Meeting the company’s goals is the responsibility of everyone in the organization, so even those who aren’t directly influencing the cost or the revenue side should know what those goals are. “No matter who you are, you are directly or indirectly affecting key company metrics,” said Vanessa Lam, Business Intelligence Manager for Optoro. Lam shared with DATAVERSITY® Enterprise Data World Conference participants her experience creating and building Optoro’s Business Intelligence (BI) team during her presentation titled A Journey Towards Building Trust and Alignment in Data and Metrics.
Optoro’s mission is to make retail more sustainable by eliminating all waste from returned merchandise. They do that by offering a software platform that uses machine learning and data-driven decision-making to help retailers find the best use for their returned goods.
Lam started with Optoro as a rotational associate, working six months in each or four departments: Client Success, Sustainability, Data, and Business Operations. During this two-year period, she developed a deep understanding of what different departments were doing across the organization, how each department thought about data, their priorities, and what their pain points were. She discovered that three conditions led to Optoro’s Business Intelligence issues: a fear of data, inconsistent use of vocabulary and metrics, and mistrust of data.
Data Chaos
They had a commitment to democratized Data Management, a practice that puts data in the hands of the people closest to the problem, allowing them solve it for themselves. The resulting expectation is that people will use data to make decisions, that they will back up their arguments with data, and that they will monitor the success of their products through data. “It makes us agile, it makes us faster, it also means that they’re able to draw the insights that they need for the people who need it.” This process works in theory, but in practice, Lam said, it wasn’t working for Optoro because users were afraid to use the data they had access to. “If users are afraid of data, they won’t use it to make decisions,” leading to fewer data-driven insights.
The reasons users were uncomfortable using data were valid. Vocabulary and naming conventions were inconsistent and not universally agreed upon. Lam shared Shakespeare’s quote:
“‘What’s in a name? That which we call a rose would, by any other name, smell as sweet,’” countering with: “I am here to tell you today: that is totally false. We cannot do this with data.”
She talked about a meeting where the VP of e-commerce reported $1.5 million in revenue, which was .15 percent over target and cause for celebration. In the same meeting, the VP of sales reported $1 million in revenue, which was .35 million under target, and cause for concern. Both results were presumably based on the same figures.
Similar metrics with different underlying calculations often had different benchmarks as well. Without a consistent baseline, performance was not clearly understood. Communications took longer and meetings were not productive. Other inconsistencies due to outdated figures and differing interpretations for the same data led to a lack of confidence in Optoro’s data. “Even if people are confident with using data, and they are using the right data, if they do not trust it, they will not use it.”
Building a Team: Building Trust
The credibility she established with multiple departments as a rotational associate allowed her to look into existing metrics across the organization and question the logic behind them:
“That was a big step in our organization because we’ve had these metrics for three or four years and they were all just assumed to be correct.”
As she went deeper into Optoro’s metrics, she found flaws in logic and areas where measurements were interpreted in different ways. She was trusted because she had been an integral part of the business side, and understood frustrations business users had with their data. She considers this trust a key factor in her success.
Building a Team: Showing Value
She started with the three overarching problems she wanted to solve and worked backward to actions that could address them. Lam needed to ensure that her proposal, which included long-term governance processes, would be noticed in Optoro’s fast-paced environment where projects with a short time-to-value get the most attention.
“I had to show the connection between Data Governance and business value, both in the short term and in the long term.” Showing directly what the gains would be — such as improving communications and enabling teams to meet goals because they were measuring the right things — was extremely important to pitching the job, she said.
Building a Team: Getting Buy-in
Lam’s existing credibility with multiple departments also helped with her conversations with stakeholders. She was able to focus on the benefit to the business and, at the same time, make clear to users that they would still be empowered to access and use data, but that there would be enough structure around it so that metrics would be more accurate and trustworthy.
Building a Team: Finding Flexibility and Oversight
Optoro’s commitment to a democratized analytics model allowed quick access to data by many users from many points. The flexibility of this model, however, had its drawbacks. In Optoro’s case, it led to multiple versions of the truth that emerged when incorrect assumptions were made, and as more users built on those incorrect assumptions, data was no longer trusted.
Lam considered their current model against two other models: centralized and federated analytics. Centralized analytics is the practice of having all data requests answered by one department, but she decided it wasn’t a good fit. “That was too much governance, too much time.” They wanted a model that retained flexibility, but also had strong oversight. “So, we landed on federatedanalytics: the practice of having several data representatives whose collective responsibility it is to enforce Data Governance.”
