Inside Hive Media: Managing 10,000+ Social Media Accounts at Scale - Full Transcript

In this in-depth technical interview, Francis Roth sits down with Matt (CEO) and Chris (Technical Infrastructure Manager) from Hive Media to explore the systems behind their massive OnlyFans management operation. Managing over 10,000 social media accounts across 60+ models generating $750K-$1M monthly, Hive represents the extreme technical end of horizontal scaling.

Francis: Alright, welcome back Matt.

Matt: Thanks for having me on again, Francis.

Francis: Yeah, absolutely. So Matt, your first appearance on the podcast was one of, it might actually at this point be the most watched video on this channel, which to me is very strong indicator that the audience really liked what you had to say. So if for anybody that's listening that did not tune into that conversation, Matt is the founder and CEO of Hive Media, which is a top 100 agency that's pulling down, correct me if I'm wrong here, but between 750K and a mil a month. And you're doing it with the exact opposite of what I typically advocate for, which is you guys are pumping out insanely high volumes of more or less undifferentiated content with Latin American salary models.

So as you might expect from somebody whose business is entirely predicated on scale and consistency, Matt is a very sharp operator. I really enjoyed our first conversation. And so I invited him back to pick his brain about some of the more technical elements of his business. And to add some depth to that conversation, we also have Chris, who built and manages a lot of that technical infrastructure for Hive. So thank you for coming on, Chris. Appreciate you being here as well.

Chris: Hello Francis, I appreciate you having me on.

Account Creation at Scale

Francis: So I want to start from the very beginning, which is account setup. Obviously that is step one. So if you had to ballpark it, how many social media accounts are you guys currently managing across all models?

Matt: So across all models within our database, within each platform we maintain a minimum around 10,000 social media accounts. But just to give you like, you know, that's just like a number of eyes on a global but to give you an exact figure — when we start a model, every model gets issued 200 accounts on every major platform. So what does that mean? That's Instagram, that is Twitter, that is Reddit, Twitter, Reddit, Instagram — am I missing anything here Chris?

Chris: No, those would be our three main ones. We test a few other ones out here and there, but those are our main bed runners.

Matt: Threads would be the last one. And then there's about a hundred that become active immediately when they start, then a hundred for reserve depending on how the model does, the LTVs returns, or replacement accounts based on obviously there's a percentage of burn even if you're organic a hundred percent. There's always a risk of having some bans. We are in the OnlyFans space and we're always, regardless, you're always kind of bending those explicit guidelines even if you're doing everything above board.

Francis: Jesus, over 10,000 accounts, so roughly 200 accounts per platform per model. I would imagine given just the sheer number of reps that you guys have, that you probably have some pretty solid best practices for actually creating those accounts. So if you could just walk us through that process and then explain why you do it this way and maybe what some of the common pitfalls with account creation are that you see in the industry writ large.

The Evolution of Account Creation

Matt: I'll start by giving like kind of a summary on like the state or the history of account creation and even just where the industry has been. So, you've talked about this in a lot of your other videos over the last few years — industry's evolved, changed. I would say, and that's mostly on the social media front, right? Three years ago with a cell phone, no proxy, the same phone, you could bend rules, break rules and push a lot harder. Social media platforms with the growth of OnlyFans, the growth of spam are continuously changing, evolving and clamping down. So what you could do three years ago, four years ago, five years ago — years ago on just your local computer, your home phone, your browser, things don't apply anymore.

So I would say two to three years ago, most people were on web, whether it be Instagram, X, Reddit, creation was happening on web using whether they're Gmail accounts using US proxies, foreign proxies — a lot of these variables did not matter in the past. Nowadays, with the growth of OnlyFans, the attack on social media, and them really trying to control the spam, the way and the quality has — the bar has risen dramatically.

So for us, we pretty much do everything to the highest standard possible. We believe that the future of the industry is almost gonna be solely on phones. Why? This is how most users interact with the platform, so it's the most natural representation of a regular user. All of these websites use telemetry to identify user patterns — IP address, fingerprint, so like what device are you on? Is it a phone? What phone? Is the phone old? Is it jailbroken? Is it a computer? Right? Most people right away if you're using a computer on Instagram, that's not regular behavior of a user. So right away it's kind of like a negative strike.

