Lately, Carrefour Italy has been working closely with Salesforce. As of 2020, two main solutions were already in place:

• “Salesforce Service Cloud”, for the call center and customer services.
• “Salesforce Marketing Cloud”, for advertising and media management and  campaign automation.

In parallel, in mid-2020, Carrefour Italy demonstrated that the country knows how to speed up, engaging in a profound digital transformation that started strong with the setup of a new e-commerce site. Salesforce was also chosen to help by bringing in its expertise as well as its robust, modern and evolutive suite “Salesforce Commerce Cloud”. This means that, in a relatively short period of time, an enormous amount of new e-commerce orders data will have to be processed, understood, analyzed and leveraged by Carrefour at the profit of its customers. Hence, data is hence essential to ensure a performant digital service to all customers.

When it comes to data, the approach from Salesforce is quite different from the one found in big data & analytics specialists. It is highly grounded in the idea of embedding data capabilities and intelligence within Salesforce solutions themselves (Service, Marketing and Commerce Cloud), as opposed to extracting info to a specialized platform (datalake) and playing with it using advanced analytics techniques like machine learning. Salesforce data capabilities are regrouped under the name of “Einstein Analytics”. Einstein helped Carrefour Italy create powerful, scalable, interactive, dynamic and intuitive dashboards. It also enabled us to prepare simple or complex datasets that can be analyzed by “Einstein AI Discovery”.

We learned that data prep is the first and most important step to start your data journey

In order to work with our data, we needed to transform the relational Salesforce objects in aggregated data sets that will be displayed in our dashboard. For this reason, we started by analyzing in depth the Order Management System (OMS) data model.

 We used “Dataflow” and “Recipes”, which are processes used to extract, transform and register your data into one or more datasets.

The advantage of the Salesforce suite (Service Cloud, Marketing Cloud, Commerce Cloud) is that there are a lot of sources that you can use to extract data (here you can find all the possibilities). In our case we used the following ones: 

• Salesforce CRM objects
• Static .csv files
• Google Big Query leveraging the native connector with Salesforce

It is worth to mention that the Dataflow builder is a very powerful tool that allows, even with no coding skills, to create datasets for your dashboards with filters, joins, append functions and more. You should keep in mind that you’ll need to design a dataset that is as scalable as possible!

Creating a dashboard

Once you have prepared a good dataset, you can move to the next step and display your data where the magic happens, in a dashboard! 

Building a good dashboard does not only require high quality and fresh inputs, but also showing correct data in a very intuitive way. In order to better understand our eCommerce business, we created different dashboards that can deep dive in the data of every  order, democratizing the access to valuable info such as :

• After sale actions
• Churn rates
• Shipment times and slot saturation
• Fulfillment

To give an example, when we say that e-commerce managers can better understand fulfillment, this means they can get the status of any order in a blink of an eye. 

Performance summary

We wrapped all these  learnings within the performance summary dashboard , which generates valuable insights on different dimensions: 

• Time – Total grouped by month for the last 12 months.
• Geolocation – Total grouped by the Italian provinces directly within an interactive map of Italy.
• Shipped method –  Total grouped by the different carriers or services (pick&pack, dropshipping).
• Order status – Total regarding order status in its life cycle in order to have a quick look at the actual situation

This interactive dashboard  includes the possibility for us to click directly on the values within each widget. As a result, e-comm managers can analyze the desired segments and select the correct filters in a very intuitive way.

It is also possible to explore the data within each widget and share them in different ways:

• With another user within Salesforce using chatter or quip.
• Downloading a .png image, a .csv or an Excel file directly on your device.

–

Mobile

All the dashboards are natively available on mobile devices (smartphones or tablets), through the Salesforce Einstein Analytics app that you can download on both Google Play and Apple Store.

Based on our experimentations to date, the dashboard seems to work better on iOS devices but Salesfoce  has been making great progress on Android for the past few years.

All in all, we can say that having all this data within the smartphone is particularly useful for business managers, that see their productivity grow by leaps and bounds.

The next step: insights and recommendations

After setting-up adequate reporting for e-commerce order management, the next step is to leverage to the max, ensuring we take the right actions in order to improve our digital business. It is in this step where the AI can help you to find the hidden data correlation.

Yet again, you need to have a deep knowledge of your data but you do not need to be an AI expert: Einstein can do it for you!

For this deeper analysis, we are planning to use another tool of Salesforce’s Einstein family: “Einstein Analytics Recommendations”. 

This tool allows us to set objectives in terms of specific data to maximize and minimize (e.g., the total of orders or the shipment time). Then, it can automatically identify the insights (WHAT happened, WHY it happened and WHAT COULD happen) and the recommendations applying an AI algorithm.

This tool provides support to the business in a very quick and smart way, without significant investment.

The results, as you can imagine, are heavily dependent on the amount of records available for analysis. For this reason we plan to start using this tool at least three month after the go-live of the new site… We’ll keep you informed of course!

You can also check our other articles regarding data projects.