EvoCon2017: Celebrating a data analytics success story

It started with a simple request.

Like a few other members of the Evo Pricing team, I work remotely and, while location-independent work has many advantages, it’s not always the best way to get an understanding of the work your colleagues are doing, particularly in a complex scientific field like predictive analytics.

So I asked for some virtual sessions with the data scientists to pick their brains. The answer was positive, but not in the way I expected. “If you’re available, why not come to our conference and meet them in person.”

And that’s how I ended up travelling to Siena for an intensive 2-day conference, EvoCon2017.

The venue

Certosa di Pontignano (www.lacertosadipontignano.com) was originally a 14th century Carthusian monastery which was acquired by the University of Siena in 1959 and is now the university’s official conference centre. It is a really beautiful and peaceful place, light years away from the hustle and bustle of a city conference centre.

As Evo Pricing CEO Fabrizio Fantini explained, this was deliberate.

“Because it is quite isolated, I joked with the team about there being no escape. But there’s a serious point behind this. Being in a magnificent historical building, surrounded by some of the most beautiful countryside in the world, you can’t fail to be inspired. And, after the intensity and knowledge dump of a full-on working day, you need space and time to reflect so your brain can process all the input. Where better than a Carthusian monastery in the heart of Tuscany?”

And this notion of isolation and inspirational surroundings is set to become a pattern. “Last year, we held our conference in a castle. This year a monastery. Maybe next year, a Norwegian fjord?”

The attendees

The Evo Pricing team and the board of advisors are diverse, both in terms of nationality and geographic location. The CEO and marketing and business development managers flew in from London, the developer and QA specialist from Athens and Saloníki, one of the pricing advisors came from Tel Aviv and the content specialist from Turkey. The Turin team of data scientists, IT, admin and media relations travelled down to Florence on the high speed Freccia Rossa train (300km per hour!!!).

The curious case of Attendee #18

Before we began, we were introduced to Sebastiano, aka Number 18. Not a job title I was familiar with. It seems that 17 is an unlucky number in Italy and, as there were due to be 17 attendees, he was invited along to avoid this misfortune. However, one person cried off and we ended up with 17 people anyway. It still turned out to be lucky for us because Sebastiano is a really savvy businessman and his input was very valuable.

The conference

Being a pre-millennial dinosaur who is little by little getting the grips with technology, EvoCon2017 was a real eye-opener. You often hear the expression “quantum leap” being bandied about, but until I attended this conference I never fully comprehended what it meant.

I listened in awe as the various presenters explained how sophisticated the company algorithms were and what kind of competitive intelligence on prices it made available. A far cry from the days sending a team of researchers armed with PalmPilots to local supermarkets.

9:30 We kicked off with an overview of the business from Fabrizio, highlighting the rapid growth of the business both in terms of revenue and staff. He emphasized the company values of learning, being open to feedback and rising to challenges. Some of the global goals he outlined were speed, simplifying the problem and making the impossible, possible.

Evo Pricing closes the gap between potential and actual demand, combining potential demand management with actual supply forecasting. We do this by integrating artificial and human intelligence – Data intelligence + Human factors + Decision Support Tools. This leads to the development of self-learning Decision Support Tools.

In effect, what we are doing is bridging the skill gap that retailers have in-house, with the addition of our “secret sauce” – the insight from store staff.

10.00 Davide, the Systems Admin, gave an overview of the Evo pricing system and its security features.

Evo Pricing implements ultra-security – something that other developers tell us is not necessary but I beg to differ. The integrity of the system is our top priority and our processes reflect that. All systems are self-contained and each client is compartmentalized on different servers.

It’s not just about security. Stability and user experience are vital, so nothing gets pushed to production (the user interface) until we are 100% sure it is stable.

We use ‘R’ as the system data language. The advantages are that it’s the language of academia and it is open source so we can quickly learn what other people are doing and react quickly.

10.30 Nick, the Developer, talked about the portal.

Basically the portal consists of the Frontend (Home page) and the Backend (Admin) with various levels of access permissions. It’s a buffer between IT and clients – generally it is the data scientists who deal with clients.

