Overview of Yandex DataLens

Let’s take our minds off of the project on receipt data collection for a while. We will speak about the project’s following steps a bit later.

Today we’ll be discussing a new service from Yandex – DataLens (the access to demo was kindly provided to me by my great friend Vasiliy Ozerov and the team Fevlake / Rebrain). Currently, the service is in Preview mode and is, in essence, a cloud BI. The main shtick of the service is that it can easily and handy work with clickhouse (Yandex Clickhouse).

Connection of data sources

Let’s review the major things: connection of a data source and dataset setting.
The selection of DBMS is not vast, nevertheless some main things are present. For the purpose of our testing, let’s take MySQL.

Selection of data sources DataLens

On the basis of the connection created, it is suggested to create a dataset:

Interface of dataset settings, definition of measurements and metrics

On this stage it’s defined which table’s attributes are becoming measurements, and which are turning into metrics. You can choose data aggregation type for the metrics.
Unfortunately, I didn’t manage to discover how it’s possible to state several interconnected tables (for example, attach a handbook for measurements) instead of a single table. Perhaps, on this stage developers suggest us to solve this issue by creating of required view.

Data visualization

Regarding the interface itself – everything is pretty easy and handy. It reminds of a cloud version of Tableau. If comparing to Redash, which is most frequently used in conjunction with Clickhouse, the opportunities of visualization are simply staggering.
Even pivot tables, in which one can use Measure Names as columns’ names are worth something:

Setting of pivot tables in DataLens

Obviously, there is an opportunity to make also basic charts in DataLens from Yandex:

Construction of a linear chart in DataLens

There are also area charts:

Construction of an area chart in DataLens

However, I didn’t manage to find out how data classification by months / quarters / weeks is carried out. According to an example of data, available in the demo version, developers are still solving this issue by creating additional attributes (DayMonth, DayWeek, etc).

Dashboards

For now, interface of dashboard blocks’ creation looks bulky, and interface windows are not always comprehensive. Here is, for instance, a window, allowing to state a parameter:

Not really apparent setting window for dashboard parameters

However, in the gallery of examples we can see highly functional and convenient dashboards with selectors, tabs and parameters:

An example of a working dashboard with parameters and tabs in DataLens

Looking forward to fixing of interface shortcomings, improving of Datalens and preparing to use it together with Clickhouse!

Share
Send
Pin
 33   12 mon   analysis   business intelligence   visualisation
Popular