Is my data visualization tool better than yours?

Have you ever tried to find your way through the jungle of data visualization tools available for companies? For data-driven companies, tools can be a large part of the budget and require a tremendous amount of time to implement, so it is important to pick the right one. At Jellysmack, we’re conducting an in-depth analysis to choose the right tool for our needs. In this article, I will be sharing our thought process with you.

There are many similar tools that can help us visualize our data

Why are we changing our tool?

A bit of context

Jellysmack has been using PowerBI (Microsoft) for four years. The techies using it are very satisfied with all of the features. This is a technical tool, a very powerful for those who know how to use it and to code a bit in SQL and DAX (PowerBI language).

However, non-techies (90% of Jellysmack employees) only have access to reports created by my team (composed of Data Analysts and Business Analysts) and sometimes to filters when we grant them permission.

With the hypergrowth of the company (we’ve gone from 300 employees in January to over 1,000 in December) and no real tool to access the data directly, business users rely on us more and more to extract data from the database, make queries, and communicate the results. This is time consuming for everyone involved and doesn’t provide much added value from the data team. Even so, we love doing in-depth analysis and having the time to solve problems based on data.

Hence the change: we are looking for a tool that everyone in the company can use for direct access to our data. It would save business users time, they would have access to more data without middlemen involved, and it would give data experts time to concentrate on higher added value tasks.

Challenges and change management

However, there are challenges that accompany giving employees access to data. What data do we give them access to? Is the data shown on the tool reliable? And more. We want to make sure we give business users actionable data, not just any data.

It would be a real game-changer for Jellysmack employees, both technically, and in terms of business vision:

• Technically, it would mean working on the data pipelines that would feed the tool. The tool is only as good as the data it presents! So, we need to have a very clean database to start from and compute KPIs that make sense for everyone. It would also change the way the data team works. They would have a tool to go to for simple exploration. The time that they save could be used on more in-depth analyses (more Python and stats, yeah!)
• On the business side, it would change the way everyone works.
  Everyone would be able to access the data, well actually, the information. They would be able to see the impact of their work almost instantly by following real-time performance KPIs, for example.
  Another benefit: It would be an opportunity for business users to discover just how complex data can be. For example, they would learn to call data quality into question, what it means to define a metric or a KPI. It would be an opportunity to show users the importance of governance in an organization. We could improve everyone’s data literacy. 🤓

1st stake : what do we want to evaluate ? 🕵️‍♀️

Choosing the tool best suited for our needs is far from trivial. We’ve decided to test two different tools. But, with 1000 different employees in the company, there are 1000 ways of using them!

For example:

• Business users want the tool to be user-friendly, so they don’t have to ask us for help every time they need access to data.
• Business and Data Analysts want the tool to have lots of features, technically and graphically, and to both present and aggregate data.
• Data Scientists want to search through and understand the data quickly via Data Visualization before creating their models.
• Our Managers want the tool to provide different levels of access for different team members to ensure confidentiality.
• Product teams, who build the tools our colleagues use on a daily basis, want to easily embed the dashboards into their own tools.
• The Finance team wants to make sure we don’t spend an unreasonable amount of money on the tool.

To reconcile these differences, our Lead Business Analyst (who is in charge of the project), is creating a survey for all potential users. For every aspect of the tool, respondents will give a score out of 5, rating how good the tool is at meeting their needs.

She will collect the data and make a data-driven decision about the tool, without letting her own personal biases take over.

2nd stake: get everyone on board

… and use our collective intelligence

To get the opinion of all of our stakeholders, the two Proofs-of-Concept (POCs) were made separately, and they both had 3 steps:

1. During the first week, the data engineering team loaded data into the tool, and tested the features and user-friendliness. The rest of the employees were not involved and kept on working on their usual tasks.
2. During the second week, the Business and Data Analysts took over. They started testing anything they could with Jellysmack data and conducted proper analyses. Vague guidelines were put in place, but nothing too specific. The idea was to allow for a maximum of creativity.
3. During the third week, they presented their dashboards to top management business users, so they could look around and give them their opinions.