High Trust Score Account Creation

Matt: So for us, it is the highest standard possible now. We moved to this about a year ago, so we are 100% cell phone creation. We try to use high authority email addresses, so Gmail's or iCloud accounts. Those again help increase everything you do in the positive column, increases the strength, and what we refer to as "trust score" of the account. We use only US proxies. During the creation process, we believe that the mobile proxies is the strongest proxy that will have the least amount of problems. We use residential proxies to actually run the account, but during creation you want to create these accounts as high authority as possible.

So yeah, that touches on like device — again, the newer the phone, the stronger the account is going to be. Mobile proxy and then making sure we're creating everything on mobile. This is kind of like the foundation of what we consider a high trust score account. And this is pretty much synonymous with any platform.

If you add additional factors to your accounts, which are, let's say like phone numbers, again, USA based or two FA verifications, this also will increase the strength of your account. So everything you do that is natural, that increases the security, that prevents it looking like a low effort account in the creation process will basically put you in a higher status starting out.

Francis: So, okay, all of that makes a ton of sense. Now, from like a practical, sort of logistical perspective, what you just said is essentially what matters most is the degree of trust that you're building, the more natural of a user you appear to be during the account creation process. But you guys are managing 10,000 plus accounts, so I can't imagine — do you just have VAs that are just sitting there all day, like manually inputting all of this, or do you have some mechanism by which you can automate it while making it still appear to look like a user? What does that process look like for you guys?

Matt: One of the only processes that we currently still have that is manual is the actual phone creation. And it involves resetting fingerprints and rotating mobile proxies. So it is pretty much one of the only processes that we still have that's almost completely manual, other than some of the system management we do. Certain specific platforms have uniqueness that requires manual intervention. But because it is so intricate, it's so precise, there's such room for error — we want it to be natural. We are still performing 100% hand created. So we have a full-time creation team that consists of five members of our team and all they are doing every day, every single day is creating Gmail's, Reddit, Instagram, X and then they go store it in our database.

Database Management and Organization

Francis: And then you guys are, in terms of managing that data, you're just saying this residential, or this proxy was associated with X account, Y account, Z account for X model, and then that's all just a database entry essentially.

Matt: Yeah, I can show you. This is a screenshot — we basically blanked out on the left the account names for obviously privacy reasons, but essentially what you see here is one part of a spreadsheet with accounts. This is like 2FA over here, there's a status of account, notes that may have happened on the account. And then I have another part here — this is part two. You have your proxy identifier here of all the accounts.

This would be the date of the creation. So this is like one model's account database essentially. Organization at scale is obviously super key. The backend of most of our systems are Excel sheets. And then we have front-end interfaces that interact better. But everything is backed into database forms.

Most of this all gets updated automatically via bots that we have or scripts that we run for these accounts. And this is kind of how we manage everything. So we have our own proxy database that we monitor what account has each proxy. We have also proxy changers. So let's say we have to change proxy at scale, we have tools for that. Over the last year, we've evolved — two years, we've developed and evolved as we scale. I would say a year ago, I remember us all sitting and changing 2000 proxies by hand over four hours.

Francis: Dear God.

Matt: Yeah, and now we have scripting for it where we can change 5,000 proxies in three minutes.

Automated Monitoring and Reporting

Francis: So how are you managing and tracking all of that? How do you even notice that an account gets banned if you have 10,000 accounts?

Matt: So from a monitoring standpoint on actual bans and monitoring, that all happens automatically. So every day we run a script because they're all social media platforms and all accounts technically are outward facing. So you can use public profile views to check the status of accounts very easily. So it's a matter of running a script across all of our accounts. It updates all our spreadsheets.

If they were banned, this would get automatically updated every single day and it would show in reports. This is like a random report — right now active on Reddit. This is actually today, and you can see total accounts running active. So we have 47 models running on Reddit with 4,000 accounts running, 3,900 are active and we got 92 that were banned. Then it actually goes over here showing the actual listing of the accounts that are banned. Now our supervisors gonna go in, identify, replace, figure out if there's a pattern to the banning. Is it a user mistake? Is it just algorithmic? You know, when you're running a scale like this, there's always gonna be a burn rate.

