchers of each and combined


I cannot think of anything else that might be useful, so need to use the above to get an "influencer score"

For each image, combine the #likes and #comments to get engagement score.

===============================

I am thinking this needs to be modified based on how old the image is. So we can weight the likes/comments based on how quick the engagement is.

Something like

engagement_score * (num_of_days_scraped - age_of_image_in_days)

i.e
100 likes and 100 comments for an image posted 4 hours ago would be
200 * (10-0) = 2000

an image that was posted 10 days ago and has 900 likes and 1000 comments would be

1900 * (10-10)

obviously we don't want x0 so if 0, call it 1 thus
1900 * (1) = 1900

so an image posted in the last 4 hours with 200 engagements is considered slightly better than an image posted 10 days ago.

===============================

Now that modified engagement score, should be modified based on the number of followers, otherwise the accounts with most followers will do better (in theory).

follower_engagement = date_engagement_ratio / num_followers

We do this for each post, for a total_follower_engagement score then divide by number of posts for a date and follower weighted engagement score.


The higher the score, the more influence they can be assumed to have.

Anyone with better mathematical knowledge have any input? Can we use any other calculations which would be better at weighting the results of multiple inputs?
 
Advertise on BHW
rootjazz said:
Have seen a few threads about finding influencers recently and has got me thinking.

What formulae would be best for deciding if an account is an influencer.


Metrics to consider:
#followers
Probably need a minimum of (x) to be considered. And this number is used to find ratios with other metrics


Last 10 days images
#likes
#comments

Stories
#watchers of each and combined


I cannot think of anything else that might be useful, so need to use the above to get an "influencer score"

For each image, combine the #likes and #comments to get engagement score.

===============================

I am thinking this needs to be modified based on how old the image is. So we can weight the likes/comments based on how quick the engagement is.

Something like

engagement_score * (num_of_days_scraped - age_of_image_in_days)

i.e
100 likes and 100 comments for an image posted 4 hours ago would be
200 * (10-0) = 2000

an image that was posted 10 days ago and has 900 likes and 1000 comments would be

1900 * (10-10)

obviously we don't want x0 so if 0, call it 1 thus
1900 * (1) = 1900

so an image posted in the last 4 hours with 200 engagements is considered slightly better than an image posted 10 days ago.

===============================

Now that modified engagement score, should be modified based on the number of followers, otherwise the accounts with most followers will do better (in theory).

follower_engagement = date_engagement_ratio / num_followers

We do this for each post, for a total_follower_engagement score then divide by number of posts for a date and follower weighted engagement score.


The higher the score, the more influence they can be assumed to have.

Anyone with better mathematical knowledge have any input? Can we use any other calculations which would be better at weighting the results of multiple inputs?
Click to expand...
I deal with influencers all the time. Tired of saying over and over again. If it's for your bot, DM me I can help you, and it's quite simple.
 
I'd like to know the formula for this if you got it all figured out
 
colorme said:
I deal with influencers all the time. Tired of saying over and over again. If it's for your bot, DM me I can help you, and it's quite simple.
Click to expand...
Would you be kind to share that knowledge please?
I'm curious to learn that.
 
How is any of that info relevant when most influencers fake everything nowadays?
 
Intexam said:
Would you be kind to share that knowledge please?
I'm curious to learn that.
Click to expand...

It's inside information. I don't want to divulge it out in the open so that those that abuse it, can just hide it a bit better.

EnhancedHuman said:
How is any of that info relevant when most influencers fake everything nowadays?
Click to expand...

Most influencers do not fake it.

MisterF said:
Most but not all, I would suggest.
Click to expand...

I would say that most are good, there are some that do abuse it for all others.
 
Deliberate said:
I'd like to know the formula for this if you got it all figured out
Click to expand...

That is what this thread is about, so we can brainstorm and hopefully some math genius can point to some equations that will just that :)

EnhancedHuman said:
How is any of that info relevant when most influencers fake everything nowadays?
Click to expand...
True, fake metrics would be hard to spot in an automated way, but as others have said, I don't think everyone is faking everything. Either you would need to manually review the output if it is critical it is all authentic, but if you are getting an 80% hit rate, that is good enough to be worth it IMO
 
rootjazz said:
That is what this thread is about, so we can brainstorm and hopefully some math genius can point to some equations that will just that :)
Click to expand...

Well I mean I have a formula in mind already. Here's what I would suggest but I was hoping for someone to suggest something else.

This is the engagement formula for a single photo - this would require some code to automate itself on an hourly basis. If you want to set a limit for how long the influencer score is calculated, we can stop it after 24 hours from being posted.

