How busy is the team or system?
№ 203 · Utilization
Definition
MetricShare of available paid time or provisioned capacity that is in productive use
Unit% utilised
Line of (busy time divided by available time) for ops resources. 75-85% is healthy. Build from capacity tracking systems. Above 90% means burnout risk; below 60% means under-utilized capacity. Example: 78% staffing utilization Q1, 92% Q4 (burnout risk emerging).
Benchmarks
| Bottom 30% | Median | Top 30% |
|---|---|---|
| below 60% billable utilisation | 66% to 69% billable utilisation | above 75% billable utilisation |
Professional services and billable-team utilisation, SPI Research Professional Services Maturity Benchmark, n=403 firms across seven industries, 2024 and 2025 measurement years. Non-people resources sit in the category table and are not comparable in level. · Three cautions. First, direction is shape, not higher_better: sustained utilisation above roughly 80% removes the bench time that funds selling, quality work and training, and the cost shows up two quarters later in pipeline and attrition. SPI treats 75% as the optimal threshold, and practitioner frameworks put the healthy band at 70 to 84%. Second, the denominator decides the answer. Large benchmark databases deliberately do not adjust for sick days, holidays, training or idle time, so internally computed figures that strip those out will look several points better than peer data for no real reason. Third, the panel is moving: billable utilisation fell to 68.9% in 2024 and 66.4% in 2025, the lowest in SPI's survey history, so a firm holding 70% today is performing better relative to peers than the same number would have implied in 2021, when the panel sat at 73.2%.
| Category | Bottom 30% | Median | Top 30% |
|---|---|---|---|
| Professional services billable people | below 60% | 66.4% to 68.9% | above 75% |
| Senior consultants and managers | below 50% | 55% to 70% | 70% |
| US industrial capacity | below 73% | 75.7% to 76.1% | above 80% |
| Kubernetes CPU | below 5% | 8% to 10% | above 40% after automated rightsizing |
| Kubernetes memory | below 15% | 20% to 23% | above 50% |
When it looks bad
Utilisation climbing quarter on quarter while revenue per head is flat or falling, which means the extra hours are being sold cheap or written off. The mirror case is utilisation holding steady while revenue per head declines, which is the same problem arriving through rate rather than volume.
Billable utilisation up from 64% to 71% across three quarters while revenue per consultant falls from 205,000 USD to 188,000 USD. The additional hours went out at a discount.
What to do about it
- Fix the denominator to total paid hours with no adjustment for holiday, sick or training time, so the number is comparable across teams and against peer data. Large benchmark databases use exactly this convention, and mixed denominators are the usual reason internal utilisation looks better than the panel (MetricNet).
- Track utilisation against revenue per head on the same chart and never in isolation. The SPI panel shows utilisation at its lowest recorded level of 66.4% while revenue per consultant rose to 210,000 USD, so the two move independently and either one alone is a false read (SPI Research via Deltek, 2026).
- Set the target as a band of 70 to 80% for billable people rather than a maximum. Sustained operation above that band consumes the bench time that funds selling, quality review and training, and SPI treats 75% as the optimal point rather than a floor (SPI Research 2025, Mosaic 2025).
- For infrastructure, rightsize automatically rather than through request review. Measured Kubernetes CPU utilisation is 8% with 69% overprovisioning, and automated rightsizing halved provisioned CPUs while cutting out-of-memory kills to near zero, so the reliability defence for holding headroom does not survive the data (Cast AI, 2026).
Sources
- SPI Research (Service Performance Insight) Billable utilisation fell to 68.9%, below the 75% optimal threshold, squeezing margins. Revenue growth slowed to 4.6% year over year from 7.8% in 2023. Level 5 maturity firms show 71% higher billable utilisation than Level 1 firms. spiresearch.com ↗
- SPI Research, distributed by Kantata Full report confirming declining utilisation rates and operational inefficiencies across the panel, with billable utilisation and fully loaded cost per consultant tracked as impact measures in the scorecard. get.kantata.com ↗
- SPI Research, reported by Deltek Billable utilisation fell to 66.4% in 2025 from 68.9% in 2024, the lowest level SPI has recorded and well below the 75% target. Project margins rose to 37.7%, revenue per consultant climbed to 210,000 USD, EBITDA held at 9.9%. deltek.com ↗
- Saibon Group, citing SPI Research Healthy billable utilisation band is 74 to 84%; below 74% revenue per consultant typically falls below break-even. The SPI panel was at 73.2% as recently as 2021, a 4.3 point drop over three years. Senior consultants and managers run 55 to 70% with significant non-billable time. saibongroup.com ↗
- Mosaic, citing SPI Research 75% has been the historically accepted benchmark, but leading firms target 70 to 80% as the sustainable band for profitability without burnout. IT consulting organisations maintain closer to 75 to 80%. mosaicapp.com ↗
- Cast AI Average CPU utilisation 8% in 2025, down from 10%; memory 20%, down from 23%. CPU overprovisioning rose from 40% to 69%. Automated rightsizing halved provisioned CPUs and reduced out-of-memory kills from 40 to 50 per interval to near zero. cast.ai ↗
- Cast AI Average CPU utilisation 10% and memory 23% across the panel. The gap between provisioned and requested CPUs averaged 40%, down from 43% the prior year. cast.ai ↗
- MetricNet Defines agent utilisation as total call handling time divided by total work hours, with no adjustment for sick days, holidays, training, project or idle time, specifically so that all contact centres worldwide are measured identically. Names agent utilisation the single most important indicator of agent productivity and the primary driver of cost per contact. metricnet.com ↗
- Federal Reserve Board Total industry capacity utilisation 76.1%, manufacturing 75.7%, mining 87.4%, utilities 69.5%, against 1972 to 2025 long-run averages of roughly 79.4% and 78.2% respectively. federalreserve.gov ↗