Daniel Gorol, Hello bank!: Classic Corporate BI Tools Would Be a Financial Suicide for Us

Until last year, it was called Cetelem, a provider of consumer credit and credit cards. In the autumn of 2017, BNP Paribas’ subsidiary was transformed into a new bank. Hello bank! caters for retail clients and has come up with several new products, so far not made public, for them this year. We visited the company for comments and news from the bank’s data back-office.

Daniel Gorol accepted our invitation for an interview. He has been working for the bank for nine years, since June 2017 as Digital Marketing Manager.

At the bank, you’re essentially a marketing veteran, and you have many years of experience working with Keboola. What were the early data-mining days like?

In the early years at least, I didn’t have an overall picture, so I’m going to speak for Marketing only – it was a bit like the Wild West. Everything was measured just in conversions – how much they cost, how many we had, and which channel generated them. Back then, we simply used the last click, which can be universally and simply communicated, but we knew that this led to overestimations. When we needed to get some insights, we needed to open AdWords, Sklik or any particular app. We then analysed everything in complex Excel files and numerous tables, which was an inflexible approach in terms of both user friendliness and analyses. When we had all the data aggregated in Excel, we couldn’t continue verifying and investigating details and new connections. What’s more, we lacked proper visualization…

When did you decide to change things?

Around 2012 or 2013, together with H1, our online agency, we dealt with some attribute modelling issues. We only had models available in each of the analytical tools, but those usually analyse only the data that they collect themselves. We wanted to put everything in one place and above all to monitor, apart from the click interactions, impressions, so that we were able to assign time stamps and values to the display too. The solutions available on the market then were extremely expensive black boxes. We had no clue what was inside them and what mechanism they used to interpret the data. And then H1 recommended Keboola to us.

What where all the data that you began to track using Keboola, and how successful were you?

We were very interested by the fact that apart from the digital marketing, we also integrated SMS campaigns with it. They are usually considered an offline channel and analysed separately. We have combined it all and thus received a more comprehensive picture of the conversion paths. Besides these marketing channels, which ideally lead to an application for credit, we also needed to incorporate CRM data in the solution, because a completed application is only the start of the entire approval process. And, of course, not every conversion ends up with financing. We saw which channels are able to generate a huge number of conversions at a fairly good price, but we didn’t know whether those were the financed conversions, i.e. those that bring real business. And only after incorporating the CRM data did we get a true revelation. We found out that we’re spending a fair amount on channels that generate 10% of all digital applications, yet the transformation into actual business was even below 1%. Only after developing an internal tool in Keboola were we able to combine all this information and transform it into something more effective and logical.

Does Keboola help you today also with other projects  than just with attribute modeling?

We’ve recently been working on developing a Keboola-based backbone platform that would aggregate multiple data sources – not just marketing channels and GA as before, but also NPS measurements, external data such as competitor spend on TV, etc. To be more specific, we’re presently using Keboola to analyse data from Analights, Collabim, our CRM, Sklik + AdWords, and Doubleclick. We’re currently working on a solution to analyse text using Geneea, because we need to use better the feedback forms completed by our clients. Overall, we want to aggregate more data sources in it given that we’re part of a large group and have a lot of data that that is anonymized and aggregated. Such data can be considered in different contexts. It’s also referred to as Big Data, which has almost become a dirty word nowadays, so I prefer to refer to it as data enrichment.

Whether we call it Big Data, data enrichment or anything else, how did you handle these aspects of your work before?

Most of the analytical work is usually done in Excel files, or better still in Tableau, but in all cases the work involves one offline data set. It’s very difficult to obtain live data, or at least data updated daily in batch form. For some major KPIs, you need enough inputs so that the results are worth considering. The only way you can do this today is to go around five or six different departments, but you’ll never get the inputs even within a week. Such an approach is, of course, frustrating, and nobody wants to do it that way. We need to eliminate this fragmentation and instead get everything into Keboola, so that our reports and analyses can be generated and carried out in just one click.

You mentioned that you have a lot of data in your group; do you receive any tools from the BNP headquarters for working with your data?

Each of the regions within our group handles this agenda on its own and, I would even go as far to say that we as Central and Eastern Europe are one of the most advanced markets. At that time, our Keboola-based attribute model was quite revolutionary; we were the first ones who could monitor users from the minute they saw a banner all the way to financing. This pioneering approach earned us quite a reputation in the BNP group. Of course, there are also group-wide corporate tools that can provide in-house data enrichment, yet at the moment these are too expensive for us and rather difficult to implement. Such systems also use up a lot of our IT resources. With Keboola, we know that we can pay its developers and that the solution implemented will not meany any work for  our IT staff.

What is your definition of Keboola? How would you assess your partnership with it?

For me, it’s an aggregator of various data sources – I call it an “external data warehouse”. A storage place where I should put all interesting sources of data that are worth considering and perhaps working with further. That’s my definition of Keboola based on how I use it today.

Overall, I would assess the work with Keboola very positively. For example, I appreciate the quick servicing, which came in handy mainly during various outages or failures – we always had the information ready.

Thanks for your time. We wish you every success in your ongoing and future projects.

‍