By clicking “Accept”, you agree to the storing of cookies on your device to enhance site navigation, analyze site usage, and assist in our marketing efforts. View our Privacy Policy for more information.
November 13, 2023

Nextbite x Sifflet

Managing data quality at scale in the restaurant services industry with Sifflet

Nextbite is an all-in-one virtual restaurant solution for existing restaurants and kitchens that pairs online ordering/delivery management systems with a selection of highly visible delivery brands. 

At Nextbite the Data Engineering team, in partnership with the broader Data Analytics and Data Science team, are responsible for the company’s data platform and data assets which includes first-party data as well as third-party data coming from external partners.

Nextbite’s data stack includes Fivetran for ingestion, Snowflake for data warehousing and processing, dbt for data transformation, Tableau and Sigma for BI, Hightouch for Data Activation, Castor for Data Governance and Slack for alerting and notifications.

The challenges

Nextbite partners with many different services in the restaurant industry and relies heavily on external data sources. Consequently, for Nextbite’s Data Engineering team, ensuring all the data is being consolidated while guaranteeing its quality is an ongoing challenge and they came to Sifflet to address that.

Nextbite's Data Stack

Sifflet in action

1. Deploying the checks 

Director of Data Engineering Ross Serven shared that thanks to Sifflet Auto-coverage feature, the data team at Nextbite was able to deploy around 100 data quality checks that are essential to monitoring the quality of their assets, in a matter of minutes, an effort that would have required weeks if not months otherwise. 

2. Leveraging ML for monitoring

Restaurant industry data is not always predictable, it can get very cyclical throughout the week, so it is not possible to expect the same numbers every day. Prior to using Sifflet, Nextbite’s data team had put some checks in place to ensure that they were receiving the right amount of data. However, after looking at the standard deviation of their data on their own, they realized that they needed a deeper look at expected trends. This is where Sifflet’s machine learning came into play by helping them train a model with expectations on their own data, which in turn allowed the team to focus on more impactful tasks. As Ross put it:

“Data teams should be more focused on driving value for the business, not stuck monitoring data feeds”

3. Incident management 

Before adopting Sifflet, the standard incident management process was a combination of dbt tests and a dedicated Slack channel. The team then would use the channel to manage the incident response. For Ross, this is fine at first. After some time, however, you need to start scrolling through slack messages to understand whether the issue was actually fixed, becoming an unnecessarily complex and time-consuming process. Sifflet Incident Management Reporting has proven to be extremely useful to the team as it allowed them to take the next step and define clear ownership. For Ross, this is an essential feature and a key differentiating factor for Sifflet.

Sifflet provided our team with the context they need

“We chose Sifflet for its wide offering. We had checked out other vendors in the space, and they have all data quality rules, but what we need is the next step. Being able to know what to do once that rule fails and ensure it’s resolved”

Results 

  • Nextbite deployed ~80 advanced monitoring rules with Sifflet. This saved them several weeks worth of custom-developing SQL/dbt based rules. A non-ML completeness rule that had been previously developed by the Nextbite team had taken 40+ hours of work between developing and testing. 
  • The team at Nextbite is saving ~3 hours per week on average by not having to manually track down data errors in their failed dbt/SQL tests. This is thanks to Sifflet’s overview of the historical data as well as the failing data points in real time.
  • Sifflet is identifying 1-2 data provider issues a week on average. These alerts arrive on the day of the issue, rather than days later by ad-hoc checks or by the business partners. Nextbite is now able to resolve these issues before they can impact reporting. 

“Let’s put Sifflet on that!”

Initially, Sifflet was used by the Data Engineering team at Nextbite. However, very quickly, the word spread in the broader Data Analytics team - which in turn also started setting up rules to monitor key business metrics. 

Conclusion

After facing data reliability issues on the data coming from external sources, the all-in-one virtual restaurant solution Nextbite was able to apply ~80 advanced monitoring rules with Sifflet, saving them weeks of work. On top of this, Nextbite is now saving ~3 hours per week because the team does not have to manually track down errors. And finally, by identifying, on average 1-2 data provider issues a week and thanks to Sifflet’s data lineage, the data engineering team at Nextbite is now able to resolve data issues before they become business issues.

The Modern Data Stack is not complete with an overseeing observability layer to monitor data quality throughout the entire data journey. Do you want to learn more about Sifflet’s Full Data Stack Observability approach? Would you like to see it applied to your specific use case? Book a demo or get in touch for a three-week free trial!