Beyond the winning Formulae

Years ago, I did a program that was presented by Landmark Education. Aptly titled “Forward from the future”. The first part of this program was called “Beyond the winning formulae”

In the period where Covid-19 and Data Analytics are dominating our minds, I found the learnings from these programs very relevant.

What qualifies me to write this is the fact that I have worked in diverse industries and geographies and have repeated the process of learning and unlearning on many occasions.

Here are some of the common traits which one experiences when one enters an unknown domain

1. Someone will tell you this (business. geography, event ) is different, has never happened before and your experience does not matter.
2. The culture will matter and one cannot work around it
3. This is complicated. It requires an understanding of the domain and it really needs years of endurance in the same industry to get it right.
4. We have seen many like you, come and go and have their say with no significant impact

I have always started my learning process with these inputs. If I am setting up distribution in Mozambique or starting to build software for knowledge management, this is what I start with.

While I do concede that I have not been successful in all my initiatives, I have been successful in learning and unlearning. The reasons for failure or success could be attributed to simple business practices like or error in deciding on them. 

Some of the reasons I have failed

• Wanting to finish the product to a high quality before releasing and starting to sell the beta. 
• Unable to convince the management to fund the business path due to the overall market size
• Market size and lack of funds or approval to market.
• Timing, just plain timing has killed more than anything else.

Never ever because I did not figure it out, did not learn and unlearn.

The challenge with Machine Learning is similar.

Its dependence on the past is extraordinary. The past has limited data and what is available may be huge but tends to be the same.

So when someone says I have 10 years’ experience, is it the same year repeated 9 times or it is actually 10 years. While most humans can have 6 out of 10 years as experience, it is a diverse set of data for learning.

Data points include emotion, contradictive perspective, the concern of impact on others, the impact of joy during the process and Yes, the key outcomes which are measured, what we call results.

But quite a few of us and ML do not get that. So, we use data with 1-year experience repeated too many times. The model that will be generated out of this is going to be amazingly consistent and non-productive. 

I will let you grapple with it and not dilute your thinking by giving an example.

This is called winning formulae. If something has been working so far, it or its extension will work in the future. Any decision based on this premise is likely a failed standard predictive model. 

An action that pushes one towards a decision that has been predicted for you is what you notice if your ad blockers are disabled in your browser. This becomes intrusive and inaccurate on many occasions unless it a linear equation. I am looking for shoes and I am seeing ads for shoes. But I looked for shoes and hence I need socks is where the problem starts.

When human beings operate like this, it’s their winning formulae. This is the reason some grandpas have winning formulae on what works and you cannot see them working. You have unlearnt and Grandpa has not. 

All ML/AI models are like Grandpa and their reliance on the past is exceptionally high.

Now, what is the big deal here?

If you design a model which is going forward from the future, how would you design and implement it?

Unlearning is not doing away with your experience and you learning out of it. It is more of letting it not impact the way you absorb new data. 

So, when you hear of Covid-19 and its impact, your learning and experience will tell you that like storms, Tsunami, Cancer, plane crash and car-accidents it’s not going to get you. Which is likely to be true and I hope it is. 

On the other hand, we have seen the launch of MacDonalds or KFC or Kellogs or Starbucks, which is the learning that needs to be unlearned? Here the learning (winning formulae) was that persistence, branding, consistency, customer service, hygiene, price, experience, and taste will work for Indians as it has worked in Nigeria. 

As you know, only Starbucks got it right in the first go. They moved forward from the future. Rest took their time to learn and unlearn.

I will again not dilute it by giving the stories as evidence. I am assuming you can see the evolution as much as I did.

But what is forward from the future?

Let’s see how many read and like this article, then I will spend an hour writing about forward from the future