While other discount e-commerce marketplaces are going bankrupt, Slevomat has been experiencing unprecedented growth year-over-year ever since its humble beginnings eight years ago.
Recently, they’ve made a name for themselves offering experiences, like events and travel, at heavily discounted rates.
To delve into why they’ve been such a success, we have to look at one of the cornerstones of their company – their approach to data.
From our experience, only a handful of European companies have been able to automate their business processes on the level that Slevomat has, so it’s advantageous to anyone trying to follow in their footsteps to look at their methods.
So what does Slevomat’s data department look like? What have they dealing with, and how has Keboola improved their business? Can they ever imagine their data management working differently?
To answer those questions, and more, we sat down with Zdeněk Linc, Michal Kolomý, and Jana Chroboková from Slevomat’s GHT department._
In Slevomat, what does a regular day of working with data look like to you?
We ensure that Keboola is being fed the entry data that it’s supposed to receive and that it’s transforming the data correctly. Also, we manage ad-hoc analyses for customers, a/b testing, and even sales. With all these responsibilities, we stay pretty busy.
We’re currently using Keboola as a BI tool – we pour data from SQL database, Zendesk, and Google Analytics into Keboola, and once the system processes it, the results go into GoodData or BizMachine.
Have you been able to start discovering new journeys that Keboola can help you undertake?
The truth is, we’re using Keboola a bit differently than we thought we would when we first started using the system. Originally, Keboola was used as a data preparation method for GoodData. We used it to work with our SQL database, but we found that running complex operations took about 20 minutes. We later figured out that we could use Keboola to integrate SnowFlake, which allowed us to do incredible things. After that, we have basically been able to work unrestricted with any data directly in Snowflake.
For example, thanks to Keboola, we’re able to send data to Facebook’s API to create more effective ads.
You could say that the introduction of Keboola with all its opportunities actually led us to create the Growth Hacking (GHT) and CRM teams.
Have you encountered any restrictions or limitations while using Keboola?
Slevomat works with large amounts of data every day, and processing all of it is incredibly time-consuming. If you look over what we have in regards to notifications and mailing, you could easily be talking about hundreds of millions of data rows. Because of the sheer size of our databases, speed is critical when considering data solutions.
Before introducing Keboola, we used to try to handle our databases on our own, but some of the more difficult operations would take hours to complete, and every dead end we encountered would waste vast amounts of precious time. Now, thanks to the symbiotic relationship between Keboola and SnowFlake, everything works together so much more quickly. It gives us the freedom to do what we want much more rapidly, saving us tons of time and accelerating the progress of data processing in our company.
You mentioned that data processing used to take a lot more time. What did your processes look like before you started using Keboola?
Honestly, we try not to think about those times. Thinking back on it, it feels like we were working in a different century.
When someone needed data, that would have to make an order, so to speak, with the IT department. Many of those tasks back then were done by Tomáš Čupr (the founder of Slevomat) himself. Back then, we didn’t have too many data requests, and the requests that did come in were mostly for data used for reporting to management. Everyone knew that data requests were a long and arduous process and that it was difficult to get new and relevant information from the system, so most people didn’t bother making a request. We also didn’t have access to all the tools that Keboola has to offer, so, for example, the marketing department was only able to work with data from Google Analytics before we made the switch.
So you’re saying that everyone was instantly able to see what was possible once Keboola was introduced, and they suddenly started making data requests that they previously hadn’t been making?
Exactly! A salesperson, for example, could tell us that they were starting a new campaign and they want to know which town or district they should target. Instead of taking ages to get back to them with the results, which may or may not be accurate, we can now complete this kind of task in mere minutes. In fact, we’re typically able to get those answers ourselves without having to consult the IT department, freeing them up to deal with more important tasks.
Before Keboola, the marketing department would have to submit a request to the IT department, and it would take time since the IT department typically had other important things to take care of. Now, we can export the results ourselves in a handy CSV document, which our marketing department can then directly import into Facebook to use in ads, saving people in both departments time and effort.
How has Slevomat’s approach to marketing shifted in the last two years?
Once we were able to automate most of our daily routine, we were able to start approaching company data, including marketing data, from new perspectives. One of the examples is that instead of using the traditional A/B testing based on cookies, we’re now able to do them on users instead. Since we’re able to work with more relevant data, we’re seeing a lot more interesting results.
Another example of this is our internal category “Good Deals.” This category is a simple model that determines which offers would work well when giving recommendations to users on the website. This model used to only work with sales, but we’ve been enriching it with new data, and now it contains the number of users that were brought to the page by a marketing offer, up-sell of other products, and so on.
Thanks to the model, we can recommend not the bestsellers, but the products that drive the most value to the business as a whole.
Have you found ways to work with the data from these best-selling deals?
Absolutely. We run daily evaluations based on Keboola data so that we can give recommendations to our marketers about which deals they should be pushing and which are likely to do well. There are many similar possibilities, like using GoodData to generate ads for Facebook directly.
How would you describe Keboola to someone who isn’t familiar with the system?
It’s something like a well-tuned ETL [tool.] (laughter) Imagine that you come into your office in the morning, and all the routine work was already done by “someone” during the night. Now you’re free to handle the requests currently coming in instead of having a large backlog of work. Instead of spending hours trying to catch up every morning, you can spend about 15 minutes reviewing what’s already been done. When we hear colleagues in London complaining about reaching the limits of their solutions, we feel like we’re talking to someone from another century.
How do you like cooperating with people from Keboola?
We’re totally happy with them, no complaints whatsoever. When we need help with something, we contact support, and they quickly and efficiently solve the occasional issue that arises.