Get 24 hours worth of data doing this:
(# of likes + # of comments)/(number of hours since posted)

It would look something like [((200 likes + 10 comments)/(1 hour))] + ((94 likes + 2 comments)/2 hours)) ... until you have 24 hours of data or whatever you find influence not significant anymore)

In my example at the 2nd hour, the photo has 294 likes and 12 comments. You only care about how many new likes and comments, so you need to make sure you get the difference.

AFTER you are done, you can divide the total sum above by the number of total likes and comments. You get a weighted influencer score of about 84.3/100.
The more time it takes for a user to get likes and comments, the less weight it has in the score and drags down the influencer score.

--


You can also try to create something that attempts to find bots. If there's a bot or private account among those who are liked, you can choose to weigh those likes/comments as 0 or even a smaller % (1 like from a bot is worth 0.5 likes from a real person). This will also weigh down the influencer score.
 
If you scrape all the users that follow an account and all the users that liked the 10 last posts and after comparing the lists you find out that those that liked the posts are not followers, the influencer used SMM panels to boost his account.
 
Mr. Burnham said:
If you scrape all the users that follow an account and all the users that liked the 10 last posts and after comparing the lists you find out that those that liked the posts are not followers, the influencer used SMM panels to boost his account.
Click to expand...
Yes, and no. Many follow hashtags, and for certain hashtags, some accounts, in some cases, may be influencers that show up at the top. Thus, these people that follow hashtags, will like and comment on the same accounts for days, yet not follow them.
You first need to recognize who is an influencer. Once you do that, you can delve further as to if they fake their engagement or not.
 
rootjazz said:
Have seen a few threads about finding influencers recently and has got me thinking.

What formulae would be best for deciding if an account is an influencer.


Metrics to consider:
#followers
Probably need a minimum of (x) to be considered. And this number is used to find ratios with other metrics


Last 10 days images
#likes
#comments

Stories
#watchers of each and combined


I cannot think of anything else that might be useful, so need to use the above to get an "influencer score"

For each image, combine the #likes and #comments to get engagement score.

===============================

I am thinking this needs to be modified based on how old the image is. So we can weight the likes/comments based on how quick the engagement is.

Something like

engagement_score * (num_of_days_scraped - age_of_image_in_days)

i.e
100 likes and 100 comments for an image posted 4 hours ago would be
200 * (10-0) = 2000

an image that was posted 10 days ago and has 900 likes and 1000 comments would be

1900 * (10-10)

obviously we don't want x0 so if 0, call it 1 thus
1900 * (1) = 1900

so an image posted in the last 4 hours with 200 engagements is considered slightly better than an image posted 10 days ago.

===============================

Now that modified engagement score, should be modified based on the number of followers, otherwise the accounts with most followers will do better (in theory).

follower_engagement = date_engagement_ratio / num_followers

We do this for each post, for a total_follower_engagement score then divide by number of posts for a date and follower weighted engagement score.


The higher the score, the more influence they can be assumed to have.

Anyone with better mathematical knowledge have any input? Can we use any other calculations which would be better at weighting the results of multiple inputs?
Click to expand...
Well, depends, I´d count me with 450k as an Influencer. But when talking to a 5m+ page they´d probably laugh, so it´s always from the perspective you´re looking onto.
 
Deliberate said:
Well I mean I have a formula in mind already. Here's what I would suggest but I was hoping for someone to suggest something else.

This is the engagement formula for a single photo - this would require some code to automate itself on an hourly basis. If you want to set a limit for how long the influencer score is calculated, we can stop it after 24 hours from being posted.

Get 24 hours worth of data doing this:
(# of likes + # of comments)/(number of hours since posted)

It would look something like [((200 likes + 10 comments)/(1 hour))] + ((94 likes + 2 comments)/2 hours)) ... until you have 24 hours of data or whatever you find influence not significant anymore)

In my example at the 2nd hour, the photo has 294 likes and 12 comments. You only care about how many new likes and comments, so you need to make sure you get the difference.

AFTER you are done, you can divide the total sum above by the number of total likes and comments. You get a weighted influencer score of about 84.3/100.
The more time it takes for a user to get likes and comments, the less weight it has in the score and drags down the influencer score.

--


You can also try to create something that attempts to find bots. If there's a bot or private account among those who are liked, you can choose to weigh those likes/comments as 0 or even a smaller % (1 like from a bot is worth 0.5 likes from a real person). This will also weigh down the influencer score.
Click to expand...

The problem with this approach would be the number of requests potentially required. If you scraped say 1000 results and want to find any influencers in that resultset, to check the 10 most recent images evvery hour would be:

1000 * 10 * 24 = 240,000

for a 24 hour period.

Also I am thinking that the test isn't exactly fair as an image posted an hour ago and watched for 24 hours will have a different pattern than an image posted 3 days and watched.

