Associate Profile

Name:  Maria Diapouli

Role: KTP Associate for Crimtan 2012-2014

Brief description of project

I was working with Crimtan, a global digital marketing company as part of my KTP project.  Crimtan’s clients include Marks and Spencer, Santander and Sky and provide services including:

• running advertising campaigns
• campaign optimisation towards customer KPI
• audience profiling and segmentation
• audience targeting/ retargeting and reporting

The project was supported by academics from the University of Brighton’s School of Computing Engineering Mathematics.  My project focussed on developing and implementing innovative data mining and modelling techniques to enhance customer profiling and optimise advert placement and effectiveness.

This included:

Phase 1: I reviewed and analysed the company’s business objectives, customer base, technology, delivery architecture and processes. Data mining and artificial intelligence techniques and methodologies were also analysed in accordance to company’s business objectives and systems architecture and business processes were documented.

Phase 2: I investigated data availability and developed processes for data collection, data cleaning, data validation and preparation. Open-source algorithms were identified, implemented and the best-performing ones were incorporated into business workflows.

Phase 3 I developed innovative approaches to enhancing placement rules by developing models for prediction for user click and conversion rate.  Additional techniques were developed for accelerated tuning models behaviour in order to refine models for achieving peak effectiveness on new campaigns. The impact of these innovations is measured through the development of business impact metrics.

Highlights from the project

I had the opportunity to take part in industry relevant conferences including Thirty-second SGAI International Conference on Artificial Intelligence and the Seventeenth UK Workshop on Case-Based Reasoning, UK KDD BCS 2014 and was also able to participate in technical workshops. Which included Hadoop, Microsoft VBA, Microsoft SQL Server, all of which supported the successful delivery of the project.

A particular project success was the final outcomes of the model implementation. We can normally expect projects in data mining area to indicate that the accuracy of the learned patterns is between 40-60% however this KTP project delivered 90% accuracy of the data mining patterns, which a huge success for the company pushing them ahead of the competition in this area.

The best thing about being a KTP Associate

Through this KTP I have gained invaluable commercial experience, particularly in online advertising. I  had professional training in technical excellence/ state of the art (e.g. Hadoop, Hive, MS SQL, VBA), had business management/leadership exposure and had the chance to be mentored by lead academics and industry experts which I would not otherwise have been able to experience.

Challenges overcome during partnership

In the beginning, this KTP seemed to be challenging due to the nature of the company – the large volumes of data, and its specialisation in terms of low level complexity (e.g. raw cookies, xml data, csv files, etc.). However, both my company and academic supervisors helped me to overcome the challenges: The academic supervisors have provided their expertise in the Data Mining research area and the company supervisors have delivered descriptions regarding their data processing and optimisation systems. With the support of my supervisors the project has followed a successful path and has delivered insightful results to the company.

What I am doing now

My current job as a Research Officer in Data Mining Project Management at the University of Brighton has a similar feel to the KTP in terms of tasks, roles and responsibilities. In particular my everyday duties are related to:

• project management
• demonstration of analytics knowledge and implementation of heterogeneous databases and Big data
• competence in programming on data analysis tools
• research on various data mining algorithms and techniques

Definitely I would recommend a KTP since it is a very well structured program that offers huge support on the KTP Associate. The whole experience is invaluable.

Top tips for maintaining a successful and rewarding KTP partnership

From the Associate perspective:

• Be responsive to any changes to the project plan
• Always consider the final company’s objectives
• Consider your personal development as early as in the process as possible
• Enjoy the KTP Residential Course modules and keep in touch with other Associates