Automated insights via Slack at Marqeta

A large focus of the centralized analytics team at Marqeta is providing increasingly sophisticated analytics and insights to foster a data driven culture. It goes without saying that when decision makers in marketing, customer success, product, and operations, are empowered with relevant insights, they can work smarter and faster. One challenge of achieving this goal has been finding flexible and user friendly formats to present insights in an automated manner. How do you incorporate automated analytics into a natural work flow and achieve human-centered automation?

It’s easy to share data in canned reports, spreadsheets, and interactive dashboards. But like many things in life, the easy way out is rarely the best. Reports leave you waiting to copy and paste the numbers out of a pdf. Spreadsheets leave you lost in email at the mercy of drafts getting “lost in translation.” Dashboards often have attractive visualizations, but the static setups can be limiting and they can take forever to update. It’s time for something better!

Of all the latest products on the market designed to provide access to insights and analytics, we are singing the praises of Slack. Marqeta is the first of many things, but we admit we are not the first to love Slack; a quick search for slack integration or slack bot returns a myriad of examples. At the risk of sounding like another set of “fanboys,” we wanted to highlight a few ways that Slack is helping us provide data insights across teams at Marqeta.

Incoming webhooks to dedicated channels

Using Slack, the DiVA (Data, Insights Visualization & Analytics) team can deliver internal exception alerts and results monitoring via incoming webhooks. These insights could be delivered via email and in some cases still are. However, alerts through Slack result in an immediate dedicated channel, rather than an unwieldy email chain. Slack channels encourage asynchronous communication that is more fluid than email, and less intrusive and more available than in-person conversations. Dedicated channels serve as logs of historical conversations. We can also and they provide security because we can control access by whitelisting secure channels for specific data to be shared. Slack allows us the flexibility to increase analytical sophistication while keeping the point of contact consistent and familiar.

Outgoing webhooks for callable alerts

Once a new alert is configured, we can set up outgoing webhooks to make the alert callable through a bot. The bot enables on-demand access to insights and eases the burden of common, yet ad-hoc, analytical requests. In order to make the requests as human-centered as possible, we leverage some natural language algorithms to translate the request to a data query or underlying analytics procedure. Effectively, the Slack bot enables a non-technical user to query without SQL. The bot is only available in certain channels which again has the added feature of a reasonable log. We have learned a lot just watching which requests are called most often. We were surprised to find, there is a wide range of people, from more technical engineers to less technical marketing savants, interested in quick insights.

Currently, Marqeta only uses Slack internally for alerts and analytics. However, Slack fits well into our long term goal of providing human-centered automation with the help of guided analytics. Some actions from analytics can be fully automated, but there is a lot that will still require human interpretation for appropriate action. In the future, we expect that Slack and products like it will help us provide data insights for our customers as well. For now, empowering our internal customers with automated insights has been illuminating for people across the organization and has helped the DiVA team move one step closer to the main goal of a centralized analytics group: provide relevant insights to foster a data driven culture.

Dave Herberich leads a team of incredibly talented individuals tasked with solving data-related problems, answering questions through analytics, and creating scalable data science solutions using modern technology, programing languages and software. He has more than eight years of experience building and leading teams focused on leveraging data to solve problems with private companies. Dave received a BS from Tufts University and a PhD from the University of Maryland.