Building a Team: Implementation
Each of Optoro’s nine departments designated a representative who became the person users could go to for answers. Representatives met monthly as a group to discuss data issues and best practices:
“Individuals could still have access to data. They were still empowered and could get quick answers, but the representative was responsible for ensuring that they were using data in the correct way.”
Because representatives were members of the user’s own team, they understood the context of user questions and were able to provide better answers than a central body would have been.
Creating a Single Source of Truth
Lam’s plan to create a single source of truth had three components: consolidating datasets, pushing data upstream, and universal dashboards. Her team worked to combine inventory and sales data so that they could be analyzed without manual joins, and all of analytical fields became available across both sets:
“This consolidation of datasets allowed analysts to more easily understand where they were going for their data because it was more centralized.”
Another way they worked to create a single source of truth was by giving analysts clear, consistent guidelines for when to push data upstream by putting their calculations into BI or ETL layers, rather than in their own workbooks or at the source.
Implementing universal dashboards required changing assumptions and habits, but provided consolidated, consistent figures that could be used from a variety of different perspectives:
“Convincing people to get rid of their existing dashboards was really difficult because people had a way of looking at their metrics and had a set of assumptions that they already had made that they didn’t want to change,” but this process “really addressed our problem of inconsistent use of vocabulary and metrics.”
A Data Education Program: New Hires
New analysts coming into the organization were required to go through a four-hour training session with mandatory homework, “before they have a chance to develop any new bad habits.” The training teaches them how to use the tool, but more importantly, they learn about Optoro’s data, data model, and best practices. Lam stresses that what matters more is that they know when to ask a question instead of guessing and making assumptions about the meaning of data.
A Data Education Program: Existing Analysts
Lam put a comprehensive communication process into place so that as soon as the BI team knew about any changes or issues, they could communicate them to stakeholders. New developments or releases were accompanied by release notes, and outages or issues were quickly and widely communicated. “This really gave people the sense that we were looking out for them and we were on top of the issues,” she said. Formerly, a small issue would lead to a distrust of the whole dataset, but this comprehensive communication process gave users confidence that they could bring data concerns to the BI team and that they would be addressed.
A Data Education Program: Explain Key Company Metrics
Part of the education process was also explaining key company metrics to “anyone who would listen,” because everyone in the company, no matter what position they hold, has a responsibility for meeting the company’s goals, she said. They invited everyone to a lunchtime presentation where they included high-level reasoning for their metrics along with a more technical, granular explanation. The higher-level explanation was for stakeholders, and the more detailed information was for analysts, but the overall effect on all listeners was to illustrate that a lot of thought had gone into the assumptions and calculations the BI team had produced.
This had the added benefit of curbing the “creative” use of metrics that had been done in the past, so users couldn’t “just make things up, because if they did, they would have to also go through all of that thought process that we did.”
A Data Education Program: Increase Data Team Visibility
Formerly, data issues were addressed through an ad hoc system using a variety of channels from emails, Slack, and random in-person communication. They decided to streamline and formalize the process so data issues were handled the same way as technical tickets. “We asked people to put in data tickets, exactly the same way that they would put in tickets if they saw a problem with OptiTurn, our software.” This provided a standard way for users to ask about data, request changes, or solve problems, and it allowed users to track the progress of submitted issues and requests.
The ticketing system was an important part of changing user perceptions of the value of the program, as well as showing leadership that “this data is just as important as the software.” By positioning themselves alongside the company’s software team, the BI team was able to convince leadership that resources dedicated to BI were just as important to the company’s success.
A Data Education Program: Hold Office Hours
Lam implemented weekly topic-focused office hours for discussions or presentations, with the topics chosen by someone in the organization. “The great thing about this was that it got people talking. It got people in the same room talking about data.” She also designates an hour once or twice daily when she encourages people to drop in to her office and ask questions. Because these office hours are informal and happen daily, people know there is time dedicated to their concerns, that they are a priority, and that they aren’t interrupting Lam’s work. “I cannot stress enough how important this was.”
Where Are They Now?
In the 18 months since the BI team was formed, Lam says the data culture has shifted significantly. “You can’t tell people to change their culture. It has to come organically.” And it has: as a result of that shift, she will soon be doing a month-long training, this time with product analysts, to instill good data practices in them as the program expands into a new area.
Check out Enterprise Data World at www.enterprisedataworld.com
Here is the video of the Enterprise Data World Presentation:
Image used under license from Shutterstock.com