This text was originally posted in BigML blog as part of series of blog posts summarizing the upcoming talks to be given in 2ML conference

In this talk we will learn how my company Frogtek helps “base-of-the-pyramid” Mexican micro-retailers to better control and grow their businesses. With Frogtek’s point of sale (POS) systems these businesses register every transaction that takes place in their shops getting easy access to metrics and value-added services fueled by their own and other shopkeepers’ data. All this transactional data is also aggregated and fed into a business intelligence and marketing analytics system that Consumer Packaged Goods companies rely on for better visibility into a traditionally opaque sector. These valuable insights are driven the engagement Frogtek receives from its customers as they result in further gains in operational efficiency.

There are four key concepts for Frogtek to achieve success, all of them make extensive use of data, algorithms, and Machine Learning techniques:

  • Shopkeeper value proposition & onboarding process
  • Shopkeeper engagement & retention
  • Customer care
  • Data explotation

Shopkeeper value proposition & onboarding process

Shopkeepers in emerging markets seldom use technology only to have a better “picture” of their business, they usually have been running their shops for many years and think they know everything they need to in order to be successful. They are drawn to much more tangible value propositions if they are to use an electronic POS system. The value prop has to address a real, instant, and tangible pain point. At the same time, on-boarding barriers and learning curves have to be taken down or reduced. In Frogtek we try to achieve this by using data and algorithms to automatically build and update a massive product database and eliminate the need to create from scratch the shop inventory before the system is rolled out, by offering a pricing feature that automatically adapts to the shop’s price-levels and provides accurate pricing suggestions when new products show up. Similarly, we advise micro-retailers on what products to buy and what the optimal inventory level for those are each time a new supplier or distributor is being considered.

Shopkeeper engagement & retention

Although Frogtek’s more visible product is the POS system, the Frogtek business model isn’t based on selling these systems to shops. Instead, we focus on selling the aggregated data from those systems to large consumer packaged goods companies that are selling their products through this retailers. Therefore, data quality is key and this emphasis is reflected in our deal with shopkeepers: “you get the product for free or at a very low price, in exchange you commit to provide us with good data”. Detecting which users are going to become good users feeding accurate data and which ones are going to fail is really important to maximize the return of the investment we are making while growing our panels and achieving a sustainable business model. Machine Learning techniques can be used to achieve early predictions on which users have more chances to succeed and which ones are doomed to fail. This in turn lets us take corrective action.

Customer care

With a growing network of more than 1,300 micro-retailers in 12 different (and large) cities in Mexico, customer care becomes an important, expensive, and usually slower than desired activity. We use data strategically to generate alerts and forecast most of the issues our shops experience. We plan on using Machine Learning techniques to automate and optimize customer care task prioritization, assignment and routing with the proper next steps in every new case as based on our past incidence outcomes.

Data exploitation

Consumer packaged goods companies are eager to have extremely high quality and high granularity data to shed light into end-consumer behavior in a traditionally opaque channel. With this sort of data in hand, opportunities for new product development are countless. Traditional descriptive analytics, market share predictions, price and cost elasticity forecasting, stock-exchange evolution prediction or even consumer behavior patterns recognition are only some of the workstreams we, along with other partners, have given a thought to in the recent months. As data quality and completeness improves, the number of products and services for both CPGs or shops will likely grow, advance data analysis techniques such as Machine Learning can be used to maximize efficiency, efficacy, profit and more importantly the positive and tangible impact we are aiming at in our install base network of shops.

If this post was interesting to you please come to see us live in Madrid May the 8th!

And… by the way, we’re hiring!