Customer Use Cases

Our customers are continually finding new and challenging projects for us and our catalogue of use cases is constantly growing. Some examples of the types of projects we have previously undertaken are:

Inventory Management

A national bakery chain wanted to reduce the time and cost of predicting sales patterns for new and existing product lines. We were given a range of business data and tasked with developing a model to automatically predict future demand.

Our task was to match the capability of the bakery’s own team of experts. The model we produced was proven to be more capable than the experts as it was able to reduce costs and fulfil otherwise missed opportunities equivalent to 2% of turnover. These savings were far in excess of the original aims of the project.

Price Prediction

A jewellery retailer wanted to be able to identify irregular pricing patterns in their supplier network, and therefore be able to negotiate with a higher degree of confidence.

We were given data on defining design details, materials and manufacturing processes and tasked with predicting the suppliers quote.

Our model was able to predict cost to within a 25% margin of error for 80% of the products. As data scientists, we were frustrated that we weren’t able to validate the model and improve upon this figure. However, we learnt that it was more than adequate for the customer as it gives them a much better negotiating position than had been achieved previously.

Image Classification

A security provider had previously purchased a commercially available image recognition software that was achieving approximately 65% accuracy when automatically categorising human features such as age, gender and ethnicity from low quality CCTV images.

We were given a catalogue of similar low quality images and tasked with improving the reliability of accurately categorising these human features.

After a short period of learning our artificial intelligence algorithm was able to achieve approximately 75% accuracy; an astounding figure given the poor quality of the images.

Our Technology Partners

We use a range of platforms to develop bespoke solutions. We select platforms based upon the project specification and the need to integrate with existing infrastructure.

Data science is a rapidly evolving field. Consequently, we maintain a close relationship with our platform providers. We aim to assist their own development with great feedback based upon our own and our customer’s experience.
If appropriate, we can advise on the pros and cons of a particular platform. However, if you have a specific preference we will endeavour to accommodate it.

Case Study


A national bakery chain wanted to reduce the time and cost of predicting sales patterns for new and existing product lines. We were given a range of business data and tasked with developing a model to automatically predict future demand.

Our task was to match the capability of the bakery’s own team of experts. The model we produced was proven to be more capable than the experts as it was able to reduce costs and fulfil otherwise missed opportunities equivalent to 2% of turnover. These savings were far in excess off the original aims of the project
  • thumb1
  • thumb1
  • thumb2
  • thumb3
  • thumb4

Terminology


Data Science
Data science is an interdisciplinary field that acts to interrogate data, often from multiple sources, to extract knowledge or insights so that it’s possible to make informed decisions and react appropriately in a given environment. Carefully considered data processing and analysis can result in a remarkably positive impact on both your bottom line and your customer’s perception of you and your business.
Machine Learning
Machine learning was defined by Arthur Samuel in 1959 as a field of study that gives computers the ability to learn without being explicitly programmed. The wider population has only recently become aware of the potential of machine learning techniques. This is primarily due to the wide availability of very large digital data sets and the computer processing power needed to execute the machine learning algorithms. In terms of human learning a machine is able to develop at a much faster rate. However, contextual understanding and comprehension is currently beyond its capability; Human beings are still required to make that big call, even if the supporting evidence and data is provided by a machine.

Links to ongoing projects, presentations and developments in data science and machine learning
www.kaggle.com
www.bristolisopen.com
opendata.bristol.gov.uk
www.irc-sphere.ac.uk