Operational efficiency gains depend on establishing data quality. But what after that? Read the second part to understand the data engineering and analytics metrics.
Once an organization has ensured data quality, data teams – comprising several sub-teams –need to be evaluated according to different metrics. An established way of doing this would be to analyze the two major sub-teams: engineering and analytics. Each sub-team is set to have different goals and impacts, and hence needs to be evaluated accordingly.
In an interesting article measuring the return-on-investment (ROI) of data work, Mikkel Dengsøe, the head of Data Science at Monzo Bank, introduced an interesting system of differentiating the systems and KPI sections of the workforce.
Data engineering metrics
The goal of the data engineering team is to build a reliable infrastructure in order to ensure Analytics have access to trusted, quality data. ESSEC Business School’s Louise de Leyritz opines:“Engineers’ work doesn’t directly impact top-level KPI. The peculiarity of their job is that it acts as a “multiplier effect”, allowing the analytics team to work faster and more efficiently. For example, if data engineers can make data move faster, analytics engineers can move quicker too and build more data models.
“Without the engineering team, data analysts and data scientists would spend 70%-80% of their time cleaning data. Having a strong engineering team increases the performance of the analytics team, which in turn positively impacts the performance of other teams.”
Data uptime, defined as the percentage number of times a dataset is delivered on time relative to expected frequency (i.e. the number of times it needs to be updated – daily, hourly, real-time, etc.), and data quality, measured by the number of data issue-triggered events such as failed tests, internal alerts or external consumer complaints, are the major metrics used to measure data engineers’ performance.
Infrastructure cost-saving: Data engineers are responsible for good data management practices – cleaning tables, archiving unnecessary ones, making full use of cloud features, automating processes for efficiency gains, etc – saving tremendously in storage costs. de Leyritz opines, “infrastructure cost-saving thus comes as a natural consequence of a well-performing data engineering team.”
Analytics teams usually have a much more direct impact on top-line business KPIs as they work closer to decision-making processes. The crucial metric here is the turnaround time between the question asked and the answer provided. The goal here is to minimize this time – based originally on a framework established by Benn Stancil for KhanAcademy’s data team.
“The great thing about measuring analytics performance this way is that it encourages analysts to focus their work on real-life decision-making, avoiding them getting lost in exploratory data analysis,” according to de Leyritz.
The direction in which a firm’s analytics team focuses sheds much light on the firm’s data return-on-investment (ROI). Simply, if the analytics team spends most of its time answering tickets for other teams (finance, marketing, etc.), it would probably mean data isn’t operationalized. “This also means your analytics team spends too much time running, solving day-to-day business problems and basic reporting instead of building and focusing on deeper level analysis.”
Although a metric for this cannot be measured adequately, it gives a good idea to help identify which metrics need focusing on. If a firm’s data is not fully operationalized, one should spend more time on data quality – i.e., clean and trusted data.