Leading development of interactive media recommendation prototype
Why does interactive media need to be recommended? A successful enterprise had a large catalogue of media that it distributed through its interactive service using a skilled marketing team to recommend items to customers manually. The question arose how to scale up this service, and how to predict the response of customers more effectively. One way would be to purchase more items to add to the catalogue, increase the marketing team and fund a lot more market research. The alternative, carried out in a team which I led because of my background in data science in earlier projects, was to develop recommender algorithms, predictive models that used data about items in the catalogue and customer’s response to earlier marketing to predict their future preferences. When tested these algorithms were as efficient as manual marketeers and have gone on to be used in the deployed service.