What is the opposite of Fragile (In Digital Marketing)?

We are intentionally being loud at Seeda lately, as my colleague JD O’Hea pointed out in his latest post, we want to build in public so, here it goes!

Antifragile

In his book Antifragile, the erudite author, Nassim Nicholas Taleb, answers this exact question in detail, with a pinch of spicy and delightful humor. I like him because he is a non-conformist, the type of mind that breaks with the status quo and brings new ideas forward with courage. Even when that means that he has to train to look like a bodyguard because he does not want to pay for one.

In a nutshell, he uncovers the fact that we do not have a word for the opposite of Fragile, most of us would think of Robust as the opposite of Fragile, but it is not quite. The opposite of a Fragile system or body should possess the exact opposite set of characteristics. If we vaguely define a fragile system as one that is highly sensible to stressors, that breaks under stress, the opposite of that would be some system or body that gains from stress (or change or volatility), he calls such a system Antifragile. A robust system is resistant to change but does not get better because of it.

This small change of perspective (a small change that requires someone like Taleb to have studied it and distilled it into a sentence everyone can understand) can be applied to everything in life that is a system (a set of individual components that interact with one another). And that includes businesses and even disciplines.

All of this got me thinking about the work that we have been doing at Seeda during the past months, specifically in the world of Digital Marketing.

The metric that defines our greatest value add is MER (Marketing Efficiency Ratio) which is the ratio of the the amount of revenue over a period to the amount of marketing dollars spent over the same period. We can provide such a metric because we seamlessly integrate all of the data from our customers (Shopify, Facebook, Google …). But enough with the self-promotion, what does this have to do with the Antifragile nonsense you were bluffing above?

The Fragile way

A lot of businesses rely on this specific metric to create a system to maximize ROI, and such systems can end up being Fragile or Antifragile. Let me illustrate.

Most of the industry, to this day, would build the system as follows:

1. To increase MER, I can increase revenue OR decrease marketing spend.
2. If I decrease marketing spend, that will also affect my revenue (decrease), so I have to do that smartly.
3. To adjust my marketing spend, I look at how much, attributed revenue there is to Facebook and Google and adjust accordingly.

That might work, if attribution data was 100% accurate, if there was no dependency between your spending in one channel and the attributed revenue of the other, and if all your revenue came from new customers. Too many assumptions must be made, and the system works as long as the assumptions hold, which is not the case often. This system is Fragile to small stressors. If attribution does not work, the system does not work. If your spending on Facebook affects the revenue from Google, it does not work. If there is a lag between the time you invest in one of the channels and when you see an ROI in that investment, the system does not work.

This relies on flawed data or metrics and does not learn from past mistakes. It is Fragile.

Now, what is the antifragile version of this?

Well, it is good old trial and error:

1. I increase/decrease spending in this and that channel this or that way this week.
2. I see if it worked, if so I iterate following that same route, if not I change paths.
3. And so on until I get to what I consider good, which will keep changing over time.

There is only one catch, these are a very expensive set of experiments to run, and we might as well have some assurance that they will work more often than not.

This is exactly what we do at Seeda (well, one of the many things). We aid our customers in finding that mixture of spend that maximizes their ROI. Follow Simona Vychytilova for a more detailed outlook on how we do this.

But in a nutshell, instead of running these expensive experiments, we look at all the past spending per channel and revenue data per week, and we run a set of ML algorithms to figure out the most optimal mix of marketing spend (known as Marketing Mix Modeling, MMM) that will contribute to increased revenue, hence MER, hence ROI. Moreover, we take into account lag effects (how your spending two weeks ago might affect your revenue this week) as well as saturation effects (when for a given channel, spending and extra dollars will not increase your ROI), which are embedded in the system.

Note that we are not trying to predict future revenue, we are just analyzing the properties of each channel (lag effect, saturation, and interdependence between channels) and drawing causal relationships between your marketing spend per channel and your ROI.

Such a system does not rely on any assumptions to hold and it is not static, it adapts to the changes in its environment. It is not trying to control anything, it is just trying to gain information from its environment and constantly adapting to it.

We would love to know what your experience is in the chaotic world of marketing spend optimization, don’t be afraid to get in touch.