Performance Management & Costing

Performance management and costing are what make shared services accountable — to its customers for service quality, to the enterprise for cost, and to itself for continuous improvement. Without rigorous measurement, shared services is a cost centre whose value is asserted, not demonstrated; with it, shared services is a capability whose value is provable and improvable.

The measurement system has two halves. Service performance — SLAs, KPIs, OKRs — measures what shared services delivers. Costing — cost-to-serve, chargeback, benchmarking — measures what it costs. Both are needed; service without cost is unaccountable to finance, cost without service is unaccountable to customers. The two together are the balanced scorecard of shared services.

This page covers the performance framework (SLAs, KPIs, OKRs), costing models, cost-to-serve, chargeback and showback, benchmarking, and the discipline of using measurement to drive improvement rather than to report.

The performance framework is layered. SLAs (service level agreements) measure the service commitment to the customer — provision time, restore time, availability. KPIs (key performance indicators) measure the process — first-contact resolution, ticket volume, backlog. OKRs (objectives and key results) measure the outcome — cost reduction, customer satisfaction improvement. Each layer has a purpose and an audience; conflating them produces metrics that serve none.

SLAs are the contract; they define what the customer can expect and what shared services commits to. SLAs must be few, measurable, and calibrated to business impact — too many SLAs become noise; unmeasurable SLAs are aspirations; miscalibrated SLAs are either too generous (undermining efficiency) or too tight (undermining credibility). The discipline is a small set of SLAs per service, each tied to a business-relevant outcome.

KPIs are the internal process measures that explain SLA performance and guide improvement. A missed SLA is explained by KPIs (where in the process did we fall short?), and improvement targets KPIs (which process change improves the SLA?). OKRs, in turn, set the direction — what outcomes shared services is committing to this period — and the KPIs and SLAs are the evidence of progress. The layers connect: OKR direction → KPI process → SLA service.

Costing is the discipline of knowing what shared services costs, by service, and why. The model allocates shared services cost (people, platforms, vendor fees) to services, producing a cost-to-serve per service. Without it, shared services cost is a single number that cannot be improved or charged; with it, cost is attributable to services that can be optimised and consumed.

The costing model has standard components: direct cost (people and platforms dedicated to a service), allocated cost (shared platforms and management allocated by a driver — FTEs, transactions), and the unit rate (cost per transaction, per user, per service). The allocation driver matters — a bad driver (allocating platform cost equally across services regardless of use) distorts unit rates and misleads decisions; a good driver (allocating by transactions) reflects real consumption.

The model must be maintained. Costing is not a one-off exercise; it is a recurring discipline (monthly or quarterly) that reflects the current organisation, volumes, and costs. A costing model that is not refreshed drifts from reality and misleads the decisions it should inform. The costing owner is accountable for its currency, as the catalog owner is for the catalog's.

Cost-to-serve is the unit economics of shared services: what does it cost to deliver one service one time? The measure shifts focus from total cost (which grows with volume and tells little) to unit cost (which reveals efficiency). A shared services organisation that reports only total cost cannot demonstrate efficiency; one that reports cost-to-serve can — and can show efficiency improving as standardisation and automation reduce unit cost.

Cost-to-serve is the metric that connects the value levers to measurement. Consolidation reduces cost-to-serve by removing duplication; standardisation reduces it by removing variance; automation reduces it by removing manual effort. Each lever's effect is visible in cost-to-serve, which is why it is the single most valuable cost metric in shared services. A programme that cannot show cost-to-serve falling has not delivered the value it claims.

The measure also guides decisions. A service with high cost-to-serve and high volume is the priority for standardisation and automation; one with low volume is not worth the investment. Cost-to-serve, combined with volume, is the prioritisation input for the improvement programme — directing effort where it returns the most.

Chargeback and showback are the funding models that shape demand by making consumption visible. Showback tells business units what they consume and what it costs, without charging — informing demand without constraining it. Chargeback actually charges business units for what they consume — constraining demand with price. Most mature shared services move from allocation (cost spread by formula) through showback to chargeback as services and metering mature.

The models shape behaviour differently. Allocation gives business units no demand signal — they consume what they want, and the cost is spread regardless. Showback gives the signal without the consequence — business units see their consumption, which often reduces it. Chargeback gives both — business units pay for what they consume, which shapes demand decisively. The choice depends on the enterprise's maturity and the demand problem it is solving.

Chargeback's risk is that it can discourage legitimate use of shared services, driving business units to cheaper shadow alternatives that cost more enterprise-wide. The mitigation is to price shared services competitively (which the value levers enable) and to govern shadow IT. Chargeback works when shared services is both cheaper than the alternative and the governed choice; it fails when shared services is expensive and the alternative is ungoverned.

Benchmarking compares shared services performance and cost to external peers — other enterprises, industry data, or specialist benchmarks. Internal-only metrics drift: a shared services organisation that improves against its own baseline may still be well behind peers, and not know it. Benchmarking provides the external reality check that keeps internal improvement honest.

The benchmarks that matter are cost-to-serve per service (am I efficient relative to peers?), service quality (am I delivering comparable service?), and the productivity trajectory (am I improving as fast as peers?). Each is a comparison that contextualises internal metrics. A cost-to-serve that looks good internally but is twice the peer benchmark is not good; one that looks flat internally but is improving faster than peers is.

Benchmarking has limits. No two enterprises are identical, and a crude comparison can mislead (different service scope, different volume, different quality). The discipline is to benchmark like-for-like (comparable services, adjusted for scope) and to use the benchmark as a directional check, not a precise target. The benchmark that drives 'why are we different?' inquiry is valuable; the one that drives 'match the number' action can mislead.

Measurement's purpose is improvement, not reporting. A shared services organisation that produces a thick monthly report no one acts on has measurement theatre, not measurement. The test is whether each metric drives a decision or action — a metric that does not is noise, however interesting it looks.

The discipline is the metric-to-action loop: measure, identify the gap, act, verify. Each metric should connect to an owner who can act on it and a review where it is discussed. A metric without an owner is unactioned; a metric without a review is unreviewed. The measurement system is the structure; the owner-and-review is what makes it drive improvement.

The culture matters as much as the system. If measurement is used to blame, people game the metrics; if it is used to improve, people surface the gaps. The leadership that treats metrics as improvement input, not judgment, gets honest measurement and real improvement; the leadership that treats metrics as verdict gets defended numbers and hidden problems. The culture is the foundation; the system is the instrument.