My PhD project is under the above title and is co-sponsored by marketing and customer analytics company Aimia. I am co-supervised by Prof. John Boylan and Dr. Nikos Kourentzes at Lancaster and by Dr. Mark Howland at Aimia.
Forecasting demand in retail has long been a fundamental issue for retailers. Long-term strategic planning is all about prediction, and demand forecasts inform such processes at the top level. At a lower level, marketing departments find the capacity to predict demand under various arrays of promotions valuable. At the micro-level, supply chain and inventory management processes are reliant on fast, accurate, tactical forecasts for each stock-keeping unit (SKU), to keep stock levels at a suitable level.
Demand forecasting techniques traditionally employed in industry have focussed on extrapolation of past sales data to predict future demand. However, as demand forecasting becomes more complex, with ever increasing ranges of products, there is an increasing need for forecasting tools which use more information. Causal factors, such as promotional activity, have a driving effect on demand patterns and accurate modelling of these can prove crucial to forecasting accuracy.
A further challenge is the huge amount of data now collected at point-of-sale in retail, right down at the micro-level of individual SKUs and transactions. The massive datasets that are compiled as a result pose challenges for forecasting, but also may hold the key to the major gains that can be obtained in the development of prescriptive models for demand.
My PhD aims to bring together these different strands of thought to develop a demand forecasting framework that harnesses the potential in big datasets and incorporates causal factors in demand, such as promotions, to produce accurate forecasts which can provide value at all levels of a retail business.
I completed two independent research projects as part of my MRes. The projects were an opportunity to explore a broader range of topics in Statistics and Operations Research.
Intermittent Demand Forecasting
Intermittent demand comes about when a product experiences several periods of zero demand. Products which might exhibit intermittent demand histories include specialist parts in the aviation, automotive and other manufacturing industries, along with products nearing the end of their life cycle. However, a consequence of the many zero values is that traditional forecasting techniques (eg. exponential smoothing) perform poorly in terms of accuracy and also for inventory control purposes.
In my report, I tried to give a summary of different approaches to the problem that have been proposed, along with a discussion around the suitability of different error metrics used to evaluate forecasting techniques.
The agent-based paradigm of simulation offers an intuitively different approach to traditional simulation modelling. Rather than prescribing at the top level all the rules in the model, an agent-based model describes behaviours in individual entities known as agents, and allows them to interact to see what behaviours will emerge. Often, macro-level effects can be recreated through this micro-level design.
By exploring different models in the social sciences, my report aimed to investigate the capabilities and impact of agent-based modelling in an active research area. Important points around the often-overlooked validation and verification stages were also discussed.
This project, entitled "An investigation into the longest edge-lengths of the nearest neighbour graph (NNG), and other selected random spatial graphs" used methods in stochastic processes and probability to analyse the limiting distributions of the longest edge-length in random spatial graphs. Ie. if P(n) is a homogeneous spatial Poisson process with intensity n, what does the distribution of the longest-edge lengths of an NNG look like as n goes to infinity? The project included a detailed investigation for the NNG on the 2 dimensional unit square and unit torus, and extended that to 3 dimensions (and more). It also detailed a framework that could be used to investigate the same question for a family of supergraphs.