How calibration make people famous on TikTok and Instagram

I found that one of the most successful methods for going viral on Tiktok, Instagram Reels, and other short form content social media, is to learn the importance of calibration when optimizing solutions for complex systems. It might sound complicated, but in reality it is really easy and one of the most powerful tools the most talented creators use to grow and become famous.

Linear systems vs Complex systems

First, let's define two types of systems. Bear with me, this sounds complicated but the practical examples in the end will clear things up.

  1. Linear systems: A linear system means that the input and output are directly proportional. If you drive a car, and you are on the same road, the same gear and press gas the same amount, on multiple different occasions, the car will move roughly the same distance each time. This is a linear system, where the input (gas) and output (distance) are directly proportional. This system has a few parameters deciding the result, which makes it relatively easy to predict what the outcome will be for a given input.
  2. Complex systems: A complex system is a system where the input and output are not directly proportional. A complex system often doesn't have a set of known predefined rules, instead it is known for being chaotic (not in order), adaptive (changes over time), non-linear (input and output are not directly proportional) and self-orginizing (it can change its own rules). An example of a complex system is the same example as for a linear system, but instead of a car we can consider a boat in water. If you place a boat in an ocean, roughly at the same place, and press the same amount of gas, it will be really hard to predict beforehand where the boat will end up. The ocean is a very complex system, and things like: wind, temperature, underwater currents, and many other factors will affect the outcome on that specific day.

Two types of modelling

How can we model the outcome of a system? With modeling an outcome of a system, we mean that we are trying to create a methodology that consistently predict an outcome for that given system. In other terms; if we want to get famous on social media and go viral over and over, we want to create our own methodology that we can reuse on every video and get the same viral results.


There are two types of modelling that are relevant here:

  1. Theoretical modelling: Means we try to come up with theories for our system, and then create a model that takes those theories and parameters into account to predict the outcome. Theoretical modelling is useful when we want to understand the underlying principles of a system and how it works. This is a very common way to model linear systems, where we can predict the outcome based on a set of known parameters. Consider the car example above, if we use a theoretical model, we can start considering the slope of that road, the weight of the car, the type of engine, introduce physics formulas and so on. This model can then be used to input a set of parameters, and we can know roughly what outcome to expect.
  2. Calibration: A.k.a. iterative optimization / empirical modelling. For this method we don't expect to ever understand what principles are actually ruling the underlying system. That's not our purpose, but we can still get predictable outcomes. This method is specifically useful for complex systems, where it would be too difficult or even impossible to understand the system fully (since it is ever changing as well). By using calibration we simply make a guess and try it out very quickly. We then measure, make a theory of an action that could improve the outcome, then we perform that action, measure again, learn, repeat. Everytime we get closer to our goal, we continue, but every time we get further away, we change our action and re-evaluate. This is an iterative process, where instead of calculating everything beforehand and making a big bet upfront, we keep making small calculations as we go. Instead of trying to understand the system, we are trying to understand how to get the outcome we want.

Example: Calibrating the brightness of a TV

Have you ever calibrated brightness on your TV? Maybe you have experienced that your TV is a bit dark and you struggle to see? Then perhaps you went into the settings and changed the brightness. This is an example of where we use a calibration process on a regular basis, and I will now compare the difference between using the two types of modelling, to show why calebration is the right way to go.


Set the right brightness with theoretical modelling: Let's try the first method to get the brightness right for our TV. With theoretical modelling we want to understand the system. So we could for instance start by measuring the distance between the TV and the couch, the light that is coming into the room from the window, the angle the light is appearing from and the amount of light that are coming from our lamps. If we do these complex calculations we might be able to create a formula, which if we do the math correctly will give us the optimal brightness setting for a TV. This doesn't sound that easy, right? This is not the recommended solution, and the method is not very practical. This is because the laws of physics that affect the perceived brightness of your TV is a complex system.


Set the right brightness with calibration: Let's try to get the brightness right with the second model, calibration, instead. This method is more appropriate for a complex system. Let's say we decide to do the calibration when we have average light in the room, that should be a good starting time. Not the brightest part of the day but also not during night and complete darkness outside. We then go into the TV settings, and we make a first guess. "I think brightness should be 56 / 100". We then sit in the sofa, look at the TV, and evaluate if 56 was a good setting. Hmmm, maybe it is still a bit dark. So we crank it up to 65. Now it turned too bright all of the sudden. We reduce to 61, but we notice quickly that that was too far in the wrong direction. We go up to 62 - that's it. Perfect. We have now used a calibration model to optimize the result of a complex system.

How to go viral using this knowledge

Let's translate this knowledge to social media and how people use it to become famous. The theoretical modelling approach, would mean we create our own theories and models of what we think the algorithms and human physchology of the social media apps look like. So we make a model of what we think will make our own videos go viral. For instance:

(Note: These are just example, and not necessarily true)


We now made a model that we think is true, and we can try to create videos where we plan very carefully so that each video fulfills all these rules, because then we increase the chances of going viral. The key here is that we are trying to understand the system, and then think that if we do lots of work upfront, we can predict the outcome.


But TikTok and Instagram are not linear systems. They are in fact very very complex system. They change over time, because people change over time. They are not predictable, because people are not predictable. There are also millions and millions of parameters that decide if your video will be successful or not on that given day. So we can not guess the outcome beforehand.


Instead, we should use calibration. When doing calibration for TikTok or Instagram, we can still have theories of what we think might work. For instance, I think the best video length is 7 seconds. But we should re-evaluate over and over, and we should expect the system to change over time. Let's use calibration to go viral:

  1. I have a video idea: a DIY home improvment technique that I came up with and want to share
  2. Instead of spending lots of time editing, optimizing beforehand etc, I quickly make a video and post it
  3. Now I go to the statistics, and look at parameters such as engagement and retention.
  4. I then make a small hypothesis: Maybe the video was too long, can I skip the intro?
  5. I make a new video, where I skip the intro and post it very quickly again
  6. I go to the new video statistics, and see if the engagement and retention increased
  7. If it did, I consider if I can make it even shorter and get to the point faster, or if I should improve something else
  8. If it didn't, I consider going back to the orginal format, and changing something completely different, like the way I'm smiling in the video
  9. I can also decide that if I don't see any improvements after a few attempts, that I will come up with a new video idea and start over
  10. And so on, continuing this process over and over...

With calibration we still create smaller theories, and can learn over time that, yeah, for me 7 seconds is probably the optimal video length. But we don't set these rules in stone, we keep re-evaluating them over time, since the system will change over time. We also use this knowledge to make our future content better, but we also don't spend loads of time on optimizing beforehand when we make new video concepts. We keep making small bets, and we keep re-evaluating.


Essentially, we don't write down a predefined set of rules that we think apply for TikTok or Instagram - instead we train ourselves and our own video making skills.

Conclusion

Making well planned out videos can be a good strategy, especially for already large creators. But when you are starting out, calibrate, calibrate, calibrate. Rather do a 100 quick videos than 10 well planned out ones. Keep doing micro-improvements, calibration and re-evaluation. Because instead of trying to understand and predict the complex system, you can iterate to get the desired result, which means you essentially train yourself into becoming a viral video machine that produces viral content over and over.