Analytics Baseline
Data Discovery
Data Discovery: Ensure that relevant data sources have been explored and understood to assess what data can potentially be used for the given purpose, by taking a hypothesis driven approach. This includes establishing and testing keys, granularity, understanding business logic, mapping data to business processes and terminology, reconciliation between sources, and validation with SMEs.
Data Quality: Identify any data quality issues with candidate data, validate with SMEs, document and communicate the impact along with any suggested initiatives to mitigate these.
Data Modelling: Establish an agreed approach for data modelling, both a standard framework (Kimball, Data Vault) and a common set of principles that cover normalisation, naming, and handling of bitemporal data. The choice should be driven by the specific needs of the customer, with emphasis on end user needs.
Reporting and BI
User-Centered Development: Develop with a specific group of end-users in mind, work with them to design a solution that best meets their analytical needs, keep them regularly updated on progress, and run formal User Acceptance Testing before deploying to production.
Multi-View Design: Build dashboards to enable users to obtain views of the data at different levels of granularity, such as a summary view highlighting key metrics, a detail view allowing the user to explore key concepts or business questions, and a tabular view allowing the user to see the underlying data in table form. Limit the number of filters and parameters, or consider splitting into multiple tabs or reports.
Data Lineage and Definitions: Include an Introduction page that explains the intended use of the product, definitions of key concepts, calculation methodologies for key metrics, data sources, refresh schedule and last refreshed date.
Usage Monitoring: Set up automated tracking on any reports or dashboards created to ensure that intended users are actually using them. Run follow-up sessions with them to gather and implement any post-launch feedback.
Self-Service Enablement: Ensure any data artefacts created as part of report/dashboard creation are well documented, reusable and discoverable for power users who want to explore and visualise data themselves.
Measuring Success Effectively
Metrics and KPIs: Follow best practices regarding the design and communication of metrics and KPIs, such as considering their relevance to strategic goals, compiling a mix of leading and lagging indicators, and clearly documenting methodology, cadence and channels for distribution.
Experimentation: Advocate for experimentation to enable statistically significant (or otherwise) conclusions to be drawn about hypotheses wherever possible. Gather sufficient information about the question being asked to determine whether a Bayesian or Frequentist approach is more appropriate.
Data Storytelling
Craft a Narrative: Data stories should contain relevant actors, important business context, a clear description of the problem/opportunity, supporting data/visuals and a call to action.
Know the Audience: Focus on the impact and outcomes of the analysis rather than the technical details of how the analysis was performed. Consider the decision-makers’ depth of understanding of the topic and the time/focus they can devote to it when crafting a data story.
Visualise Concisely: Use simple visualisations as supporting evidence for the story, leveraging colour and annotations sparingly to call out the most important points. Call out only a select few compelling and memorable numbers to avoid overwhelming the audience with figures.
Be Transparent: Build trust and confidence in decision-makers by proactively mentioning any assumptions and limitations in the data and/or analysis.
Align With Engineering
One-Team Approach: The analyst knows what the engineering team is currently working on and how it relates to their workload, and vice versa.
Ways of Working: Align with engineering documentation practices and ways of working where possible.
Architecture Design: Use knowledge of data analytics tooling and end user needs to contribute to software architecture design decisions.