Fidenda five series - five reasons that Anaplan and machine learning make a great match

September 2021
Written by
Tristan Colgate
minute read

If you’ve clicked through to this blog, then congratulations .  The IT industry is in hyperbole overdrive when it comes to Artificial Intelligence and Machine Learning and hopefully your own interest is what has brought you here.  I’m going to reward your curiosity with a little cynicism about the over-inflated claims of the possibilities of ‘AI’, and give you comfort that you’re not about to be replaced by a computer program.  Then we’ll get on and talk about Machine Learning, which is a term I’m more comfortable with, and an area where there are some genuine benefits to be gleaned - especially in the context of Anaplan.

A lot of what is claimed to be ‘Artificial Intelligence’ is really just advanced statistics.  Take neural nets for example; these are remarkably successful at emulating lower level human sensory perception and can be taught to recognise handwriting, detect cancerous cells in x-rays, and work out whether a photo has a cat in it.  They do so by applying a mathematical algorithm called back-propagation on a large volume of training data.  For example, if you want to train a neural net to recognise handwriting, you provide it with thousands of examples of handwritten letters, as well as the ‘answer’ for each one (i.e. which letter they represent).  It actually works in a similar way to the brain by storing the results of the back-propagation as strengths of connections between notional ‘neurons’.  But brains come with some pre-wired abilities, and are connected to eyes, ears, fingers, stomachs, and a world that responds to what the brain does and doesn’t give us a clear ‘answer’ when we’re learning.  Perception is not intelligence.  Neural nets are not artificial ‘intelligence’ in a meaningful way that threatens our utility as humans.

But that ability to train a machine to learn patterns in data is useful and powerful, more powerful than humans’ capacity to do so.  This is why I am personally more comfortable with the term Machine Learning.

Let’s take a business scenario that lends itself well to Machine Learning, Demand Planning.  Traditionally, Demand Planning has used statistical techniques for looking at past demand, and using it to extrapolate forwards.  Algorithms are easily able to separate out seasonal trends from an underlying trend in demand, and have been successful in reaching a certain level of accuracy in predicting future demand.  Anaplan is a recognised market leader for Demand Planning and supports these techniques easily using its Hyperblock modelling platform.

Yet, past demand is just one factor that we know, anecdotally, predicts future demand.  There are a plethora of others:  macro economics (inflation, GDP, employment rates, market sentiment), promotional activity (whether your own company’s, or your competitors’), weather, competitor activity, new products entering the market.  The relationships between these and demand are subtle and lend themselves well to Machine Learning techniques that can take this data alongside historical demand and understand the patterns in a way that humans cannot.  These relationships can then be used to predict future demand (this also requires predicting the future factors such as GDP and competitor activity of course).

At the forefront of the race for EPM and analytics vendors to integrate ML capabilities into their toolset has been Anaplan.  Early partnerships with leading ML technology partners such as Google Tensorflow have ensured that ML generated insights can be integrated easily into the Anaplan platform.  The recent acquisition of Israeli-founded predictive marketing and sales analytics company, Mintigo, demonstrates a commitment to bringing ML capability in-house.  With this in mind, we present below five key reasons why you should consider Anaplan as your platform if you are looking to use ML to enhance your business decision making.

1. Data

Machine Learning relies on having detailed and accurate data in large volumes.  Often this data comes from a plethora of sources and needs to be harmonised and joined up in a meaningful way before it can be useful to a Machine Learning engine.  Anaplan is the perfect platform for achieving this being cloud-based and easy to integrate into a wide variety of sources, with a modelling engine perfectly suited to the task of manipulating large volumes of data.

2. Process

The input to and output from Machine Learning algorithms must sit within an overall business process if they are to be useful to a business.  The output of statistical or ML Demand Planning, for example, must then be passed into an overall S&OP process for decisions to be made about how to meet that demand.  These processes require features that ML engines don’t have – the ability for multiple users to collaborate on data in a way that is organised around a business process unique to the company.  Anaplan supplies all of these abilities.

3. Visualisation

ML algorithms produce their insights in the form of data, sometimes quite a lot of it.  Anaplan is an ideal platform to present this data in a meaningful way, in business terms, to the people that will consume that insight.  As well as being able to present data in charts and tables, Anaplan puts a security layer around the data to ensure that only the right people access the right data at the right time.

4. Ownership

Machine Learning is complex and an area of Data Science.  However, for it to be truly useful, employees without deep data science skills need to be able to access its powers and the analysis it produces.  By ring-fencing the complexity in a Machine Learning tool, but surfacing it in Anaplan, this brings the power of ML to the fingertips of finance, and supply chain professionals in an environment which is user-friendly and can be maintained by those without deep technical skills.

5. Humans

Finally, ML platforms miss one vital aspect – us.  ML algorithms are only as good as the data they are given, yet as we go about our day to day lives, we are constantly confronted with potentially new sources of information that informs our understanding of what has happened, or our predictions of what will happen.  This is true intelligence, and something that current day AI cannot deliver on.  Anaplan is a platform for humans, where humans can input the results of their thinking and understanding and override what the algorithm is telling us.

Share this insight:

Related insights