Posting Automation Systems

Francis: What is the actual process of posting on these accounts?

Matt: We have different categories for different platforms. The way I look at it is you have two or three categories of platforms. You have Instagram and X which are very linear in the actions that you can do, right? What I mean by that is you are either posting — whether you're posting a reel or you're posting a picture or carousel, it is posting mechanism to get traffic — or you're doing commenting to get traffic to get views. Those are really the only mechanisms there is to get traffic on Instagram and X.

Those mechanisms are very linear in the sense that it is... create post, attach a document, attach a piece of media, and then post. So they're very easy to automate. Instagram and X are completely on automation for us. So there's very little to no actual physical posting going on on the platforms.

We have multiple systems. We have a phone farm that is completely remotely operated and controlled. That is in Vietnam. We have an operator in Vietnam that we brought on board two years ago that this is what he was doing for the last seven years — Instagram phone farm management. He was actually selling t-shirts via drop shipping on phone farms. He was spamming reels. We found him selling some accounts. We bought some accounts from him. We recruited him and now three years later, we have a couple of 300 phones fully on automations out in Vietnam.

We essentially are running that completely from a device monitoring panel. That phone farm is completely remotely operated. Every model has assigned phones. Some girls have four phones, some models have five phones, this model has two phones. The higher return, the more money and the more resources we're gonna put into the model.

Francis: So that's all directly tied to performance.

Matt: Exactly. It's exactly that type of formula. The higher return, the more we're going to put into the model.

Reddit Strategy: The Human Element

Matt: Let's talk about more nuanced skill-based platforms like Reddit. Reddit is not linear, right? There is thousands and thousands of subreddits. Each subreddit has its own type of content that needs to get posted, right? There is different captioning for each type of subreddit, different rules for each subreddit, so it's not linear. It's much more complex to automate and quality is a lot more important on Reddit than algorithm because you're targeting specific content to a specific niche to specific user base.

Francis: With Reddit, you're serving the content to a very specific community intentionally. Whereas you don't have that optionality with Instagram, where it's just you're putting the content out and you're hoping that the algorithm serves it to the correct people. On Reddit, you are the algorithm, essentially.

Matt: Exactly. So we have a hundred VAs on Reddit that are manual. They all have physical devices in hand — whether they're in an in-house location or remote. Within the first four weeks of a VA starting, they have a phone in their hands. All of our Reddit operations are still 100% manual, they're 100% skill driven and it is a matter of matching niche content, niche captioning and targeting the right content for the models in the right area.

Francis: Just out of curiosity, because I think Reddit especially has become a little bit more contested — what you just described is you have 100 employees whose exclusive job it is to post on Reddit. You have these incredibly complex automation systems for every other social media platform. Is the return that you're getting from Reddit worth it?

Reddit ROI Analysis

Matt: So our cost per acquisition of a Reddit subscriber is anywhere from 25 cents to a dollar. That is our cost for a free subscriber. And we are getting anywhere from four to eight dollars return. So a VA doesn't have to do much to break even on at an acquisition cost of $1. So let's say a VA is $125 a week — they're using Filipino or Brazilian labor, which is good pay for them and their demographics.

Their return for us is if they get 20 subscribers a day, which — each VA runs about 20 accounts, 20 to 30 accounts. One free subscriber a day per account. So it's a very low threshold for the return. So we have some VAs who bring in just 100, 150 free subs a week because their skill level is low. We have other VAs who are bringing in 100 a day. And it's very unique to the VA because their level of effort, level of skill, level of creativity — the best VAs are out there actively looking at the results every day.

We get like a daily report from our Reddit team. Every day we get this report so we can identify the outliers and then every single day be working with them to improve their results. Continuous monitoring for anything staff related is the most important thing.

Content Production at Scale

Francis: In terms of actually posting content and scaling creative, how are you guys producing all of this content at scale?

Matt: From actually from a content front, it's honestly been one of the easiest things we've had to overcome. Because we started this almost three years ago now, we had our initial base model. We were really able to dial in our content approach with models. Understanding what it is to work with models that don't want to do a lot of work and having a standard.

