Implementing Data Quality Systems

How to build a great data quality system

Insight from Avenue x Athena Case Study

Do you have a ton of moving pieces to keep track of?

So many moving pieces that data issues in your operations have caused you to lose thousands? (I definitely have)

You're not alone.

Athena was facing this exact problem. Athena helps businesses by placing thousands of virtual assistants (VA's) a year.

Athena manages onboarding, account management, payroll, and more for those VA's.

Data accuracy became a challenge because of the amount of data coming into their system. Clients not billed or employees not paid on time were because of these data issues. 

A few other issues included:

• Multiple deals with incorrect pricing = thousands of dollars at risk

• Blank VA address = missing payments = churn

• Inaccurate VA employment status = customer paired incorrectly with a VA (or former VA) = task delays = poor customer experience

So what did Athena do about these operational data problems?

Enter Peter.

Peter Dimov is the director of data science at Athena. He tried building a janky QA system but knew this wouldn't scale.

When Peter sought out a quality control engine for Athena, he had a few requirements in mind.

His must-haves included:

• Plug-and-Play integrations: connect existing data sources in minutes (Snowflake, HubSpot, Coupa, Greenhouse, Zendesk).

• Easy to use: non-technical members can set up monitors without knowing SQL

• High security: the ability to manage who has view/edit access to underlying data with role based permissions

• Reporting: Track performance across individuals and teams (SLA’s, efficiency, accuracy). 

• Smart task assignment: set custom working hours and auto-assign feature. 

• Communication syncing: don’t want to force people to learn another tool. Let operators stay in existing tools like email, Slack and Zendesk (unlike Zapier).

• Affordable - as a start-up you dont have the resources to hire an engineer to build a reliable internal tool or staff a QA team. 

• Automated: the ability to auto-resolve certain issues that shouldn’t require human input to fix

When sourcing or building the right QA system, just like Peter, you must scope out what works best for your system.

Maybe the above works as a guide to building your own system.

Maybe you need to build a whole new list.

Just don't spend your time on data inaccuracies. It'll kill you.