Wednesday, May 6, 2015

From Sales Cycles to Shopping Carts - Analytics on Display at the Toronto Data Mining Forum

I'd been looking forward to today's meeting of the Toronto Data Mining Forum for some time. This group consistently offers extremely technical papers - it's really a statistician's dream, to be honest. Not being a 'man of numbers' myself, I often find myself struggling to keep up with the complex algorithms, formulae and terminology which our speakers throw around with some casual abandon. This meeting, however, promised something a bit different - a focus on sales and retail analytics.

The meeting began with Pramod Dogra and Iqbal Habib of Shoppers Drug Mart discussing the recommender system implemented in a retail environment to help better serve up relevant, timely offerings to their customers. Shoppers has a huge customer ecosystem-  nearly 1 in 3 Canadians is a member of their loyalty program called 'Optimum' - so there is a rich pool of data to draw upon. Interestingly, they are now beginning to develop more robust online offerings and a richer web experience to help mine customer data from real-time clicks, purchase intent, traffic patterns and more. Having spent quite a bit of time lately in presentations around re-marketing and programmatized advertising, I couldn't help but see opportunity for Shoppers to leverage the social web in a much more powerful way. Having said that, Pramod was happy to walk through their methodology for developing scoring/propensity models from a high level. With thousands of products and millions of customers, figuring out who is most likely to respond to offers given similar buying profiles of other individuals was quite the challenge. It was illuminating to see the Shoppers approach.

Our next speaker kept up the sales/retail theme by looking at the data side of the equation - how Enterprise Miner nodes could be used to score, test, validate and adjust models which predicted buying behavior or offered sales forecasts. I must single out our speaker Pat Valente from SAS Canada for some well-earned praise. Pat stepped in to deliver this talk with less than 48 hours notice as our original speaker became quite ill and quarantined under Doctor's orders. Pat familiarized himself with someone else's slide deck and was able to step up and not only speak to the topic at hand, but to demo it through the Enterprise Miner software itself. As if this wasn't enough, he fielded questions expertly and assuredly. Our Solution Specialist team here at SAS Canada is truly the best! Fantastic work, Pat. Contextually, the subject material was right in line with the previous talk - it gave a technical wrapper to the business situations previously outlined. I love synchronicity at a meeting.

Our final speaker was none other than SAS Global Forum Chair Elect (2017) John Amrhein of MacDougall Scientific. John is a statistician of no small skill - I'm envious of his ability to not only think through and address complex business issues, but to be able to easily explain it to a non-statistician like myself. His talk was fascinating. The topic was 'B2B Sales Analytics' and it really exposed how a typical sales cycle could be informed by Bayesian statistical analysis. Inputs included whether or not competitive intelligence was used, the experience level of the sales rep, the maturity of the product/service being sold... plus a whole host of others. By building Bayesian models John was able to fine-tune the expected outcome - especially as new data was introduced over time - to really have strong predictors of the sales process. Beyond this, by tweaking certain variables within the model the relative weight and impact upon success in any given situation could easily be explored. I was absolutely riveted. John also suggested (rightly so) that this methodology could be used for virtually any process wherein the outcome was in question. I left his talk knowing much more than I did previously about the sheer power of Bayes statistics.

I can honestly say that this entire meeting was a real eye-opener. As a marketer, I could easily relate to the challenges of properly identifying and offering value to customers - a task being tackled by the Shoppers team with some skill. As a SAS user, Pat Valente's Enterprise Miner talk brought a technical framework to the same challenge. Finally, John's talk layered in a statistical approach which was truly 'next wave'. His method seemed to me to be a sound, viable approach to mitigating risk and accentuating the possibility of success. My thanks to all speakers for their great work and effort! If you weren't able to attend, have no fear - the talks will be posted shortly here.

Next up for me: I'm off to Quebec. I'll be voyaging with the Section Chief for Statistics Canada in charge of SAS, a true figurehead of SAS expertise and a great individual as well. I'll have more on that trip next week.

Until then...