ShopUp CDP – Customer Data Platform (CDP) with Machine Learning models tailored for eCommerce
Retrospective 2020
The year 2020 was an interesting and challenging for all of us with all the turbulence and unprecedented events around COVID 19. There was less interaction with clients, but this gave us time to learn, experiment, interview various experts in the eCommerce digital world and focus on creating value for the sector. It was a good opportunity to design our new products and start adding various features to them.
In April, we designed a new prototype with focus on Retail, which is able to analyze and predict physical customers’ patterns in Hypermarkets with high precision using the latest advancement in technologies. We have the opportunity to reach very high precision outcomes based on statistical models rather than on the technology itself. It is a three layer system, consisting of sensors, IoTs and cloud servers.
In May, we participated in a global Datathon organized by Data Science Society, qualifying all the way to the finals . We developed a cutting edge hybrid article recommender applying BERT and RNN neural networks.
Since then we started modifying our tailor made Customer Data Platform, initially, all the data was stored on Relational Data Base, since November everything is scalable and on Spark clusters, and now the sky is the limit. Furthermore, we added new, interfaces and connectors. During that period we worked with Kupinauka (ecommerce) and Bulgarian science digital magazine in order to experiment and build our first case. So to sum up, it was a very beneficial and challenging 2020.
ShopUp CDP – the new beginning
This January was full of new endeavors and events. We, as a product company, finally introduced our new product ShopUp CDP, which aims to optimize products, audiences, and Facebook ad campaigns of every eCommerce website. It is relying on ShopUp’s customized Customer Data Platform (CDP), which we are building for the last two years. It is also applying various Machine Learning models in order to suggest the best products to the right audience.
Currently, we are focusing to improve Return On Ad Spend (ROAS) for Facebook ads and our two main directions are products and audience recommendations.
Products recommendation
CDP is able to identify the behavior of various customers, their interests, and patterns. We used millions of records on sessions or user level in order to identify which products are similar or suit best based on data we have. Our models are able to identify:
- Products with the highest growth potential and recommend them
- Products with high seasonalities – XYZ analysis
- Product bundles
- Product catalogs based on customer interests and patterns.
We know how complex it is to find a cross-sell or upsell product especially if you have thousands of products, so we do it for our customers.
Audiences recommendations
Google Analytics, Shopify, or Facebook pixel are not providing concise and detailed information about your audience when we speak about their sequence of actions and how to predict which ones have intentions to buy. We all know that creating or defining an audience the way marketers love their demographics is a tough task. Therefore we decided to challenge it and support marketers with audiences that can easily feed into their Facebook ads campaigns. The audience recommendation is dependant on the level of data we have access to:
-
Level 1 anonymized data
We identify patterns of visitor behavior their product preferences, website places they browse, etc. We can thus define segments and product catalogs based on their behavior and their physical location.
-
Level 2 with sales data
When we use historical sales data, we can go deeper into our analysis, finding seasonalities, similar patterns, and differentiate several segments of customer interests. There we can run a segmented approach of targeting over social channels or personalized over emails or mobile channels.
-
Level 3 access to Facebook or google ads API
This is where the magic comes, there are millions of records and variations of your behavior over social channels targeting, and based on the models we can identify which audiences behave best for the respective products you have. We are able to make all the processes so transparent that you can focus on winning combinations and wait for facebook to optimize them. The benefit is that we match the information from the main three layers and we have a full picture of how various campaigns behave.
Now, it is so easy and transparent for every eCommerce site to understand their audience and to increase their revenue on ads spent by using the historical data-driven ShopUp CDP AI.
Something which we are considering is to evaluate and different targeting channels and creatives.
New customer from UK
In January we welcomed a new customer – EDINA at the University of Edinburgh with whom we will work on improving customer experience for two of their flag ship services Digimap for Schools and DataNation. Digimap For Schools is an award-winning online mapping service designed to be used by teachers and pupils. The service is designed for use by children as young as 5, and up to 16 years old.
DataNation is a mapping service that combines official 2011 census data with authoritative Ordnance Survey maps. It allows You to investigate local socio-economic conditions, view, analyze and personalize this data to engender critical thinking.
Shopup will provide a 360-degree unified customer view achieved by combining different channels of information and monitoring customer behavior. The aimed result is to convert website visitors into loyal customers and foster long-lasting relationships supported by the ShopUp CDP platform – Customer Data Platform (CDP).
New Features
There is a major feature that we are launching in February, which is relevant to CDP and visualization of data after its cleaning, augmentations from external databases, transformation and cross-referenced with other sources. We are using an open-source platform called Superset, if you want to implement it into your business we wrote an article for the steps.
We also plan to start a complete redesign of our website with a more appealing look and information about our products, services, and clients.
In Conclusion
We are so proud to introduce and share what we are doing and how Machine Learning can be applied to daily work in the eCommerce sector in order to increase profit for each business. We believe that when the data is used in the right way it will increase the revenue and can automate a lot of processes, so if you what to stay tuned subscribe to our updates or contact us at info@shopup.me
Images by anncapictures, Merry Christmas , Wokandapix from Pixabay