Fivetran

Fivetran "läuft einfach". Ob die Daten aktuell sind ist kein Diskussionsthema mehr. Dank den OAuth-Connectoren kann man innerhalb von wenigen Minuten einen neuen Connector im UI hinzufügen. Jede Person die schon einmal selbst API-Konnektoren geschrieben und gewartet (!) hat wird die Zuverlässigkeit zu schätzen wissen.

Manchmal gibt es neue Tabellen bei einer Quelle die dann per-default aktiviert sind, obwohl man sie nicht braucht. Das ist dann allerdings auch in wenigen Minuten wieder deaktiviert. Jedoch sollte man - wie bei jedem Produkt - schon aktiv die monatlichen Kosten im Blick haben und ungenutzte Tabellen/Quellen ausschalten um nicht böse Überraschungen zu erleben.

die Entwicklung und Wartung von API-Konnektoren wird zur "commodity". Neuer Marketing-Kanal? Wenn Fivetran den Connector hat sind die ersten Daten in den nächsten 60 Minuten im Data Warehouse um weiterverarbeitet zu werden.

Durch Fivetran ist es uns möglich alle Daten zu sammeln und zu aggregieren. Man kann seinen Account ganz easy mit fievtran verbinden und nur durch wenige Klicks kann man beispielsweise einen Connect von seinem FB Account zu fivetran herstellen.

Leider kommt es öfter mal zu Fehlermeldungen, die nicht auf den ersten Blick zuzordnen sind. Verbindungen brechen ab und Daten werden somit nicht übermittelt. Zusätzlich können die Daten von Facbeook beispielsweise nur über ein Profil verknüpft werden und nicht anhand einer API ohne, dass ein Profil dazwischengeschaltet werden muss.

Wir nutzen fivetran für unser Performance Marketing uns so können wir notwendigen Daten jedes einzelnen Kanals in ein Dashboard pushen. Durch die einfache Anbindung können Anpassungen wie neue Konten oder Ähnliches schnellstmöglich umgesetzt werden.

Fivetran is very simple to use and robust. It is definitely a tool that you can rely on and you know it will get the job done. New functionalities are being added regularly (e.g. transformations and dbt integration) and the company is making big steps in different ETL/ELT relevant directions.

It would be great to have the possibility to view a sample of the extracts and choose for a specific timeframe to refetch the data.

With Fivetran one can be sure that their ETL/ELT jobs will work without interruptions. The dev team is taking care of any changes in the APIs and the support team is super responsive and always helpful. Other than that, the tool is very intuitive and easy to use, and there is a very wide range of connectors.

The ease of use of Fivetran is what I like about it the most. I used to develop custom connectors in Python so I know how difficult the process can be, however, with Fivetran you can ingest a new data source into the data warehouse within minutes by just creating a connector and configuring a few fields in the user interface (service account email, account password, account keys etc). I also like how there is a Fivetran API where you can programmatically edit multiple connectors using a few lines of Python. This is especially useful when you need to edit the schedules of multiple connectors or enable/disable a lot of connectors at once.

The only main downside to Fivetran is that not all data sources have connectors, so you need to still implement custom Python connectors for sources that Fivetran do not support (AWIN, Adjust etc.). The other thing is that I do not really ever use their sql transformation function as there are better tools to do so out there (dbt) so I find it quite obsolete. Although I do understand why they include it as some other companies might find it useful.

The main problem that Fivetran solves is it simplfies super complex ELT pipelines using Talend or Python scripts. With Fivetran, ELT is so much easier as it takes care of a lot of the big data ingestion jobs and the connector is maintained by Fivetran so all updates to the API is handled by Fivetran. So all you need to do is maintain a few number of custom connectors not supported by Fivetran, which really makes ELT a lot more simple.

It is a tool that doesn't take more than a day to learn how to use yet very adaptable. Very easy to configure and no coding required. It may be one of the easiest ETL (or ELT in this case) tools I have ever used. It works great with cloud datawarehouses and I never had an issue in any of my connectors so far.

There are still some properties that could be added to make it more flexible and reduce workarounds due to tool dependencies. As it is not a complex ETL tool to use, Fivetran automatise several processes and in some cases, it is actually not very useful. A very simple example is that Fivetran specifies the data type based on the data from the source. In the cases that you need to get an error (like having a string instead of integer), it changes the type in the target and runs without any errors. Which might be very useful in some cases but I wish it was configurable.

Extract, load and transform data from a huge variety of sources into data warehouse. It works great with cloud datawarehouses and I never had an issue in any of my connectors so far. No delays or any issues related to data. For anyone using it day-to-day, it is very time and cost efficient.