All three steps were equally important. The whole company has to be on board for the tool to be implemented well. The old tool and old processes were in place for 4 years. This makes the transition more difficult.

How did it go?

Looker (acquired by Google)

The main feature that sets Looker apart is LookML, the Looker language that allows us to hide the complexity of data and provide a semantic layer. Otherwise, like in every discussion with C-levels, one may ask “But, what do you mean by retention rate?”
KPIs and metrics are discussed with the business. They are defined and calculated in LookML, directly via the interface of Looker. LookML also allows us to create “explores” which are a kind of data mart. The “explores” and aggregation are then put into production with a script that can be versioned on github (by users who have the Developer role on Looker).
Then, when a metric is presented, you can be sure that everyone in the company understands what it means. It helps us save time and avoid confusion. Business users aren’t able to build metrics that haven’t been approved beforehand. Metrics are also quite easy to display on “looks” or dashboards.

Unfortunately, there is a major disadvantage for us at Jellysmack. The tool is not adapted to Data and Business Analysts. We are a diverse team with different levels of coding knowledge. Not everyone has the knowledge to use the features necessary to dive into the details. Plus, the reporting features are not as extensive as what we were used to with PowerBi.

ThoughtSpot

• The main feature that sets ThoughtSpot apart is their search bar. You can type anything in natural language, just like on Google (ThoughtSpot was created by former Google employees.), and it automatically creates a graph. For example, you can type “Number of video views per month on Facebook in 2021,” and a graph will appear. Simple as that.
• They’ve also added an AI brick in the software, which shows us important data we should take into account to take our analysis further. The brick is called Spot IQ. For example, if we want to investigate a sudden increase in revenue on a channel in a given month, we simply click on the curve and launch Spot IQ. It gives us information about the probable cause of the increase. In one case, it pointed us to the 4 videos that were responsible.
• Like Looker, ThoughtSpot is quite limited in terms of graphics for Analysts when compared to PowerBI, which allows personalized pixel-perfect dashboards. We can work faster with Looker and Thoughtspot than with PowerBI, but we can’t go as far. But, for business users who don’t need to go into that much detail, both tools are life-changing because they have full access to the data.

Under the hood

A point about database administration and performances

We noticed a real difference between Thoughtspot and Looker in the storage of the temporary table/views: In Looker, we had access to them and could reuse them once they had been created. In Thoughtspot, we didn’t, which made it kind of like a black hole when it came to identifying what affects performance.
For that reason, performance monitoring was a lot easier with Looker than with Thoughtspot.

With Thoughtspot, the creation of metrics and KPIs must be done beforehand in a database to be accessible to users. On Looker, the LookML allows you to take raw data from the database (You still need to have a centralized and organized data warehouse plugged into it.) and create a KPI on top of the data warehouse. Because it’s done on the interface of LookML, it is accessible to anyone who learns LookML. It would save Data and Analytics Engineers time (which, as you may know, is scarce these days).

The choice was not easy for business users either

When we asked business users for their opinions, we were surprised by one of their answers. They trusted the data more on Looker than on Thoughtspot. It is interesting to take into consideration. Even though we put quality controls and data governance processes in place, they like to check the data for themselves. They want to be able to put it on a table when they see a graphic. Looker seems to do a better job at letting anyone check and play with the data.

On the other hand, they were very impressed, as were we, by the search bar on Thoughtspot, which I described earlier.

Overall, they were very satisfied with both tools. They appreciated the design, the UI/UX, and other features such as the ability to set up automated alerts when a KPI reaches a certain threshold, for example.

Conclusion

We haven’t chosen our new tool yet, but we know it will change the way that everyone works at Jellysmack. We are excited to find out which tool will be implemented and how it will drive the evolution of our company culture. One of our five core values is precision, and this next step will definitely allow us to be more precise.

The biggest challenge now is getting everyone on board and managing change, data acculturation, data literacy training, and a lot of expressions with the word “data” in them.