Analyzing Marketing Data: Behind the Scenes at the ThoughtSpot Implementation (Part I)

One of ThoughtSpot’s new partners in EMEA is Celebrus. They provides their client’s marketers with data about how individuals are interacting with their brands across digital properties (website, mobile apps, social media, etc.). But some of Celebrus’ customers need ad-hoc access to this rich customer data and their traditional BI tools can’t give them that. So we recently started working together on a prototype that would do exactly that. As part of the partnership kick-off I was explaining to their Marketing VP, Katharine Hulls, a little about how ThoughtSpot’s search-driven analytics is helping business people do their own data analysis across terabytes of data.


Now Celebrus eat their own dogfood, and as Marketing VP, Katharine is their number one user. And suddenly it clicked - never mind what their product manager might want for customers, she wanted ThoughtSpot for her own team to get ad-hoc access to all this data too. Katharine told me that she and her staff often need to answer ad hoc questions that come up in the moment, such as a spike or dip in lead flow. These are queries that their existing analytics product doesn’t already include and which the BI team isn’t able to prioritise just for a one-off request for Katharine - precisely because they are one-offs.


But ad-hoc data exploration is exactly where ThoughtSpot shines. And with Katharine fired up, we now had a real-world use case for a proof of concept (PoC) project. So she and I spent the rest of the meeting sketching out use cases and sharing them with the implementation team - one person from each of our tech teams.


Just like any PoC, first we needed a set of data. For this case, we decided to go with click stream data from the company website. The data load and and schema metadata build went through smoothly; but then comes the stage where you discover that the users have various terms that aren’t reflected in the data, either because the database tables and column names are obscure, or because these terms are user jargon for some calculations (lots of those in digital marketing!) So we needed to add those names to the Thoughtspot metadata. Once that was completed, we quickly put together a pinboard of about 15 charts that answered most of Katharine’s questions.


The next step will be to review those with Katharine - and to find out what she really wants for a couple of the more obscure stories.


You can read Katharine’s side of the story here and i’ll post an update when we’ve reached the next stage of the project.