It's understanding what are our models — our models are models only going to work four to six hours a week, right? So that is our — we built everything around our model base and then associated with our marketing strategies. So understanding that we only have four to six hours per model, we needed a really clean and crisp format that was easy for the models to do, but also have enough content to produce and feed the machine, which is Instagram, Reddit, and also obviously have some shit to sell on the page.

So we have what's called a "set format." Every week the models have to deliver two sets. A set consists of A, B, and C.

Set A involves SFW content which involves like taking some pictures of their daily life — getting ready, going to the gym. This is stuff that would just show realism, show that she's a real girl. She's a person. This is just casual daily content. It might go into a story, maybe it goes into an SFW reddit post. It's warming up an account. Maybe it's a profile picture. This goes to just general content to show that she's real. That's section A. This is about 20 to 30 pictures, all with variation and 10 to 15 short clips, 30 seconds.

Set B is sexy content. This is bikini, lingerie, maybe it's teasing content where they're covering their boobs. So this is either Instagram, sexy Instagram, or X-content that's compliant that still follows an SFW that's non-nude, or this makes also really good Reddit SFW content. So these are like teasers and borderline explicit that's showing things, that's being super sexy, super sexual without showing anything. So this is your classic OnlyFans bait bullshit.

Francis: Thirst traps.

Matt: Thirst traps, exactly. And then we have Set C, which is usually either a sexting set or a pay-per-view. This is the selling portion of it. So this is A, B, they see that on the socials, and then C, they come to the page and there's a matching sexting set or a pay-per-view of some kind.

Character Development and Branding

Francis: Do you guys have a team that's coming up with new content ideas, or do you just basically say: We have this box of stuff that works, here's an exact template, everyone go do exactly this?

Matt: As much as we do have a box, the first thing we do when a model signs is we sit them down and we create a story. Just like you would with any more quality approach — branding, etc. We create a pretty lengthy character sheet — what's her occupation, what's her personality, what's her persona. It's pretty in-depth. And then we do try to build the content — her storyline, her pay-per-views — we are always trying to gear it, her outfits, her scene, everything within her niche.

We do try and then we're more specific. We don't do as strong — we do decent storytelling, but what we do really well is really good niche. We do have a full creative team that puts a lot of attention and drive into these nuanced details. This also goes to the chatters.

It's very much dependent on the character. Like this character is obviously young, she's blonde, she's got that whole — her category of fans is going to be guys who like that type of girls or have like a white knight syndrome or like sugar daddies, right? These are the people that are going to have a predisposition to her. If it's a MILF, maybe she has a kid, maybe she's struggling to get by and has had a bad separation.

We're always trying to target people and demographics and spending habits based on that person's look, right? So there is some strong branding, so profiles, the way they dress, all that — the content doesn't necessarily always feed real to the character, but we can fix that with captioning, profiles and have it kind of work. So at least there's some tie into the character.

Performance Tracking and KPIs

Francis: You guys are distributing accounts to models based on performance. So how are you measuring that?

Matt: So our main indicators on a weekly basis that we document rigidly, especially during that first four weeks. We put the initial marketing campaign, which is two Reddit VAs, Instagram and X, and we are monitoring overall views that they're achieving — views to conversion. So views to clicks, clicks to conversion. This will tell us how well our model converts and is accepted on social medias.

But the main indicator for us is LTV, right? And again, it's not always linear because chatter skill, etc. can play variance, but we are always monitoring free LTV and paid LTV over a seven day and 30 day period.

We have free LTV, paid LTV, and then we have a weighted total LTV for the model. These are our key indicators. So seven day monitors our trends week to week. This is telling us based on 30 days — are we extracting well this week? Are we not extracting well? Has there been variance week to week? And then their 30 day gives us a more longer term pattern, removing some of the dependencies of luck.

We're always trying to get in that three, four dollar range. Obviously you have some that are very high, right? So if they're getting a very high return, both on their paid and their free, well this is a good model to pump money into. We're gonna get a higher return.

Francis: And just to clarify, anything below $4 is what makes you investigate that issue?

