Home‎ > ‎

Our approach

Let's get something out of the way first. Our business is data modelling and analytics. We can handle high volumes and a large variety of data, but it is not about big data. To re-purpose a cliché - "Size does not matter. It's how you use it."

Data modelling & analytics

A model is a simplified representation of the world. It captures just enough complexity to answer questions that we care about and ignore the rest. Therefore, all models are false. Some, if constructed with care, are useful.

Hedgehog & Fox approach any problem by first understanding the business. We identify the questions that need to be answered, the relevant factors, and the assumptions behind business. Next we construct multiple models to represent the moving parts, how they interact with each other, and how they affect the outcome. Then we test them with a combination of your data, data from public sources, and our own proprietary data.  The result is a model that is expressed in terms of concepts from your business that you are already familiar with. You can interpret the model in a meaningful way and derive actionable insights from it. If this sounds a bit abstract, have a read of our case studies and find out how we apply our approach to real world problems.

This is the opposite of the purely statistics driven approaches used in many artificial intelligence problems.  There is a place for a statistical approach. If you are given a photo of a hand-written alphabet, you may say it is the letter "b". It is very difficult to argue why it is a "b" and not a "g". We just know it is. A statistics-based approach mimics this intuition by using the relative brightness of each pixel and its correlation with its neighbours, and comparing them with the same information from known pictures of the letter "b". When applied well it can imitate our intuitive understanding of the world. However, the outcome of such an analysis is difficult to interpret and reason.

Big data must be handled with care

Once upon a time, very few human activities were recorded. You ate, you slept, you bought a cake and visited your friend. But you did all of that without swiping a credit card, touching your oyster card, taking pictures of the cake with Instagram or telling your friends about it on Facebook. Capturing and storing information was expensive so we chose very carefully what we recorded - rainfall, temperate, crop yield, tractor production etc. We recorded what we believed to be relevant to give us insights and wisdom about the future.

The Internet changed all of that. Now most of us carry a smartphone. If you go out for a jog, your mobile phone company knows about it from the mobile signal. If you reward yourself with an ice cream afterwards your bank knows about it through your contact-less payment. Every single activity leaves a trail of digital footprints. However, it does not necessarily mean there is more wisdom out there. It just means there is a lot more data.

In some cases the new data allows us to look at the world in ways that were previously not possible. However, it also gives us many more spurious coincidences that can lead us to false insights. If you toss a coin enough times, you will eventually get ten heads in a row by pure chance. Therefore, in the era of big data each stage of the distillation process from data to wisdom must work harder to filter out the noise. A model based approach to analytics allows us to do exactly that - we can augment the knowledge mined from the data with our experience and judgement.