In 2021 I helped an outdoor lighting client review a 3PL contract in Riverside — 120,000 square feet, and right there in black and white: "order accuracy ≥ 99.5%."
That November, peak season, one mis-shipment: 200 units. The client had two SKUs with duplicate barcodes from an old batch that never got purged. The picker scanned, the system didn't complain, and a full case went out wrong. The client filed a claim. The 3PL opened the contract and walked through it line by line: accuracy was measured against "total outbound order lines for the month" — those 200 units counted as a single line, and out of 40,000-plus lines that month only three were wrong. Accuracy: 99.99%. Target met. And the carve-outs excluded "errors caused by duplicate customer SKUs or non-compliant labeling" — which was exactly what happened. Not a dime paid. The client even ate the return freight.
After that I started telling everyone the same thing: SLA numbers are paper. Definitions are steel. When you negotiate with a 3PL, don't start with the numbers. Start with the definitions. Once the definitions are nailed down, the numbers are just fill-in-the-blank.
The Denominator Decides Whether the Number Is Real
Order accuracy is the most common metric and the one with the most traps. The same "99.5% accuracy" means completely different things depending on the denominator. Measured per order, one wrong line out of twenty sinks the whole order. Measured per order line, one wrong line out of twenty is a 5% error on that order. Same warehouse, same error rate — per-order math might show 97% while per-line math shows 99.9%.
There is a genuine trade-off here. Per-order measurement favors the client, but the 3PL prices the risk in — I've watched an Inland Empire warehouse raise its handling fees 12% before signing a client who insisted on 99.8% per order. Per-line is the more common formulation. Example language: "Order-line accuracy ≥ 99.5%, measured against total system outbound order lines for the month, excluding return orders." Treat that as common example language only — negotiate your own, don't mistake it for an industry standard.
Loss rate has the same problem. What counts as "lost"? Physically missing for more than 30 days, or 90? Some contracts quietly reclassify "system shows stock, can't find it" as a cycle-count variance handled in a separate process — and the loss rate stays a beautiful zero forever. Example language: "Loss rate ≤ 0.1% of total monthly outbound units; units physically missing and unrecovered after 30 days count as lost."
Dock-to-stock time is the core receiving metric, but "when does the clock start" varies by contract: truck arrival at the dock? Unload complete? ASN data received? System receipt posted? I got burned on this one back in 2018 — a client's ASN arrived two days late, and the 3PL counted those two idle days inside its "receiving timeliness." The month-end report showed green. Lesson learned: the clause has to pin down the starting gun. Example language: "Standard receiving completed and system-stocked within 48 hours of unload completion; waiting time caused by late customer ASNs or non-compliant labeling is excluded, provided the 3PL notifies the client in writing within 4 hours."
On-time shipment comes down to two things: the cutoff time, and what "shipped" means. Is it when the carrier label prints, or when the carrier picks up and scans at the dock? That one-day gap is customer complaints and bad reviews during peak. Example language: "Orders released by 3:00 PM ship the same day via carrier pickup scan; on-time rate ≥ 99%."
Inventory accuracy depends on the count methodology: full physical or cycle counts? Are high-value SKUs scored separately from low-value ones, or blended into one flattering average? Example language: "Monthly cycle counts covering all SKUs; inventory accuracy ≥ 99.5% per SKU line, with SKUs showing variances over $500 reported separately."
And the one everyone forgets: who produces the numbers. Plenty of contracts define metrics but never say who measures them or with what data. The 3PL hands you a polished month-end report and you have nothing to challenge it with. Put it in the contract: the client gets real-time access to raw operational data in the WMS, monthly SLA reports are based on system data, and the client has 10 business days to dispute and request a recount.
Credits and Carve-Outs Are the Second Half of the SLA
Metrics without a working remedy are decoration. The standard mechanism is a service credit: for a missed month, a percentage of that month's warehousing plus handling fees gets credited back — commonly 5% to 15% in example language, with a monthly cap around 25%. The devil lives in three details. Does the credit base include storage fees, or only handling? Some contracts credit handling alone, and the real payout shrinks fast. Can credits accumulate, and do they expire? Some say "use it or lose it," and if you're too busy to file, it's gone. And is the credit automatic, or do you have to claim it every month? Make it automatic — no 3PL finance team will remind you.
Carve-outs are where the traps hide, so read every single one. "Force majeure" sounds reasonable, but during the February 2021 Texas freeze I watched California warehouses try to hide behind "extreme weather" too. The contract should bound force majeure: written notice within 24 hours with supporting evidence, or no exemption. "Delays caused by the client" needs the same treatment. Your WMS goes down for two hours, your EDI feed sends bad data — fine, exempt the directly affected orders. But write the boundary: only those orders, not every metric for the entire day.
Peak season deserves its own addendum, not one fixed standard for the whole year. November through December, relaxing time-based metrics by 24 to 48 hours is common practice; accuracy and loss rate usually stay put. Cap the peak-season surcharge in the same addendum so the 3PL can't invent a "labor cost increase" in mid-November. And spell out the exact start and end dates — never write something vague like "during peak season."
Finally, the exit clause — your ultimate leverage. If core metrics miss for three consecutive months (or four cumulative), the client can terminate without penalty, and the 3PL must cooperate on inventory transfer within 60 days. Without this, credits are pocket change: the 3PL will happily pay you a little every month and never fix anything.
Before you sign, run through this list: is every metric defined, with its denominator and measurement period? Who measures, and can you see the raw data? Credit base, rate, cap, expiry. The boundary of every carve-out. Peak addendum dates and relaxations. Penalty-free exit after sustained misses. And in the negotiation room, don't ask "can you hit 99.8%?" Ask three questions instead: what's the denominator, who measures it, and what happens when you miss. Ask those three and they'll know they're dealing with someone who's done this before — and the padding in the quote tends to shrink on its own.