It is our aim to make possible as many data visualization tools as possible. To this end we use different content types (SSRS, Shiny, Charts, Dashboards and Tables) which enables us to support 10 types of charts. An example of the high granularity of our data can be seen in Tables. Clients can update one cell at a time with a user interface similar to Excel, changing as many parameters as they want, and all the related visualizations update instantly. This is very important because it allows clients to see thousands of permutations very quickly so they can do laser-targeted research and drill down very deep.

10.45 Tina, the QA specialist, with an intro to the Evo Pricing add-in for Excel and testing methodology.

Clients can download updates via the plugin. Evo Pricing has a very robust Quality Assurance system in place to ensure clients have no problems with updates.

Our Software Development Life Cycle (SDLC) uses the Waterfall system. Within this we do Regression Testing ( making sure older programs still work with new updates) and Functional Testing (black box testing in a compartmentalized environment separate from the live system).

There is a huge emphasis on writing high quality bug reports using detailed templates containing step-by-step instructions. You can’t fix anything complex without a good bug report.

11.00 Giuseppe, Senior Scientist, gave a demo of the application.

There are two ways we produce results – through the Excel add-in and the Portal. We can give information in 3 ways – by showing what happened, what would have happened if something had changed in the past, and based on client interaction.

Our Competition Monitoring tool shows how the market is moving, based on information collected daily. Our Dashboard tool illustrates changes by merchant, price range and category in the last month. Finally, the Excel tool delivers a complete analysis of prices, markdowns and changes between selected dates.

These tools are so sophisticated that it’s possible to measure something like the Brexit effect on prices, even factoring in different hedging expiry dates.

Promotions are a big pain point for businesses because a lot of retailers don’t know if they work. That’s what Evo pricing does so well. We measure the impact of promotions – the expected pieces, revenues, and margins without promo. We are able to tell you which promos perform better, and in which metric.

11.30 Elena, Data Scientist, gave a demo of the replenishment aspect of the application.

How do we decide what is the right number of a particular size of a particular item to send to a particular store in a particular week?

We do it via a replenishment proposal at article/size level for each store. We continuously rework the proposal based on store managers’ feedback. Data collection from stores and the calculation of optimal goods allocation leads to the knowledge we need for the execution of shipments between stores and warehouses.

Monitoring of the process is done via fulfilment rate of requests (number of store managers’ requests fulfilled), stores’ activity (stores’ reaction to proposed replenishment), and the popularity index (most and least requested items).

What we find very surprising is the small number of items that stock out and that the long tail is really long.

Our biggest challenge is not about the data. We need to work hard to convince senior management that store managers know better than Head Office – “the collective wisdom of the stores”.

12.00 Enrico, Junior Data Scientist, gave a demo of price change recommendations.

Should I change the prices of the items I sell? Which ones? How? These are the big questions our clients ask us.

We support their decision-making process via data acquisition, automated processing with elasticity and sales forecast, followed by visualization in Excel via Evo Pricing add-in. Once they have decided on a price change, or changes, we monitor performance and adjust over time. The time needed to see the effectiveness of the price changes is usually a minimum of 4 weeks.

After lunch we broke up into smaller teams for 3 break-out sessions on Marketing, Science & Applications, and IT.

EvoCon2017 concluded with an outline of our expansion plans for the rest of 2017. Stay tuned, it’s going to be big!

Closing message from Fabrizio

Thank you for having made EvoCon2017 such a great, memorable event! Sun-blessed, and luck-blessed, with such a great group.

We have also been blessed with a fantastic past year of growth and rewards.

But let’s not forget that the challenge ahead of us is still significant, and there is no time to rest on our laurels just yet.

We live in such a chaotic world, where things change fast, and there is dire need for the kind of trusted guidance that we can provide our clients. So we must keep going at it, harder than ever.

Which, I am sure, is exactly what the last few days together have inspired us to do!

A final thought

I was very impressed that Fabrizio had managed to gather together such a diverse, intelligent and ‘simpatico’ group of people together in one company. In answering my question about this, he revealed that It was no accident.

“Life is too short to work with people you don’t like.”

There was time for fun and culture, too!