Ask for the SLA Documentation: Pick Accuracy, On-Time Ship and Misses

By Bernardo Campelo, Forbes Business Council E-Commerce Leader, Amazon SPN Certified provider, Amazon SP-API authorized partner, and Founder of PrepVia.

In March 2026, a buyer posted a single line in a private seller forum that reset how a lot of operations teams now open a 3PL vetting call. Before pricing, before a facility tour, before a single question about turnaround: ask for the SLA documentation. Not a capabilities deck. Not a page on a website with a percentage in bold. An actual document, with defined metrics, real numbers behind them, and a stated consequence when the numbers miss.

Most prep centers do not have that document. They have a sales sheet that says something like ninety-nine percent accuracy and a fast turnaround, and when a seller asks where that number came from, the honest answer is usually nowhere in particular. It is a phrase that has been on the website for two years, unattached to any report anyone actually runs.

This post is about the document itself. What it should contain, which numbers inside it actually matter, how those numbers get measured in ways that quietly change the result, and what is supposed to happen on the page when a prep center misses its own target.

Why Buyers Now Ask for the SLA Documentation First

Vetting a prep center used to start with price. Then it shifted to turnaround time, once sellers learned that a cheap rate with a slow dock was not actually cheap. The next shift, visible in that March 2026 thread and in a growing number of RFPs since, is a request for the underlying document rather than the claim on top of it.

The reason is straightforward. A number on a website costs nothing to write. A number inside a document that a buyer can hold a vendor to, month over month, costs the vendor something if it turns out to be wrong. Buyers who have been burned once stop asking what your accuracy is. They ask to see it, on paper, for a period that already happened.

Most operators answer the first version of that question well. Fewer answer the second version at all, because the report either does not exist or has never left an internal spreadsheet. That gap is the whole subject of this post.

What Counts as Real SLA Documentation

A real SLA document does three things a marketing page cannot. It defines each metric precisely enough that two people reading it would calculate the same number from the same data. It shows a track record, not a target, meaning actual results from actual months rather than an aspiration. And it names a remedy, in writing, for what happens when a metric misses.

None of that is the same question as what a prep center agreement should contain overall. Our prep center agreement checklist covers the ten contract clauses worth reading before you sign, things like liability limits and offboarding terms. And it is a different question from what a turnaround SLA actually covers end to end, which we broke down in why your prep center SLA does not mean what you think. This post sits between those two. It is about the performance document a prep center should be able to produce on request, the one with your actual numbers on it, not the clause that promises one exists.

Pick Accuracy: Same Word, Three Different Numbers

Pick accuracy sounds like a single figure. In practice it is at least three different figures, depending on what gets counted as one unit of measurement, and a prep center can quote whichever version flatters the operation without ever stating which one it used.

Per-line accuracy scores each line on a pick list separately. Miss one item on a ten-line order and that order contributes one error out of ten lines, a ninety percent result for that order alone. Per-order accuracy is harsher. It treats the entire order as wrong the moment any single line is wrong, so that same order contributes a flat zero, regardless of whether nine of the ten lines were perfect. Per-unit accuracy is usually the most forgiving of the three, because it spreads the same single error across every unit physically handled that day, not just the units on the one affected order.

None of these three methods is dishonest by itself. The problem is a document that reports a single number, ninety-nine point five percent, without stating which method produced it. A buyer comparing that figure against a competitor using a different method is not comparing accuracy at all. Require the document to name the method next to every accuracy figure it reports, and ask what happens to a partial error, a correct SKU with a wrong unit count inside an otherwise right box, since that edge case is where a lot of quoted numbers quietly diverge from what actually happened on the floor.

On-Time Ship Rate: What Starts the Clock and What Stops It

On-time ship rate looks simple on a scorecard, a single percentage next to a target. The definition underneath it is where prep centers differ the most, because on-time can be measured against several different start and stop points, and a favorable choice of either can flatter a slow operation.

The start point can be the date a shipment plan was created, the date units were fully received and counted, or the date a carrier appointment was confirmed. The stop point can be the moment a carrier physically picks up the freight, the moment a bill of lading is signed, or the moment tracking shows the load in transit. A center that measures from carrier pickup confirmation to carrier pickup, instead of from full receiving to pickup, is measuring almost nothing, because it excludes the entire prep window where most delay actually accumulates.

A real SLA document states both endpoints explicitly, in writing, not as a footnote. Ask what date range each on-time percentage on the report actually spans, and ask what happens to a shipment that ships on time but was originally promised a week earlier through a change nobody flagged. That second question surfaces whether the metric tracks the original commitment or a commitment that was quietly revised after the fact.

Damage Rate and Label Error Rate: The Metrics Nobody Volunteers

Pick accuracy and on-time ship rate show up on most sales pages, because both numbers are usually favorable enough to advertise. Damage rate and label error rate almost never appear, and the reason is not that they do not exist. It is that they are harder to make look good, and few prep centers track them with enough rigor to report them at all.

Damage rate measures the share of units damaged during receiving, prep, or handling before they ever leave the building, expressed as a percentage of units processed. Label error rate measures FNSKU labels that are missing, illegible, duplicated, or mismatched to the wrong unit, the kind of error that causes an Amazon fulfillment center to reject, mis-shelve, or stall a shipment on arrival. Both metrics are operationally more predictive of a bad quarter than accuracy or turnaround alone, because a damaged unit or a bad label creates downstream cost that a clean pick and an on-time truck do not.

Ask for both figures broken out on their own line, separate from pick accuracy, with a stated denominator. A center that folds label errors into a general accuracy number is making its own report harder to audit, whether or not that is the intent.

