If a new 3PL is a fit for your volume, you learn it from a pilot: a limited slice of SKUs and live orders run through the new warehouse under agreed measures, before any full cutover. Write the pilot into the contract with pass criteria and a rollback right, because a test that is not in the contract is a test you cannot hold anyone to.
What is a pilot, and what is it not?
A pilot is limited in three ways: scope (a SKU subset that represents your order profile), volume (a slice of live orders, not a staging exercise) and duration (agreed in weeks, not open-ended). It is not a demo, not a warehouse tour, and not a few sample packages mailed to yourself.
Many checklist articles reduce testing to "send a test shipment" and move on. One shipment tells you the label printed. It does not tell you whether receiving reconciles, whether invoices match the quote, or what the operation does on its third bad day.
How do I structure a pilot?
| Element | What to agree in writing |
|---|---|
| Scope | SKU subset covering single-item, multi-item, kits and returns |
| Volume | A defined slice of live orders from one channel |
| Duration | Enough weeks to cover a full order cycle and one receiving wave |
| Measures | Order accuracy, cutoff-to-ship, receiving count variance, invoice accuracy |
| Rollback | If measures are missed, volume returns to the incumbent without penalty |
| Billing | Pilot work billed at contracted rates, not "project" rates |
The rollback row is the one brands skip and later regret. If the pilot fails, you need the right to stop the migration without a penalty, and your current 3PL needs the capacity to take the volume back.
What should the pilot measure?
- Order accuracy and cutoff-to-ship against the written SLA, not a verbal "we ship fast"
- Receiving counts against the pack list, and how long putaway takes
- Invoice accuracy: every pilot invoice checked line by line against the quote
- Inventory visibility: whether the portal distinguishes committed from available stock
- How problems surface: who calls whom, and how fast the answer comes
For the numbers those measures should sit inside, see what SLA should I expect from a 3PL?. The pilot also validates what a 3PL site visit showed you on the floor.
How long should a pilot run?
Commonly two to six weeks. Shorter than that and you may not cover a full weekly order cycle; much longer and you pay the cost of running two operations at once. Time it so the pilot covers at least one receiving wave, and one launch or promotion-sized day if your calendar has one.
What does the contract need to say about the pilot?
- The measures and the pass thresholds, with numbers.
- The ramp window after go-live and what stays measured during it. Some agreements suspend SLA remedies while a new operation stabilizes; a remedy holiday should not become a measurement holiday, so spell out what is still tracked.
- Billing during the pilot and during the parallel run, when two warehouses hold stock at once.
- Rollback rights, and the notice that triggers them.
If parcel and freight contracts sit in your brand's name, the pilot exercises the same carrier setup you would use at full volume, and it keeps the rates decision separate from the warehouse decision.
When does a pilot fail?
Week-one misses that nobody flags, invoices that drift from the quote, counts that do not reconcile, and answers that take days are each a finding, not a hiccup. A pilot exists to surface the operating behavior you would otherwise discover after a full cutover. For the questions that belong in the contract before the pilot starts, see what to ask a 3PL before signing a high-volume fulfillment contract.
new3PL is an independent adviser and placement firm. It treats a written pilot program as part of 3PL selection rather than an optional extra, and never recommends a specific provider.
FAQ
How big should a pilot be?
Big enough to exercise every order type you sell, small enough that your current 3PL can absorb the volume back if it fails. One channel or one region is a common way to bound it.
Can I run a pilot without moving inventory?
No. The pilot needs real stock to receive, pick and measure; one receiving wave is enough, and you do not need a full transfer to learn what the operation does.
What if a 3PL refuses to run a pilot?
Ask what their process would show that a measured pilot does not. A refusal to test against written criteria is a finding about how the operation runs, not a scheduling problem.