Matt: Exactly. It's purely based on cost per acquisition in our business model. So it's understanding what my cost per acquisition is, understanding what my overhead and chat costs are, and then understanding where our goals are. So if our goal is 40 to 50% margin, our $4 benchmark is to achieve a 50% margin, right? So it's always based on those metrics and then the assumption of a $1 cost per acquisition.

So our base fundamental is based on a 50% margin. So those initial Reddit, Instagram, and X, we aim for a 50% margin. If we're not hitting it and our cost per acquisition is high, it's probably because something needs to be tweaked on marketing. Maybe the VA's bad, maybe we're doing something wrong, maybe the sales are bad, maybe the accounts are bad. Maybe the model. But it alerts somebody to investigate that issue.

Software Integration and Tools

Francis: At the scale that you're operating at, there's obviously a lot of software involved. Do you just use an amalgamation of CRMs? You guys obviously have built in-house proprietary tools. What is kind of the playbook there?

Matt: I'll shout out a couple here. I'm gonna shout out to Inflo and shout out to SuperCreator. So people may say we're crazy running multiple CRMs at the cost they're at. But for us it just makes sense. We do have a lot of our own custom solutions with an API on the marketing front. But then obviously we're super integrated with Inflo, who's been a strong partner for us and then SuperCreator as well, who has some amazing AI technology, AI chatting, AI tooling, then their analytics are very good as well.

Both tools we leverage for both reporting, monitoring at scale — Inflo being our main tool for sales monitoring, employee sales monitoring, and then SuperCreator being more on creator analytics and then AI integrations into the chatting structure. And then I would say most of the other features are in-house mostly at this point because we've built our own monitoring panels, we've built our own monitoring system.

Understanding where your money is, analytics and how your business — monitoring it daily is key to running a good financial back end of your business. Understanding what's working and what's not working.

AI Integration for Chat Management

Matt: We are working on a tool where we can export these reports daily from Inflow's message board. Every single day we run a script — and this was Live Rich that gave us this idea — we pull reports and we can actually go in through with AI, in an AI LLM model that we were training and that we've trained through the last year. Based on two years worth of data of our own chats, we're able to pull out a very detailed report where it tells us exactly what happened during the shift. If they left people unread, how many interest signals — so that means buy signals were there — versus how many total offers they sent. It also gives them advice and examples on things and opportunities they've missed out on, how to improve them, KPIs of their shift versus other people.

This is like a full almost live monitoring system at scale using AI and chat history to give daily feedback in extreme detail to your chatter so that they can improve and grow. And also this gives visibility to management. You have 90 accounts hooked up to your Inflo. Even if you're five guys looking at the chats every single shift, you miss things. We have 40 chatters on, 30 chatters on. It's impossible to see everything. AI integration has been a huge tool for us.

Final Thoughts on Scaling

Francis: What I'd love to get you guys to touch on briefly is what are you working on improving? Are there any places that you wish that you had more insight or maybe that there were better tools for you?

Matt: I would say from a social media standpoint, there's tons and tons and tons of third party tools. There's tons of integrations because it's widely applicable to other markets. So whether you're marketing shampoo or whatever you're trying to market or a service or something local on social media, there's tons of analytics out there for social media, right? Because it's not unique to OnlyFans.

Where I believe that on the sales side, you do have the CRM, but there's not as many widely available solutions for analytics, monitoring, etc. Inflo does an amazing job, SuperCreator does a good job. I would say Only Monster does some good stuff too. They all do good things, but there is still lots of gaps to be held on the chat side. So yeah, this is something that the industry I know is improving, the CRM guys are improving, we're improving, but there's also the fact that we also know that AI chatting is coming and something that we are actively working with and working on.

Analytics and storytelling are really gonna be more key as well to AI optimization. So where is AI doing good? Where are they breaking? Where does it need to be trained to do better? This is a key fundamental and a key tooling and understanding this so that as the AI improves, develops, you're gonna be on the front end of it, using it the best way possible.

I think that just chatting overall is probably the weaker part. I think everybody struggles with this a little bit. I think you hear it most from most agencies — there's just less knowledge out there about it. And because it's a taboo industry, it's not always readily available to everybody. But I think it's the one that's evolving and changing the fastest with the revolution of AI.

