Building an Internal People Analytics Capability in a Leading Professional Services Firm

Keith McNulty

About me



  • PhD, Pure Mathematics
  • 5 years as a McKinsey Consultant
  • 20 years as an I-O Psychologist Specializing in Human Capital Measurement and Analytics


Keith McNulty

Some key context

  • 2005-2015: Led the development of McKinsey’s first dedicated Human Capital assessment capability
    • Built psychometric assessments for selection, development, team dynamics, leadership
    • Developed data infrastructure to support large-scale assessment programs
    • Consulted clients on Human Capital Assessment
    • Developed quantitative KPIs for assessment quality and effectiveness
  • 2015-present: Led the development of McKinsey’s internal People Analytics capability
    • Inherited a small team of mostly Excel users with almost no automation capabilities or self-service
    • Modernized data infrastructure to support high quality automation and self-serve analytics
    • Introduced advanced analytics capabilities around explanatory modeling and network analysis
    • Grew team by several multiples to support a broad range of People Analytics use cases
    • Consulted clients on People Analytics strategy and implementation
    • Published technical textbooks on advanced methods

Professional services firms


  • People are the primary asset -> no need to justify investment in people analytics
  • High level of analytical interest/sophistication -> appetite for detail and interest in advanced methods
  • Data rich environments -> lots of potential data sources
  • Highly collaborative project based work and non-hierarchical org structures -> networks are important
  • High ‘metabolic rate’ with rapid changes in environment -> need for agility and adaptability in analytical approaches
  • Hypothesis driven mindsets -> significant dangers of confirmation bias from leaders

Transformation required

From:

  • Lots of data, but in poor condition
  • Random people accessing line level data and calculating their own metrics
  • Large manual reporting demand
  • No major self-serve or automation
  • No data science or psychometric capabilities
  • Poor communication between analysts and internal clients

To:

  • High quality, well-governed data, ready for use
  • Well defined metrics and centrally controlled calculations
  • 80%+ of regular reporting needs automated/self-serve
  • Advanced analytics capabilities embedded in core offerings
  • Well functioning and well supported ‘playbook’ for our services
  • Highly agile and adaptable capabilities and technology

Team structure reflects this journey

Dedicated data engineering


  • First hire was a dedicated data engineer, who now leads a team of data engineers
  • Deeply understand context in which data is consumed - transform it to make it easy to consume
  • Data tables feed directly into self-serve products
  • <10% of analysts/data scientists work is data cleaning
  • Advanced capabilities in specialized data such as survey data and communications metadata
  • Engineer non-graph data into graph format

Highly technically skilled BI Analysts


  • Analysts have strong technical skills in SQL, Python, R.
  • All work done in code. Zero use of low/no-code like Excel, Tableau, PowerBI.
  • Analysts can publish their own visualizations/documents/reports for clients using data science publishing platform.
  • Reproducibility of work is a core principle - all analyses are scripted, commented and version controlled.
  • Analysts work closely with data engineers to ensure data quality and availability.
  • All ad hoc work ticketed in JIRA, with clear SLAs for delivery. Tickets reviewed annually to determine if self-serve products should be enhanced based on new common queries.

Dedicated product development team


  • High profile self-service products built using a dedicated team of product developers
  • Front-end/back-end javascript programmers and UX designers specializing in data vizualization
  • Products built using agile development methodologies with frequent releases and rapid iteration
  • Close collaboration between product developers, data engineers, analysts, and translators
  • Products built to be modular and extensible to support rapid addition of new features and data sources.

Intentional advanced data science capabilities


  • Focus on explanatory - not predictive - statistical modeling to support strategic decision making
  • Quantitative psychometricians on team to support design, validation, and analysis across products and services
  • Graph data science capabilities supporting rapid growth in need for network analysis
  • Strong emphasis on interpretable models and explainable AI
  • Careful use of LLM technology to augment - not replace - human expertise

Translation capability unlocks technical capacity

  • Dedicated translators work with clients to understand needs and plan delivery of work
  • Consultants with deep interest in human capital and in data and analytics
  • Translate business questions into technical requirements for analysts/data scientists
  • Translate technical results into balanced practical insights for clients
  • Free up analysts/data scientists to focus on technical work rather than client management/meeting paralysis
The translator holds the umbrella