The 80/20 Rule (also known as the Pareto principle, the law of the vital few, and the principle of factor scarcity) is a concept that I trust many of you will be familiar with. For those who haven’t heard of it, put simply the principle suggests that often in business (and more generally in life) 20% of the inputs produce >80% of the desired outputs. The implicit reversal of that, is that 80% of your input produces 20% of your output. The real numbers may be quite different e.g. 1% of inputs produce 75% of desired outputs, though the point remains the same that some activities are more valuable than others.
Applying the 80/20 mindset to the day-to-day of a business analyst, we could say that tasks such as debugging code, cleaning datasets, and waiting for an Excel model to update are the suboptimal 80% tasks that produce <20% of your returns. The 20% tasks that produce 80% of your results may include your original idea on the output that you’re looking for, the dataset/s that you decide to use, or that ‘gut feeling’ that you apply as your last filter to business decisions.
At Remi AI, this is a problem we had in mind when we started building the Artificial Intelligence (A.I) platform that would become Catalyst. We believe that your 20% tasks should be pushed to the limit, allowing you to generate new levels of insight and results for your business. Wouldn’t it be nice if your time could be focused on the 20% tasks and have the 80% tasks fully or at least partly handled by A.I?
A.I and BI
The application of A.I to the space of Business Intelligence (BI) processes is hugely exciting, and something that has been attempted by some of the larger players in the market. By introducing the data heavy world of business to the data hungry methods of A.I, we can produce valuable and actionable outputs and insights.
Fusing the two has remained further from the mainstream than you might expect, limited primarily by the technological knowledge required to action such an undertaking, and secondly by the price tag attached to the teams of data scientists and PhD mathematicians required to utilise said knowledge. As it stands, there are very few ‘plug and play’ tools available for the everyday analyst to apply A.I methods. Yet where there are tools to use, another facet of a slow uptake is the resistance to or lack of time to change on the part of the analyst, in terms of up skilling right now in A.I methods.
Just as Microsoft Excel literacy is an assumed for any role requiring even basic analysis, an understanding of A.I methods will likely be an advantage for the next couple of years, and assumed thereafter. So the question becomes “How do I learn the required methods to improve my A.I knowledge and prepare for the future of business analyst roles and BI tools?” For the committed, there are many online courses out there (Including Andrew Ng’s Machine Learning course on Coursera as I have written in the past), though for the analyst who has never learned Matlab, Python, or C++, you might be looking for more of a synoptic view or at least a simpler way to learn.
Catalyst: A.I in Action
At some point in your career, have you been slaving through an Excel model and thought “Surely I can do XYZ?” and then found through a quick Google search that you can easily reach your goal through a previously unknown Excel function? The idea behind Catalyst is that by using it you will firstly have the use of powerful A.I algorithms without the previously mentioned mathematical and coding ability, but you will also come to understand the situations where different A.I methods are best applied. When you upload your data and select which inputs you hope will generate the desired output, Catalyst runs your data through more than 15 cutting edge algorithms to see if your data has any form of predictability.
Looking to predict a stock price at the next time step? Catalyst might use an LSTM. Considering how you might predict your revenue for the next quarter for budgeting purposes? The platform might apply an old school probability approach. Want to know the impact of rainfall on productivity? A recurrent neural network might be best. Having an understanding of when to use which approach will become invaluable to an analyst, and for many this ‘learn for a purpose’ approach, rather than ‘learn for the sake of it’ is a step in the right direction towards an understanding of A.I methods in context.
As a founder of Remi AI, I can acknowledge that I am obviously not an objective reviewer of this service. It’s worth mentioning that, given we are still in start-up mode, I have a day job, in procurement at a publicly listed healthcare group. The genesis of Catalyst came about designing a tool for people like me, who have access to terabytes of data but are unsure of precisely what to do with it, or perhaps lack the skills required. Believe me when I say that focussing your time on the 20% tasks that produce 80% outputs is wondrously satisfying.
Catalyst, which moves ever closer towards release even as I type this, will be a ‘drag and drop’ prediction tool.
An analyst could:
1. Upload any type of data e.g. profit and loss for the previous few years
2. Connect in any relevant external data streams e.g. the weather or stock market
3. Select the variable that they wish to predict
4. Run Catalyst to assess whether there is any predictably in the data, which the analyst can then act upon
5. Act on any insights generated e.g. ‘sell that stock’ or ‘increase inventory for the winter months’
Once these methods do become mainstream (and probably even earlier), the question on many minds when A.I is raised will be: will this replace me one day? Who could blame you when the technology being regularly released is so good? We at Remi have never had the goal of removing the human element of all business processes. We believe that A.I + BI are yet another toolbox that then allow humans to add the ‘gut feeling’ to a decision without doing the boring leg work (80% tasks). The ‘gut feeling’ element to business processes is where great analysts, managers, and executives come into their own, and we believe these 20% tasks need to be pushed as far as possible.