EmPower Online, 29th May. Keboola Roadmap Reveal.
Don’t miss it!EmPower Online, 29th May. Keboola Roadmap Reveal.
Don’t miss it!
Join our newsletter
#noSpamWePromise
Oops! Something went wrong while submitting the form.
Keboola is a key driver for operational efficiency & profitability and is used across all data processes.
Rohlik is a data-driven e-commerce business that bridges the gap between food producers on the one hand and retail consumers on the other.
By delivering food at record speeds and without compromising on quality, Rohlik has grown over 50% year-over-year, becoming the first Czech unicorn in 2021. They’ve now expanded their operations across Europe, operating multiple brands including Knuspr, Gurkeri, Sezamo, and Kifli, to name a few.
Rohlik chose Keboola to drive their data-driven processes from day one, making it a core part of their data operations and running over 3,500 jobs daily.
We spoke to two of Rohlik’s head honchos, Machine Learning Team Leader Ondřej Hubálek and Group Head of Business Intelligence Petr Podaný, who together manage a team of 30 analysts, engineers, and machine learning experts, to understand how Rohlik grew its operations and surpassed the $1 billion valuation mark.
Ondřej tells us that his team uses Keboola across all the unicorn’s data processes and are massive fans.
If you’re starting a company with a BI or data analytics solution and you’re just figuring out what to do and how to do it, get Keboola. You can build a very complex BI or data solution quickly and relatively cheaply. And Rohlik is proof of that.
So what are some funky use cases that helped Rohlik grow?
Rohlik’s core business is the delivery of groceries. When it comes to deliveries, Keboola helps Rohlik optimize its logistics and supply chain in three ways:
To understand how delivery logistics are run, Keboola is used to (1) gather information from the logistics system, (2) clean and refine data, and (3) automatically send it to Tableau, where the data team at Rohlik visualizes and tracks metrics in delivery reports, such as:
When making deliveries, Rohlik needs to pinpoint optimal cargo fulfillment and adjust it in real-time.
Sending half-empty cars or cars with more bags than their actual carrying capacity would have a detrimental effect on the speed and efficiency of deliveries, as well as on customer satisfaction in the case of damaged food.
Keboola uses real-time machine learning to predict the number of bags that will be used to pack a customer’s order. The model is then used together with other optimization tools to plan courier routes as efficiently as possible.
Rohlik offers over 21.000 different products in one fulfillment center. Keboola helps them assess how many products are still in stock in their warehouses and build machine learning models for anticipating future demand.
Acting as a data hub, Keboola collects the data needed for the models, sends it to AWS (where machine learning predictions are made), and ingests the predictions from AWS.
Data is then sent to the ordering module, where the supply chain team uses daily predictions, as well as reports from Keboola about the accuracy of these predictions. Thanks to this, Rohlik’s supply chain team can control important metrics that track and drive the company’s profitability.
But Rohlik doesn’t just use data for advanced logistics optimization, it also harnesses it to inform and streamline growth activities.
With so many orders under its belt, Rohlik has a rich history of data it can use to anticipate trends. With the help of Keboola, they were able to utilize customer information and purchasing patterns to develop a recommender engine.
The recommender engine is used to predict products that have a high likelihood of being added to a customer’s order. Rohlik sends this prediction data from Keboola directly to its IT team, which integrates the personalized recommendations on its website and mobile application.
One successful example is the recommended product strip on the checkout page. This showcases recommended products so customers can add new products to their orders or replace existing ones just before they complete their purchase.
At first, it was difficult to establish if personalized product recommendations would have an effect on order frequency, total order value, or the price of items purchased.
Extensive A/B tests were run which found that the recommender system consistently increases the Average Order Value (AOV). The catch? The impact is low, with personalization increasing AOV by only 1%.
But that 1% is cumulative. With over 1M customers frequently shopping with Rohllik, the effect of personalized recommendations on the bottom line accumulates, bringing in a new stream of revenue.
Thanks to built-in reporting, it can also help spot bugs on websites and mobile applications. In one instance, the team noticed the recommender’s revenue was dropping due to a bug on the mobile app that hid recommendations. Here, data helped Rohlik flag a bug they weren’t even aware of!
A big obstacle for any e-grocery is food waste.
Rohlik sells fresh food, vegetables, fruit, dairy products, and other expirable food products. If the food doesn’t get bought, it ends up in the bin.
To tackle this issue, Rohlik uses Keboola and AWS to build models that can propose in real-time a discount for any given product, taking into account information about the current stock, historical sales of the product, expected remaining time to sell the product, and other parameters.
This way, price-conscious customers can acquire items at a discounted rate, while at the same time Rohlik can drastically cut food waste and increase its revenue.
Having the right data at the right time is crucial for Rohlik’s decision-making. One example of this is its ‘Daily Operations’ report.
Every day at 8 AM, the country team, consisting of the CEO, marketing, finance, operations, and other personnel, checks the numbers from the previous day and evaluates key metrics on what happened, what to improve, and what to avoid in the future.
This iteration helps Rohlik quickly adjust its operations and marketing activities to improve iteratively.
For the BI team, this means reports need to be ready first thing in the morning. Keboola’s reliability is important here because business users rely on the BI team, and the team relies on Keboola.
As you can see, data permeates almost every operation at Rohlik, helping the company gain an edge through insight. Some other important areas where data has helped the company grow include:
So with data touching almost every corner of Rohlik’s business, what does the future of data hold for the company?
Ondřej is enthusiastic about the experimental initiatives they’re just launching at Rohlik. With an endless stream of ideas, he points to two areas of work in progress that he’s extremely optimistic about:
Ready to discover what Keboola can do to help your business? Get in touch and start building a better data future.
