BI product — Evolution and delivery

What lets the product evolve without breaking it.

≈ 3 min readReading

On this page

Industrialization and deployment#

How does a change reach the user in a repeatable way?

Level	Signature	Criteria to meet
0 · Handcrafted	The file is edited in place, then copied over.	Starting point: no criteria.
1 · Locally tooled	New versions of the file are copied manually.	Versions are kept and named.
2 · Shared	Numbered versions, documented manual deployment.	The sources are in a version control system. The deployment procedure is written down.
3 · Industrialized	Build pipeline, separate environments, repeatable deployment.	An automated build produces the deliverable from the sources. Development, test and production environments are separate. A deployment gives the same result every time.
4 · Integrated and governed	Controlled version lifecycle: several versions, migrations, documented deliveries.	Several versions are maintained in parallel under a clear policy. Migrations from one version to another are tooled and tested. Each version is delivered with its documentation and release notes.
5 · Autonomous	Continuous, automatic deployment, with rollback.	Deployment is fully automatic, including to customers' environments. Rollback is possible and tested. Recurring updates and loads of the environments are scheduled and automatic.

Effect on the user

The user receives changes without any intervention or interruption.

How do we know it works, and that a fix breaks nothing?

Level	Signature	Criteria to meet
0 · Handcrafted	Problems are noticed in use.	Starting point: no criteria.
1 · Locally tooled	Occasional manual checks.	The author checks by hand before distributing.
2 · Shared	Test data sets and checklists.	Test data sets exist. A delivery checklist is applied.
3 · Industrialized	Automated regression tests.	Automated tests cover the critical components. The tests run at every build. A failing test blocks delivery.
4 · Integrated and governed	Data, load and security tests; automated acceptance testing of the interface.	The tests cover data (equivalence, invariants), load and security. Interface user journeys are tested automatically. Coverage and escaped defects are measured.
5 · Autonomous	Tests triggered on every change, automatic fixes for known anomalies.	Every change triggers the full test suite. Known anomalies are fixed automatically, with a trace. Defects found in production automatically become tests.

Effect on the user

The user no longer serves as a test bench.

Who decides what evolves, and on what basis?

Level	Signature	Criteria to meet
0 · Handcrafted	Changes follow whatever is asked for at the moment.	Starting point: no criteria.
1 · Locally tooled	Requests received by email or verbally.	Change requests are kept.
2 · Shared	A single backlog.	A single backlog gathers the requests. Each request has a requester and a priority.
3 · Industrialized	Structured planning and tracking (iterations, milestones).	Work is planned by iterations or by milestones. Progress is tracked and shared. Needs are gathered in a structured way (workshops, specifications).
4 · Integrated and governed	Roadmap based on requests, product maturity and constraints; priorities arbitrated.	A roadmap links changes to needs and to the product's maturity. A committee or a ritual arbitrates priorities. Development, testing and quality standards are defined and applied.
5 · Autonomous	Value-driven steering: measured usage and value feed the roadmap.	Usage and value indicators are tracked for each delivered change. The roadmap is reviewed regularly based on these measurements. Little-used changes are removed or reconsidered.

Effect on the user

The user sees their need taken into account, and the product evolving in the direction of their usage.