Organizational Structure

Francis: Before we wrap up, I'm curious about your org structure. What does that look like with 300 staff?

Matt: We have 300 staff total across the business and 65 models currently signed. We have what you would consider an operator — people like me and you — at the head of every division. So I want essentially specialists like me and you who are obsessive, who are high level people, people who have that owner mentality and entrepreneurial spirit leading the division, obsessing, doing the research, looking at everything they need.

To do that, you need to pay people. This is something that doesn't get talked about enough in this industry — people are so hungry for the margins that they forget about what drives your business, which is people. And these high level people, people who are gonna think like me and you, they need to be compensated and paid. So most of these people are either partners, profit sharing partners in the business, or they have a very high salary range, first world, tier one salary ranges, like a traditional corporate job.

We have our upper management, this would be our executive team. Then each division has a manager, supervisor. Multiple supervisors depending on the size of the team — I would say one supervisor for every 10 to 15 staff. And then you have lead hands on top of that. I believe recruitment is the most important part of any business, period. So we do have a full-time HR manager that all she does is hire, recruit and maintains the people.

Hiring Philosophy

Matt: What do we look for in employees? I look for people who are looking for something long-term, who communicate really effectively. May not have all the skills but they know how to communicate and they're hungry. We can teach skills because at the end of the day this is a weird industry. It's not something — you're not gonna go out like a construction industry, like a professional engineering industry, like any profession and hire people who have long-term experience. So you essentially got to breed your own people.

I would say 90% of our staff are hired from outside the space. We train them raw. Good technology, good personality, good communication, responsible and want something long term. These are fundamentally the main things we look for. And then obviously you need to be somebody who's quick and can reason, troubleshoot.

We take a thousand resumes, we interview a hundred, we hire 10, we keep one. It's about a 1% to 0.7% hire rate because that is the level that you need to go through to find the right people. For any young guys out there looking — we are using Instagram ads, we are using meta ads, we're using local job boards. We are also using multiple labor pools. I don't believe in one labor pool over another. I believe that every country has its own predispositions.

Hiring is one of the main key things that we do. And we have multi-layer management systems and everybody essentially supporting everybody from labor to supervisors to managers. Everything flows from the top down basically.

Francis: This has been absolutely incredible. The raw volume of information that you guys just shared on this. I very much appreciate you guys' time. Thank you for coming on. And I'm sure the audience appreciates it as well. As you've repeatedly iterated throughout this whole thing, people are not very open about their internal systems. I think a lot of OFMs think they've built all of this, they're the only one that's doing this, they have to hide it all away.

When in reality, all of this is stuff that you guys constructed from first principles. You understood — I have a business need here, how can we build around this? And in reality, everybody should be doing that. So it's refreshing to talk to somebody else in the space who just, there's no gatekeeping whatsoever. So very much appreciate that, guys.

Matt: No problem. We run a community — Hive, honestly I haven't been super active there recently but we try to put information out and knowledge that's gonna help young entrepreneurs. We can't change the entire industry but we can try to help anybody we come into contact with. It doesn't hurt us to help so anybody who's looking to connect, looking to learn, please check the group. There's tons of good information from Francis as well. I share a lot of your videos to our team because there's just a lot of really good information and you got to be hungry. You got to be learning.

Feel free to check us out, feel free to reach out. We're always looking for people to work with, employees, people who might be looking to grow in the industry, anything and all opportunities we're open to and willing to help anybody along the line as well. Francis has been fantastic and if there's anything, feel free to reach out anytime.

Francis: Absolutely, man. And if for anybody that stuck around until the very end, please make sure to subscribe to the channel. Leave a comment thanking these guys for the sauce because this is just a ridiculous amount of information to share for absolutely free. And check the description for link to the Hive Telegram community, link to my Telegram community, early access to my upcoming OFM course, all my professional services, all that. Chris, it was great to meet you, and Matt, I'm sure that we will talk again sometime soon. You guys have a fantastic day.

Matt: Francis, take care.

Chris: Thank you guys.