On-Demand Webinar · 1 hr 3 min

Solve the Analytics Last-Mile Problem with a DataOps Process Hub

World-class engineers and a best-in-class toolchain, and it is still hard to answer an analytic question. Chip Bloche and Chris Bergh argue the bottleneck is process, and walk through a DataOps Process Hub that lets analysts answer stakeholder questions without queueing behind central IT. Recorded November 2021; updated August 2026.

Presented by Chip Bloche, Chris Bergh

What you'll learn 6 points
  • Business analytics teams describe the gap between the warehouse and the answer as a last-mile problem. In their own words: 80 percent of the effort goes into getting data together and 20 percent into insight, and delivering insight to the business is a 90 percent business analytics problem and a 10 percent IT problem.
  • The process hub inverts the table count: instead of a thousand tables from the data lake, each business analytics team gets about a dozen tailored tables it controls, with current data dictionaries and the ability to change data and schema daily or weekly instead of waiting on a multi-month IT cycle.
  • A data hub is a repository for data and a one-way street from raw to processed layers. A process hub holds the processes that act on the data, and it exists because valid data is the product of an effective process. There is no single version of truth without a single version of process.
  • A process hub raises productivity three ways. The Wedge is trusted, analyst-ready data under business analytics control, which lowers the cost per question. The Hammer is full automation from source to delivery, which removes recurring work. The Store is a version-controlled repository holding all business analytics intellectual property for reuse and sharing.
  • DataOps engineering automates eight specific things: production orchestration, production data monitoring and testing, self-service environments, development regression and functional tests, test data automation, deployment automation, shared components, and process measurement.
  • The measured shift is deployment latency from weeks and months down to hours and minutes, with production errors falling from high to low.

Prefer to read it? The written version is in Centralize Your Data Processes With a DataOps Process Hub.

Slides

45 slides

Questions from this session

What is the analytics last-mile problem?

The last mile is the stretch between the data IT delivers and the insight a business customer actually needs. Analysts receive many raw tables and spend most of their time on fragile, repetitive, error-prone data work rather than on original analysis. Because there is no tailor-made data representation, no current documentation, and no reusable business logic, ad hoc analysis takes longer and training new analysts is difficult.

What is the difference between a data hub and a process hub?

A data hub is a repository, exchange, and collection point for enterprise data, and it moves one way from raw to processed layers. A process hub holds the people and processes that act on data, works alongside the data hub, and handles the fact that data flows between teams and functions rather than simply down levels. The argument for it is that valid data is a product of an effective process, so there is no single version of truth without a single version of process.

How does a process hub increase analytics productivity?

Three mechanisms. The Wedge gives business analytics trusted, analyst-ready data sets under its own control with rapid changes and high reuse, lowering the cost per question. The Hammer automates end to end from source to delivery so recurring production work stops consuming staff. The Store keeps all business analytics intellectual property, including SQL, models, reports, tests, and scripts, in a version-controlled repository, cutting risk and redundancy.

What tasks does DataOps engineering automate?

Eight of them: production orchestration, production data monitoring and testing, self-service environments, development regression and functional tests, test data automation, deployment automation, shared components, and process measurement. DataOps engineers take nuggets of existing code, meaning ETL, SQL, Python, XML, and Tableau workbooks, put them into pipelines, wrap them in tests, and run the resulting factory.

Why can't a process hub be owned by IT alone?

A process hub needs business expertise, analytics expertise, and a meta-understanding of which parts of a process generalize and to what level, and it is subject to constant change as conditions and questions change. Analysts alone cannot own it either: it requires enterprise priorities, cross-functional cooperation across multiple teams and vendors, and a process orientation rather than a deliverable orientation. IT and business analytics work toward different suns, IT toward its development process and tech standards, analytics toward customer insight.

What is the risk of leaving business logic on a vendor's laptop?

Undocumented SQL on individual laptops is tribal knowledge, and tribal knowledge means high cost and slow speed to insight. No one can answer what the code is, where it is used, who wrote it, whether it is tested, whether it is current, or how to find and share it, and it can be lost when a person or vendor rotates off. Storing that logic in a version-controlled process hub makes it reusable across projects and survives team turnover.

Where to go next