On-Demand Webinar · 55 min

Driving Data Analytic Team Excellence Through Agility, Efficiency, and Aphorisms

James Royster of Karuna Therapeutics on twenty-five years of running commercial pharma analytics teams: ship at 70%, prevent problems instead of getting good at fixing them, and automate anything you will do twice.

Presented by Chris Bergh

What you'll learn 6 points
  • Ship at 70–80% complete and get feedback, rather than holding a 100% answer nobody has seen yet. Customers form their thinking on the draft.
  • Your competition is Amazon. Expectations for speed are set by people's lives outside work, not by what a data team considers reasonable.
  • Prevent problems instead of getting good at fixing them. Repeatable process and standardisation beat firefighting, even though firefighting feels like the job.
  • Automate anything that will be done more than once, so the team's attention moves to the next question instead of the last one.
  • Hinge activities are the tasks that open other doors — data quality testing and structuring data for easy access are two of them, and they compete with today's urgent work.
  • People need to feel connected to the customer. If team members are going through the motions, the organisation should reassess its approach rather than push harder.

Prefer to read it? The written version is in Webinar Summary: Driving Data Analytic Team Excellence Through Agility, Efficiency, and Aphorisms.

Slides

17 slides

Questions from this session

Why does analytics speed matter so much during a pharmaceutical product launch?

It costs $2.6 billion to bring a drug to market, according to the Tufts Center for the Study of Drug Development, and the first six to twelve months of a launch are decisive: how fast a product grows during that window shapes its overall lifetime revenue. Analytics that arrive a quarter late cannot influence the period that matters most.

What mindset changes does DataOps ask of an analytics team?

Six shifts, presented as a from-and-to list: from fear of change to velocity of change, from manual operations to automated operations, from hoping for quality to integrating quality, from a hero mentality to repeatable processes, and from perfection to being 70 percent right the first time. As James Royster of Karuna Therapeutics puts it in this session, not being a hero really means not solving problems but figuring out how never to create them.

What does 70 percent right the first time mean in practice?

It means shipping a draft early enough that feedback can still change it. On a new large data set the analytics team builds scrappy dashboards, sends feedback to data engineering, shows the early version to users, runs active build and design sessions making as many changes live as possible, and only then publishes the production dashboard. Multiple takes lead to highlight reels.

What is a scrappy star schema?

It is the data engineering counterpart to a scrappy dashboard: a rough star schema built quickly in a data mart from a new large data set, rather than a finished model built in isolation. The team sends questions back to the data supplier while keeping analysts in the loop, adds data tests so speed does not cost quality, shares the star with the analyst team for feedback, iterates over several Agile sprints, and then releases a solid star.

What does a commercial pharma analytics stack look like?

Syndicated data, prescription and claims data, specialty pharmacy feeds, sales data, and non-personal promotion events and campaigns land in Snowflake, where data engineers build the models and analysts maintain sales alignments, product hierarchies, and specialty mappings. The output goes through Tableau Online to hundreds of sales people plus marketing and executives, and covers ongoing production deliverables such as a weekly launch tracker, ad hoc answers to business questions, and resource allocation and predictive models.

What can an analytics leader actually control?

Not the source systems, not IT, not what marketing and sales ask for, and not the customers. What sits inside an analytics leader's span of control is the team's own process: how quickly it can absorb a change, whether quality is built in or hoped for, and how it measures itself. The framing in this session is that the problem is not too much data, it is processes too slow to take advantage of the data.

Where to go next