Although, this may be a good method for watching a finalised list. So you watch an account for a few days, only watching new posts as detected to keep the test fair


Mr. Burnham said:
If you scrape all the users that follow an account and all the users that liked the 10 last posts and after comparing the lists you find out that those that liked the posts are not followers, the influencer used SMM panels to boost his account.
Click to expand...

Good point, you would probably expect (x)% of likers / commenters to be followers. Again the problem here is requests, if you have accounts with large follower bases, you need to scrape all followers to see if likers are followers. If trying to scrape 400k followers for 100+ profiles, it starts to add up again


red333 said:
Well, depends, I´d count me with 450k as an Influencer. But when talking to a 5m+ page they´d probably laugh, so it´s always from the perspective you´re looking onto.
Click to expand...

#followers is the initial filter. Ignore everyone with less than 200,000 followers perhaps.
But then trying to work out if those are engaging followers, is the account actual able to "influence" those followers. Make sure to ignore accounts with large follower bases but are basically useless.



But perhaps this is all over thinking, just need to ignore all accounts with less than X00k followers is good enough for most? <shrug>
 
rootjazz said:
The problem with this approach would be the number of requests potentially required. If you scraped say 1000 results and want to find any influencers in that resultset, to check the 10 most recent images evvery hour would be:

1000 * 10 * 24 = 240,000

for a 24 hour period.

Also I am thinking that the test isn't exactly fair as an image posted an hour ago and watched for 24 hours will have a different pattern than an image posted 3 days and watched.

Although, this may be a good method for watching a finalised list. So you watch an account for a few days, only watching new posts as detected to keep the test fair
Click to expand...
Yeah, this would only more so for new images rather than older ones that you can watch.
The only recommendation I can suggest if the number of requests is an issue is to check less frequently, but the data might become less accurate.

It's harder to determine an accurate influence score on older ones because you'd be missing the relevant data of when their followers (or non-followers) interacted with the photo.
 
red333 said:
Well, depends, I´d count me with 450k as an Influencer. But when talking to a 5m+ page they´d probably laugh, so it´s always from the perspective you´re looking onto.
Click to expand...
An X number of followers, does not make an account an influencer. There are 5K influencers, 15K, 50K, 100K, etc., real influencers. Brands have already studied and come up with certain filters to see who is an influencer. Some 100K influencers are paid more than some 500K influencers. It has to do with quality and engagement.

For example, a 1M account with 30K likes and 70 comments per post, is NOT an influencer. The account may put billboards on their account saying "I'm an influencer" "hire me", etc., but the filters used by brands will flag this account as a fake, and rightly so.
 
colorme said:
An X number of followers, does not make an account an influencer. There are 5K influencers, 15K, 50K, 100K, etc., real influencers. Brands have already studied and come up with certain filters to see who is an influencer. Some 100K influencers are paid more than some 500K influencers. It has to do with quality and engagement.

For example, a 1M account with 30K likes and 70 comments per post, is NOT an influencer. The account may put billboards on their account saying "I'm an influencer" "hire me", etc., but the filters used by brands will flag this account as a fake, and rightly so.
Click to expand...
So my current main account has ~350k followers, 30-50k Likes per picture and an average comment count of 300-600.
Does that qualify me as an influencer?
 
red333 said:
So my current main account has ~350k followers, 30-50k Likes per picture and an average comment count of 300-600.
Does that qualify me as an influencer?
Click to expand...
If you have original images/videos, organic growth and engagement, and passes a few minor tests, then yes. If the account just reposts others' content, then no.

And by organic growth, it can be a botted account, as long as it's done in a human way.
 
colorme said:
If you have original images/videos, organic growth and engagement, and passes a few minor tests, then yes. If the account just reposts others' content, then no.

And by organic growth, it can be a botted account, as long as it's done in a human way.
Click to expand...
Original images / videos: either I’m taking the ideas of someone’s jokes and post them with my design, logo and background or my follows suggest them to me.

I’m not botting and getting bout 400-800 followers daily. (I’m often trending the last few weeks)
 
Advertise on BHW
You must log in or register to reply here.
Sign up now!

Main Menu

Home Main Forum List Black Hat SEO White Hat SEO BHW Newbie Guide Blogging Black Hat Tools Social Media Downloads

Marketplace

Account Selling Content / Copywriting EBooks / Methods Hosting Misc Proxies For Sale SEO - Link building SEO - Packages Social Media Social Media - Panels Web Design More

Making Money

Affiliate Programs Hire a Freelancer IM Journeys Making Money Pay Per Click Site Flipping Want to Buy

BlackHat World

Dispute Resolution Forum Suggestions & Feedback Introductions Lounge
Back
Top