The Remedy Clause: Credit, Refund of Prep, or Nothing

A metric without a remedy is an observation, not a commitment. Most prep center agreements fall into one of three categories once a metric misses its stated target, and the category is often invisible until a seller reads the actual contract language rather than the marketing summary next to it.

The first category is nothing. The metric is disclosed as informational, with no consequence attached, which is common and not inherently dishonest as long as the document says so plainly. The second is a service credit, a defined amount applied toward a future invoice once a threshold is missed for a stated period. The third is a refund of the prep fee tied specifically to the affected units or shipment, the strongest version, because it applies immediately rather than as a future discount a seller has to remember to redeem.

Our own FastLane 35H guarantee is built as the third kind: thirty-five hours end to end from the point units are ready to move through prep, or the prep is free on that shipment. A remedy clause only works if it names the trigger precisely, states whether it applies automatically or requires a written claim within a deadline, and says whether it caps at a percentage of the invoice. Ask a prospective partner to walk through an actual past instance of the remedy being paid, not a hypothetical one. A center that has never had to pay its own remedy has either never missed, which is unlikely at real volume, or has never actually enforced the clause against itself.

The Monthly Scorecard, and the Question That Exposes a Vendor

Everything above should live inside one recurring document: a monthly scorecard, issued on a set schedule, comparing actual results against contract targets across the metrics that matter. Below is the shape that document should take, with illustrative numbers standing in for a real three-month run.

A document shaped like that does something a sales page cannot. It shows a bad month, Month 2 in the example, sitting next to two good ones, along with what happened as a result. A center willing to show you its own bad month is telling you the remedy clause is real. A center that only ever produces a clean average across a whole year is telling you something else, whether that is the intent or not.

Which brings the whole vetting conversation down to one question, the one worth asking before any pricing discussion at all: send me the scorecard for the last three months, not the sales PDF. A prep center that already runs this report internally for its own quality control produces it in minutes, because the data already exists in a system somewhere. A prep center that has to build something in response to the question is telling you, by the delay itself, that no such report has ever been generated before yours.

We publish our own operating numbers rather than keep them behind a sales call. Our prep window runs 24 to 36 hours, backed by the FastLane 35H guarantee described above. Separately, and this matters because it is a measured average rather than a promise, the time we observe from shipment creation to Amazon's own receiving confirmation runs at approximately 32 hours across our floor. Our counting accuracy runs at 99.9 percent. We are Amazon SPN Certified, one of only 63 providers in that directory, and we operate on Net-30 terms once an account is established. None of that replaces a monthly scorecard specific to your account, and we still recommend asking any partner, including us, for the actual document rather than the summary of it.

Frequently Asked Questions

What is SLA documentation for an FBA prep center?

SLA documentation is a written report that defines exactly how a prep center measures its own performance, such as pick accuracy, on-time ship rate, damage rate, and label error rate, along with actual historical results against stated targets and a defined remedy for a miss. It is different from a marketing page listing a single percentage, because it can be checked against real months and held the provider to, rather than taken on faith.

Why does pick accuracy change depending on how it is measured?

Pick accuracy can be calculated per line item, per order, or per unit, and each method treats the same underlying error differently. A single wrong item on a ten-line order produces a ninety percent result measured per line, a zero percent result measured per order, and a result close to one hundred percent measured per unit across a whole day of picking. A number without its measurement method stated cannot be compared against another vendor's number honestly.

What on-time ship rate should a seller expect from a prep center?

There is no single figure that applies everywhere, because the start and stop points behind the percentage vary by provider. What matters more than the number itself is whether the document states the exact date range the metric covers, typically from full receiving through carrier pickup, and whether that range matches the commitment the seller was actually given at the time the shipment was booked.

Why do damage rate and label error rate matter as much as pick accuracy?

Both metrics predict downstream cost that pick accuracy alone does not capture. A damaged unit or a mismatched FNSKU label can cause an Amazon fulfillment center to reject, mis-shelve, or delay an otherwise correctly picked shipment, so a report that only shows accuracy and turnaround is missing two of the more operationally expensive failure points.

What should happen when a prep center misses its own SLA target?

The agreement should name one of three outcomes explicitly: no consequence beyond disclosure, a service credit applied to a future invoice once a threshold is missed, or a refund of the prep fee tied to the affected shipment. Ask for an example of the remedy actually being paid in the past, since a clause that has never been enforced tells you little about whether it would be honored.

How often should a prep center provide a performance scorecard?

Monthly is the standard that lets a seller catch a slipping trend before it becomes a quarter-long problem. A center that only reports quarterly or annually is averaging away exactly the bad month that a buyer most needs to see, and a center that has never produced one on any schedule likely has never built the underlying report at all.

Does PrepVia publish its own SLA metrics for review?

Yes. Our prep window is 24 to 36 hours with the FastLane 35H guarantee behind it, our counting accuracy runs at 99.9 percent, and the observed average time from shipment creation to Amazon's receiving confirmation runs at approximately 32 hours across our floor. Full detail is on our SLA guarantee page and our facts page, and we provide account-specific scorecards to clients directly rather than relying on the published averages alone.

Final Take

The seller who asked for SLA documentation instead of a sales pitch was not being difficult. That buyer had almost certainly been burned once by a number that sounded precise and meant nothing, and had learned that the fastest way to separate a real operation from a confident one is to ask for the report that already exists, if it exists, rather than a new promise built for the sales call.

Most of what this post describes is not complicated. Define the metric. Show the trailing months. Name the remedy. What is hard is finding a partner who has actually built that document before you ask for it, rather than assembling something plausible after you do. The gap between those two is measured in minutes on the phone and it tells you almost everything else you need to know.

Before you sign with any prep center, ask the one question this post is built around: send the scorecard for the last three months, not the sales PDF. A partner who has one already is showing you how the operation actually runs. A partner who does not is showing you that too.