Leading the QM function — the four-stage maturity ladder

Quality · ~6 minute read

Quality functions don’t become strategic by accident. There is a four-stage ladder — ad-hoc, operational, embedded, strategic — and the leader’s job is to locate the function honestly, name the next-level move, and operate the disciplines that earn the operation’s dependence. The honest location is usually lower than the self-assessment.

The four stages

Stage 1 — ad-hoc: a small team, requests-driven, reactive; scoring happens, coaching is inconsistent, calibration exists on paper. Stage 2 — operational: the foundations in place — deliberate scorecard, calibration running, a coaching frame, some routing; behaviour change visible at the agent level. Most centres reach Stage 2 within one to three years if they invest.

Stage 3 — embedded: the advanced disciplines operational — data model, statistical rigour, diagnostic depth, some predictive work, the three-mode operating model, named-ownership governance; QM integrated into operational decision-making rather than just scoring it. Stage 4 — strategic: predictive and prescriptive work routinely informing major decisions, the function leading capability shifts rather than reacting to events. Few operations reach Stage 4, and those that do take five-plus years of deliberate investment.

Your stage is your weakest leg

Self-assess on seven dimensions, rated 1–4: data and substrate; governance; statistical rigour; diagnostic depth; predictive capability; operating model; talent and capability. The function’s stage is the lowest of the seven — not the average, and certainly not the highest. A function with Stage-3 ambitions and Stage-1 statistical rigour is a Stage-1 function with expensive tooling.

That rule disciplines the investment plan: address the weakest leg first. One function rated itself substrate 2, governance 2, statistical rigour 1, diagnostic 1, predictive 1 — honestly early Stage 2 — and built an 18-month plan around the ones: data model and observability, SPC and significance training, a structured diagnostic process, one piloted predictive model. Eighteen months later it had crossed into solid Stage 3. The deltas compounded because they went where the weakness was.

The role-mix the next stage needs

The function evolves from one or two generalist analysts toward a deliberate mix: QM analysts running the disciplined cycle; senior analysts carrying the diagnostic, statistical and predictive depth; a data engineer for substrate and observability; a speech-analytics specialist for the platform, model tuning and calibration overlap; an insight lead translating between analysts and leadership; and the function leader on strategy, governance and growth.

Hire on dispositions over techniques: the coaching-with-data stance, statistical rigour, causal discipline, calibrated honesty. Techniques are teachable; the dispositions are harder to instil. And develop deliberately — the analyst who joined at Stage 1 should be operating at Stage 3 within three to five years, and if they aren’t, the development practice is the thing to fix.

The leader disciplines

Invest in the substrate — data model, observability, governance — unglamorous, expensive, and the reason functions plateau at Stage 2 when skipped. Run the function as a function: data-product specs, named ownership and SLAs applied to your own assets, not just preached to others. Retire what doesn’t work — scorecards, models, routines and meetings alike; a function that only accumulates calcifies. Track and communicate the value as operational outcomes enabled, attributed fairly, not activity reported.

And cultivate epistemic humility. The lead who refuses analytical work that doesn’t fit — the prediction without enough data, the causal claim without a test — earns long-term credibility; the lead who always has an answer loses it. Calibrated honesty is the function’s currency, and the leader sets the exchange rate.

The only test that matters

The closing principle of the whole discipline: the only test of QM that matters is whether it was acted on. The data model, the statistics, the diagnostic and predictive techniques, the speech-analytics integration, the governance — all of it is means. Acted-on behaviour change is the end: agent-level coaching that lands, operation-level patterns that get fixed, leadership decisions backed by trustworthy evidence.

The leader who keeps that front of mind — building the disciplines that produce it and refusing the work that doesn’t — runs a function the operation comes to depend on. That dependence is the seat at the table, and it is built, year on year, weakest leg first.

The closing principle

Locate the function honestly — weakest leg, not fondest hope — name the next-level move with investments and dates, and hold every technique to the one test that matters: was it acted on? Everything else is means.