The line between opportunity and risk can be thin, especially when it comes to data. It’s an invaluable resource that creates significant competitive advantages and distinct ethical challenges. How can a company collect the data it needs to strategically drive business forward while protecting itself and its customers? Here are the five steps you should take to build an ethical framework for collecting and using data within your company.

The first question you should ask is whether or not you need to collect data in the first place. The answer is likely obvious; you know the needs of your business. But asking this question will help you hone in on exactly the type of data to collect and its purpose. Too often, companies are tempted to collect too much and figure out their exact needs later, and that introduces ethical pitfalls and bad practices.

Discussions around collecting or using user data should begin and end with privacy. Work with your legal and security departments, create a code of conduct and a set of expectations surrounding privacy, even when not legally required. Build a “privacy-first” mentality directly into your culture and work hard to audit your practices to ensure your team is in compliance.

Depending on where you are operating, there may be strict laws or governing rules about data collection, and the rights of individuals to know about or control their data. Ensuring you are compliant with all the laws, including GDPR, will not only protect you and your customers, but it will also help you avoid many situations that might otherwise be ethically murky.

Sometimes it might seem that data science and machine learning are smarter than they really are. If you plan to use collected data to make forecasts, perform analyses, or build predictive models—especially in a way that might impact your users—spend the time and energy into establishing proper data collection practices. Data models can only be as good as the data they use, so if you accidentally encode culture biases into your data, you unwittingly encode it into your models as well. 

Work with your data science team to understand the implications of your data collection on its eventual use, and find potential cultural and institutional biases prior to rolling out new products or features that rely on the data you’ve collected. You shouldn’t expect perfection here—or to magically erase all bias—but you can ensure you’ve done everything you can to limit the amount of it you let seep into your product.

It may not always be obvious or practical to predict the actual impact your company’s use of data will have on real people, but you should try. When you roll out a new feature or product, or when you make a decision based on the data you’ve collected, how will it affect a human being down the line? What kinds of problems could you introduce to an individual user if their data was mishandled or misused? What if the data was used to make predictions about them that had a tangible impact on their lives? Consider hiring external ethical auditors to come and review the features that rely on data as well as your team’s data practices. 

It is likely that, as a society, we will come to better agreements and practices surrounding data as time moves forward and global leaders work hard to make sensible and enforceable guidelines. Until then, work with an overly-powerful ethical compass, and hold both yourself and your team accountable for the impact your data practices have on your users, your employees, and the world.