A Look at Looker
August 18, 2015
A Look at Looker
Score 9 out of 10
Overall Satisfaction with Looker
Looker is being used across all of Lyve. Looker allows our business users to explore the data in our data warehouse and answer business relevant questions. The Data Insights team are the primary users, defining the data model and providing dashboards and reports that provide simple answers for other users but more importantly act as "jumping off points" for other questions.
- Looker is very good at answering "ad-hoc" questions once the data model is defined. A simple example for instance is when we show the number of users added to the system, we start with just the basic number, Looker makes it easy to explore from that starting point and ask other questions like "how many of those users were added in the last X days", or "how many of those users are on android versus ios."
- Looker has a feature called "Persistent Derived Tables" (PDTs) which allows for highly analytical queries which may take a long time to execute against the raw data to be pre-computed and the results quickly available to end users.
- Looker provides a lot of flexibility in how the data model is defined, you can start simply with individual tables from your data warehouse, but you can also define parts of the model based on SQL queries.
- Beyond the technical aspects of the product, the Looker team has been very responsive to questions from my team, as well as to requests for new features (and the occasional discovered bug).
- The evaluation process was very straightforward, we were able to use Looker against our own data and even begin the development of the data model during the evaluation, it gave us a real sense of how we could use for real use cases.
- Looker has had multiple product updates in the few months we've been using the product, they seem to have a good steady cadence of adding new features, improving existing features and addressing existing shortcomings in each new release.
- Looker needs some improvement in the visualization aspects of the product, they are very good at getting the data, but the choices for visualizing that data are somewhat limited, there are workarounds for some of the weaknesses but they tend to be labor intensive and fragile. For example, there is no way to define a custom color palette for use in our charts, we could define a set of colors on an individual graph and tie those colors to individual values, but it's more trouble than it's worth.
- Looker dashboards are an area where the needs often exceed the capabilities of the product. The dashboard layouts are done with a drag-and-drop interface which is not very responsive, and we often wind up with a dashboard that is "good enough" rather than what we were aiming for. This isn't a show stopper for us as the data is all present, but it can be frustrating.
- Looker does not provide a lot of feedback to users when it is processing data or even when there is a problem getting the data (possibly because of user/modeler error), you have to know to look for little spinning circles to see that it is still "thinking", or know that if a dashboard shows "No Data Available" that you have to dig deeper to find out what the actual problem is.
- Looker allows you to explore a large amount of data, this is both good and bad, it's great because you can probably find whatever you're looking for, but on the flip side, it's easy for users to get lost or overwhelmed with the number of choices they are being given.
Looker is much better at allowing users to interact with the data and "explore" to answer their "ad-hoc" questions. Tableau is better at pixel perfect visualizations. On balance, the interactivity is more important for our uses rather than "polished" visualizations.
Looker works very well against Amazon's Redshift. Some of the questions I asked during a reference check with another customer:
* How much data are you storing in Redshift and querying with Looker?
* How large is your Redshift cluster?
* How is the performance of Looker against that cluster?
* How "real time" is the data? Do you add data on an ongoing basis or have more of a "traditional" ETL model where you update your data warehouse once a day?
* How large a team do you have working with Looker? How many people are authoring LookML versus just exploring the data?
* What if anything were you using before Looker?
* What's the one thing about Looker you know now that you wish you knew before making the decision?
Looker Feature Ratings
20 - Product Management, Engineering, Executive, Operations, and Data Insights
- SQL Knowledge
- LookML (internal Looker modeling language)