A quick window into a restaurant forecasting model

A quick window into a restaurant forecasting model

5out.io forecasts sales for restaurants. With 5out's insights, restaurants know how to staff their restaurant and how to shop. They can plan better, run more efficiently, and make more money.


Internally, 5out needs to peek under the hood of their forecasting model. They do this for a few reasons:

  • Sales - Branden, their CEO, needs to understand how 5out's models are performing relative to their competitors.

  • Customer service - Their customer service reps need to investigate forecasts that customers have questions about.

  • Engineering - Mike, their CTO, needs to peek under the hood to understand the model's strengths and weaknesses and to prioritize engineering work.

Possible solutions

Engineer responds manuallyInefficient, expensive
Custom built dashboardExpensive, overkill
BI (Tableau, Looker)Can't run an AWS Lambda
Drag and drop (Retool)Overkill, no history of old queries
FlankSimple ✅ Cheap ✅ History ✅

How does it work?

Rob Churchill is 5out's Chief Data Scientist. Rob has written all the backend code to peek under the hood, but his code isn't accessible to anyone else.

Enter Flank. Flank autogenerated a webpage for each one of Rob's "functions" that peeked under the hood in a specific way. Rob shared that functionality with Mike Marian, 5out's CTO. Mike shared that functionality with the rest of the team. Windows constructed, in a matter of minutes.

In their words

In the past week, Rob has alternatively described Flank as "crack" (that was a compliment) and "infuriating" (the idle logout timer was too short). It's not hard to figure out how Rob feels about something.

From Mike Marian, their CTO:

I'm impressed with how much time Flank saves us on a daily basis, and how intuitive it is to use once it is configured. We use it to monitor the performance of our forecasting modules, and we've made it so easy that even non-technical members of our support and executive staff can get the information they need themselves, without having to pull in a technical team member to write a series of queries and calculations to get the data. The on-demand nature of Flank also allows us to use it to troubleshoot issues with specific customers, often proactively, and often with our customer support team being able to triage issues and offer solutions without needing to undertake inefficient manual investigations.

If you're thinking about internal tooling, I'd love to help! Even if it's by recommending a tool other than Flank. Drop me a note. angus@flank.cloud