You don't trust my data? It was a stupid project anyway! How to build more trust in the analytics process

Many people believe that trust in analytics comes from trust in the data, if you have good data you have good analytics. While good data is definitely a foundational element to good analytics, there are also many other areas where you need to put a great degree of trust in order to be successful.

1. Trust in the project prioritization. Achieving excellence in your planning methodologies is critical for a data team's success. Prioritizing projects with very clear business impact and stakeholders' needs over shiny, more interesting projects is definitely important. This process isn't as clear cut sometimes because stakeholder needs and the measurable business impact can be in conflict at times.

2. Trust in the execution. Setting clear and realistic guidelines of when and what will be delivered. Data people are often overwhelmed with little "can you pull this data, it shouldn't take too long" requests and it can get in the way of delivering on the longer-lasting, more robust solutions. Being able to say "NO" as a data person so you can deliver on the most mission-critical projects will be both empowering for your mental health and improve your overall productivity as well.

3. Trust in the data. Nothing erodes trust in your data team more than publishing bad data. Making sure that you have automated ways of evaluating the quality of data through statistical checks such as distinct counts (e.g. there should be 50 states in the data), week-over-week row count checks (we should have within 5% the amount of data as last week), and null checks (e.g. these columns should never be null)

4. Trust in the model. Having a model that is easy to explain (e.g. linear regression) is oftentimes more valuable than a model that is harder to explain (e.g. random forest). This is because it's much easier for the business to make sense and incorporate the model into their overall thinking. This can be a conflict for data people because we want to use more fancy models because they can be more accurate in some cases.