Open source operationalisation: Why you need to think like a three-star restaurant chef.

I am going to make a bold statement: without operationalisation, analytical modelling is just an intellectual activity.

The process of operationalising a model is what allows us to gain insight from data, and unlock business value through data-driven decisioning. To get value, we need to move the model from where it was built, tested and validated, and put it into production, where it can access new data.

What does that have to do with open source? In 2019, Kaggle reported that 87% of AI developers depended on open source technologies. RedHat’s 2020 state of enterprise open source report showed that open source was important for 95% of the interviewees, and that 77% were planning to increase its use in the next 12 months. In 2021, there is no sign of the use of open source slowing down. Being able to get value effectively and consistently from open source technology through operationalisation is therefore a huge focus for businesses.

Defining open source

I am fairly confident that everyone reading this article will have heard of open source. However, we may not have a shared understanding of what it means. According to the Open Source Initiative, open source software is software released through a licence that makes its source code legally and easily available and distributable. This means people can access, alter and change the software to improve its use for the original function, or alter its use for something else in their own environment. This was originally seen as a way to speed up innovation, through crowdsourcing. Being able to take software from one environment to another is key here, linking back to operationalisation.

What are examples of open source analytical software? If I were a betting man, I would be that most people would think of Python and R. However, it goes beyond programming languages. Software likely to be used in businesses includes RapidMiner and Apache Spark. In other words, open source analytical software includes both languages and software whose source code is readily available.

Operationalising open source

It is difficult to explain everything that needs to be considered in operationalising open source. One way to think about it is to consider the difference between cooking in your kitchen at home and in a restaurant. Cooking in your home is like building a model on your laptop. It’s a small scale, individual activity. You know what everything is, how it’s used and where it’s stored. Take your oven, for example. You know if you need to turn it to 10 degrees above the suggested level because it doesn’t get as warm as it should, or that to get your gas hob to ignite, you need to turn the knob to its lowest level, not the highest. You don’t waste time wondering if the bread is kept in the fridge, bread bin or pantry, or where to find the bread knife. A friend trying to cook in your kitchen, however, could well end up with undercooked meat or having to cut bread with a butter knife!

The same would happen if someone tried to use your laptop to build a model. You probably have certain libraries downloaded but not others. You save data in a particular directory and output files in a certain folder and not that one. You can only imagine how much can go wrong even at this small scale!

I do not claim to be an expert cook. However, watch any TV show about restaurant cooking, from Masterchef to Ramsay’s Kitchen Nightmares, and you know that there is a lot more going on in a restaurant kitchen. Here’s Remy from the Disney film Ratatouille explaining who is involved.

The video gives us a brief glimpse into the number of different people in the restaurant kitchen, working on different things. Some have specialities: the sauce chef, or the pastry chef, for example. Others have responsibility for overseeing the work. No one person is solely responsible for every part of the meal, like you are when you cook at home.

We can think about operationalising open source as a bit like moving from your kitchen to a restaurant. We can summarise the similarities into four categories:

1. People

2. Ingredients and Data

3. Equipment and Technology

4. Processes

In my next article, we’ll explore these categories in more detail.

For those that would rather watch than read, you can view the on-demand webinar where I run through these ideas too.