Throughput per hour is the metric that 3PL robotics evaluations almost always lead with — and it's also the metric most likely to produce a misleading ROI model if it's the only number in the business case. In our experience reviewing pick automation deployments, the operations teams that get the strongest multi-year returns are the ones who expand their metrics framework before the purchase decision, not after the deployment reveals that throughput alone doesn't tell the full story.
Why Throughput Per Hour Is Necessary but Not Sufficient
Throughput per hour (TPH) — units picked per station per hour — is a reasonable headline metric because it's easy to measure, easy to compare, and maps directly to the primary value of a pick station. A station doing 600 UPH is doing more useful work than one doing 400 UPH at the same cost. That logic is sound.
The problem is that TPH is a best-condition metric. It's typically measured during vendor demos on pre-selected SKU catalogs under ideal conditions. Production TPH in Year 1 on a real 3PL catalog with novel SKUs, mixed-geometry bins, and first-shift operators still learning the system is usually 15-30% below demo TPH. A business case built on demo TPH will miss Year 1 actuals, which creates budget problems and credibility issues for the operations team that made the recommendation.
The Metrics That Actually Drive ROI
The more complete ROI framework for 3PL pick automation includes five metric categories:
| Metric Category | What to Measure | Why It Matters |
|---|---|---|
| Throughput | UPH broken down by product SKU category and shift | Baseline productivity, drives labor offset |
| Quality | Net error rate (post-recovery), chargeback events | Rework cost, client SLA compliance |
| Uptime | Station uptime %, MTTR on failures | Realized vs. rated throughput |
| Labor profile | Headcount by role, turnover rate, ramp cost | True labor cost reduction including turnover |
| SKU coverage | % of catalog volume handled robotically | Automation utilization and growth path |
Uptime as a ROI Driver
Station uptime is probably the most underweighted metric in 3PL robotics evaluations. A pick station rated at 600 UPH but running at 75% uptime is delivering an effective throughput of 450 UPH. The gap between rated and effective throughput maps directly to unfulfilled ROI. At a facility running 20 hours per day, the difference between 85% and 75% uptime on a 600 UPH station is 1,200 picks per day — roughly $4,800-9,600 in recovered labor cost at current rates, every day.
Uptime data is something vendors often don't lead with. Ask for mean time between failures (MTBF) and mean time to repair (MTTR) data from production deployments, not just specifications. Ask what the most common failure modes are and how long they take to resolve. That data tells you the real uptime profile, which feeds directly into the effective throughput number in your ROI model.
SKU Coverage Percentage and Its Trajectory
Not every SKU in a 3PL catalog is automatable by current pick technology. Items that are too flexible, too light, too reflective, or too irregularly shaped may fall outside reliable grasp parameters for any current system. The relevant metric is SKU coverage percentage: what fraction of your total pick volume can the robotic station handle reliably, and how does that fraction change over time as the system improves and catalog shifts.
A system covering 65% of your pick volume on day one but growing to 80% by month 18 has a meaningfully better ROI trajectory than a system that peaks at 65% and stays there. Ask vendors specifically about coverage trajectory and what drives coverage expansion — whether it's model updates, hardware upgrades, or operational tuning. That trajectory is part of the return, not a footnote.
Chargeback Reduction as a Measurable Return
Pick accuracy improvements from automation translate directly into retail chargeback reduction, and chargebacks are a cost line that 3PL finance teams track precisely. Industry data puts retail chargeback cost for pick errors at $4-9 per event in rework, reshipment, and client penalty fees. At 500,000 picks per month with a 2% manual error rate, that's 10,000 error events — $40,000-90,000 in monthly chargeback exposure.
If robotic pick stations reduce the net error rate to 0.6% (achievable in Year 2 deployments with in-station recovery), chargeback exposure drops to roughly $12,000-27,000 per month — a $28,000-63,000 monthly reduction. That's a measurable, finance-validated return that belongs in the ROI model alongside labor offset, and it's one that typically requires no capital assumptions beyond the station deployment itself.
Throughput per hour is the starting point, not the finish line. The 3PLs that get the best long-term returns from pick automation are the ones that model quality improvement, uptime trajectory, SKU coverage expansion, and labor profile change alongside throughput — and then hold vendors accountable to those metrics in deployment contracts, not just in demo settings.