Helios Retail: A Scorecard That Is Working Against Itself

The scorecard is locally rational and globally destructive.

• SLA attainment at 99.4% is measured against negotiated targets with pause rules. When attainment is the bonus trigger, the cheapest path is to close or pause tickets rather than resolve the underlying cause. The 10% reopen rate within 24 hours is the tell: those tickets were counted as met, then failed in reality.
• Utilization at 91% as a target removes slack. With high variability, near-saturated staff cannot investigate, pair, or absorb the expedites that follow changes, which is consistent with MTTR more than doubling while total effort stayed flat. Utilization measures busyness; the queue it creates is invisible in that number.
• Velocity stable at 42 points measures output volume, not outcome. Stability here is suspicious rather than reassuring: it says the team is delivering the same quantity of work while quality signals deteriorate.
• Change success at 88% looks acceptable in isolation, but combined with 62% of high-severity incidents originating in the previous 72 hours, it means failures are being absorbed as restoration work rather than being prevented. That inflates restoration effort and flatters the change metric.
• Three-minute password resets framed as efficiency are actually a demand problem: a trivial, automatable transaction is consuming engineer time and being rewarded for speed rather than for deflection.

This is Goodhart's law operating in the open: SLA attainment and utilization have become targets, so they have stopped being informative measures. The bonus design is the single biggest driver of the behaviour behind every falling customer and quality metric.

Keep, with clarified definitions. Change failure rate, with severity thresholds and a stated window. MTTR, split into detection, diagnosis and restoration, so the improvement lever is visible. Reopen rate as the quality counterpart to SLA. CSAT, segmented by contact type so deflections do not flatter it.

Retire or demote. Individual utilization from the bonus, immediately. Raw velocity as a performance measure; it can stay as a planning input. Handle time on automatable transaction types. SLA attainment stops being the sole scorecard line: keep it as a contractual and technical measure, and publish pause-clock rules and reopen counts alongside it so the number cannot be gamed silently.

Add, capped at about six measures. Lead time from commit to production, deployment frequency, first-contact resolution with its definition, percentage of incidents attributable to change, flow efficiency or work-item age, and self-service deflection rate.

Sequencing, because people optimise for what pays them. Run old and new measures in parallel for one quarter and publish baselines before changing anything. Remove utilization from the bonus first, since it does most of the damage, and replace it with a team-level flow and quality component with an explicit transition quarter. Keep SLA in the bonus but paired with reopen rate, so closing early is no longer free. Announce definitions and gaming checks in the same note. Changing the whole set and the pay formula in one month produces defensive behaviour and a plausible claim that performance fell because the metrics changed.

Explain construct validity in plain terms. A single number blends flow, stability, customer outcome and capability into an index whose weights are arbitrary. Once it is on an executive dashboard it will be managed, and teams will optimise the components easiest to move, which are usually ceremony adoption and ticket counts. Helios already has direct evidence of what happens when a composite proxy becomes a target: SLA attainment and utilization are precisely that experiment.

Offer a better artifact. A four-quadrant view with two or three measures each: flow (lead time p85, throughput), stability (change failure rate, MTTR split), customer (CSAT by segment, first-contact resolution), and people or capability (attrition, unplanned work share). Show trends with baselines and a short narrative naming the top constraints and what is being done about them. Trends over quarters answer the director's real question, which is whether things are improving and where to intervene.

If the director still insists on a number, provide it as a clearly labelled index with its components and weights published, alongside the quadrant, and state in writing that it is a communication device, not a management target. Never let it become a bonus input. A composite that prompts a conversation is useful; a composite that triggers payment is a liability, and the organisation has already demonstrated it will take the shortest path to the number.