Site Activity Rates: Calibrating Productivity from Your Own Actuals

Every warehouse cost model is built on productivity rates — cartons packed per hour, pallets put away per hour, lines picked per hour. Almost every model starts those rates from an industry benchmark, because at the beginning there is nothing else. The problem is that most operators never replace them, and go on quoting a five-year-old average against sites that stopped resembling it long ago.

Three tiers, and the one most operators are missing

A mature productivity model has three levels, each overriding the one above:

Most operators have the top and the bottom. The industry number is baked into the spreadsheet; a determined estimator sometimes types a different figure into a particular tender. The middle tier — a durable, reviewed, per-warehouse rate — is the one that rarely exists, and it is the one that carries all the value.

Why two warehouses doing the same work don't achieve the same rate

Nothing about a productivity rate is universal. The same picking task moves at genuinely different speeds depending on:

• Layout and travel — pick-face density, aisle length, whether the pick path was designed or inherited
• Racking and reach — how much of the work needs equipment, how high the fast-movers sit
• Systems — paper, RF, voice or goods-to-person, and how well the WMS sequences work
• Team tenure — a site with two years of stability outperforms one with 40% annual churn, every time
• Product profile — unit size, weight, fragility, whether items are scannable at pace
• Dock and constraint — a site that loses two hours a day to dock congestion pays for it in every rate

Differences of 15–30% between two of your own sites doing nominally identical work are entirely normal. Pricing both from the same number means one of them is quoted too cheaply and the other too dearly — and you will lose the wrong one of the two tenders.

Measuring the real rate

You do not need per-activity time studies to get a usable answer. If the site records actual labour cost by period, and the plan for that period is known, the difference is measurable:

A factor of 1.06 means the site is running 6% dearer than modelled, so the true achieved rate is about 6% lower than assumed. A factor of 0.94 means the opposite — the site outperforms the model and you have been over-quoting.

Two thresholds that keep it honest

A calibration system that fires on every wobble is worse than none, because people stop reading it. Two bars do most of the work:

• Enough evidence. Require a meaningful window — at least six usable months of actuals in a rolling twelve — before any rate is proposed. A quarter is a season, not a trend.
• Enough difference. Only propose a change when the measured rate is meaningfully off the current one. A 5% floor is a sensible starting point; below that you are chasing measurement noise and re-baselining for no commercial gain.

Applied together, a well-run site produces a handful of proposed corrections a year, each one worth the two minutes it takes to review.

Future quotes only — never a retrospective re-price

This rule is not negotiable and it is worth stating explicitly to anyone nervous about automated rate changes: an updated site rate applies to quotes built from that point forward. It never reaches back into a contract that has been signed. Signed is signed. A live contract's baseline stays frozen; measured performance feeds the next deal, and the renewal.

Alongside that, every accepted correction should record who approved it, when, and the evidence behind it. Six months later, when someone asks why this site's put-away rate is 12% above the group standard, the answer is a record, not a memory.

The compounding advantage

This is where an operator's own data quietly becomes a competitive weapon. Two providers bid the same tender:

• One prices from an industry benchmark, adds a safety buffer because they know the benchmark is rough, and quotes a number that is either uncompetitive or accidentally thin.
• The other prices from what their nominated site demonstrably achieved over the last twelve months, buffered by the actual variability they've measured.

The second operator can quote closer to the bone with more confidence, and can defend every rate in the room. Each contract they run makes the model better, which makes the next quote sharper. Nothing about that is available from a spreadsheet full of hand-typed assumptions.

Governance: who owns the number

Productivity rates are commercially sensitive settings, not casual fields. Sensible practice:

1. Proposed corrections are surfaced with their evidence, never applied silently.
2. Acceptance is an administrator-level action, recorded against a person.
3. A rejection carries a reason and quietens the suggestion — until the measured number moves materially again.
4. A per-quote override remains available for genuinely unusual work, without disturbing the site default.

The principle

Industry benchmarks are a starting position, not an answer. The rate that should price your next tender is the rate your warehouse actually achieved last year, measured from your own actuals, reviewed by a person, and applied to future work only. That is the difference between a cost model that ages badly and one that gets more accurate every month it runs.