Pick and pack: the SOP for fewer wrong orders (and when to change methods)

Short answer: pick and pack is the two-step job of pulling ordered items from storage (pick) and boxing them for shipment (pack), and almost every wrong-order complaint traces to one root design flaw: picking from memory instead of from verification. The process below is built so that every step produces physical evidence — a picked tote, a packing-slip check, a barcode confirm — and then we attack it deliberately, because the fastest way to find where your process leaks errors is to red-team it before your customers do.

Pick methods: three, not twelve

Everything else is a variant of these:

• Discrete (order-by-order) picking. One order, one list, one pass through the space. Simple, zero coordination, and the right default until volume makes the walking the dominant cost. Error rate is low because each order is one person's undivided attention.
• Batch picking. Collect items for, say, 5–10 orders in one walk (a tote per order in one cart), then pack from the totes. Walking drops dramatically — one pass instead of ten. The error mode shifts: items can land in the wrong tote, so batch picking requires tote labels and a verification step at pack.
• Zone picking. Each picker owns a zone; orders travel (via cart or conveyor) through zones and get completed downstream. This is the method of multi-person operations, and it introduces handoffs — where multi-person processes typically leak errors (treat that as an operator's assumption, not a law) — so it only pays when no single picker can cover the space.

The rule of thumb: discrete until walking hurts, batch when it does, zones when the building does. Change methods when your pick time per order starts climbing week over week, or when errors cluster at handoffs — not because a method is fashionable.

The SOP

Print this. It assumes a small space (garage to small unit), one or two people, discrete or batch picking.

Pick

1. Generate the pick list in SKU-and-bin order, not in order-of-arrival. A list sorted by shelf location turns a 40-item walk into one pass; Shopify's fulfillment views or a simple export sorted by bin does this. If your bins aren't labeled yet, that's the prerequisite — the small warehouse setup guide covers zones and slotting; bin and barcode discipline covers the codes themselves.
2. Pick against the line, not the product name. Read the SKU and variant — "GRN-STM-L" not "green stove, large." Product names lie; SKUs (when your naming is disciplined) don't.
3. Physically verify each pick. Barcode scan if equipped; otherwise a pencil tick per line at the shelf, never at the end. A tick made later is a guess wearing a tick's clothes.
4. One order, one tote, tote labeled with the order number. In batch mode this is the whole ballgame — an unlabeled tote in a batch cart is a randomizer.
5. Shorts and damages get logged at the shelf. If the bin says 4 and the shelf holds 3, log it now; a phantom count discovered during picking is a count problem, not a pick problem.

Pack

6. Pack from the packing slip against the tote contents, line by line. This is the verification step that catches every upstream mistake — wrong variant picked, wrong tote, short pick. Thirty seconds per order.
7. Weight check where it matters. A cheap shipping scale catches the wrong-variant pack on anything with meaningfully different weights — a spare-parts kit versus its single-part cousin, for example. Even a rough expected-weight column on the slip does the job.
8. Final-in, then label. Packing slip or insert goes in last, box gets sealed, label goes on, and the order is marked fulfilled in Shopify after sealing — not before, or your system says shipped while the box says otherwise.

Red team: attacking this process on purpose

Now assume the process above is your plan, and attack it the way your busiest week will:

• "The pick list is stale." Between generating the list and walking the shelf, a PO was received and shelved, a return was restocked, or a POS sale moved stock. Your list says bin B3 holds 6; it holds 11. Failure mode: over-picks are rare, but the picker who "helpfully" takes 6 when 4 are ordered creates stray stock in a tote. Fix: pick lists are generated at pick time, not at breakfast; receiving and picking don't run simultaneously in the same aisle.
• "Two orders, same customer, same day." Combine-eligible orders that ship separately cost double pick, double pack, double postage, and annoy the customer. Fix: a merge check before pick-list generation — any order-window where merging applies is a policy decision made in advance, not a judgment call at the bench.
• "The tote jumped carts." In batch mode, one tote gets moved to the neighboring cart mid-pick. Every order downstream of that is wrong. Fix: one cart per batch, totes never cross carts, and step 6 catches it at pack.
• "The substitution instinct." Bin is empty for the exact variant; the picker substitutes the near-variant because "the customer won't care." Customers care. Fix: substitutions are never a picker's call — short it, log it, and let the pack step or a support message handle it.
• "Fulfilled-before-sealed." Someone batch-marks orders fulfilled in Shopify at 3 p.m. while boxes sit sealed-but-unlabeled until 5. During those two hours your "unfulfilled" queue lies, and any count or cover math built on fulfillment state inherits the lie. Fix: the mark-fulfilled action happens at sealing, per the SOP, no exceptions — or if batching truly helps, the batch closes the same hour it opens.

Every one of these attacks was found in the process as written, not by imagining a careless worker — which is the point: error-proofing the system beats training the person, because the busiest week of your year is the least capable of remembering training.

When to change methods — or hand it over

Three triggers, in order of cost:

1. Pick time per order trending up across weeks. Walking has become the job. Move discrete → batch before anything else; it's the cheapest fix with the biggest return.
2. Errors clustering at a new stage. Wrong-tote errors rising means batch without verification discipline — fix step 6 before changing methods.
3. The build-versus-buy line. When outsourced pick-and-pack fees get close to your own labor-plus-space cost, or when you're paying for space your pick faces dominate, it's time to price fulfillment partners seriously. The relationship side — SLAs, data sync, counting through someone else's WMS — is covered in the 3PL guide and the network option in the SFN comparison; the physical process you've built above is exactly what you'll demand they replicate.

One measurement to keep regardless of method: pick accuracy, errors caught per hundred orders. It's the number that tells you the SOP is alive. If you're not measuring it, you're not picking at a known error rate — you're guessing, which is the thing this whole page exists to end.

Common questions

What does pick and pack mean? Pick and pack is the fulfillment process of pulling the ordered items from storage (picking, against a pick list sorted by location) and packing them into a shipment (packing, verified against the packing slip) with the label applied. It's the workflow every ecommerce order passes through, whether done at your own bench or outsourced to a fulfillment partner.

What's the difference between discrete, batch, and zone picking? Discrete picks one order at a time — simple and low-error for small volume. Batch picks multiple orders in one walk with a tote per order — fewer steps, but requires tote discipline and pack-stage verification. Zone picking assigns each picker a fixed area and passes orders through zones — built for multi-person operations, at the cost of error-prone handoffs.

How do I reduce picking errors? Verify at the shelf (scan or per-line tick), pick by SKU and bin rather than product name, keep one labeled tote per order, and check the tote contents against the packing slip before sealing. Most errors are system gaps — stale lists, unlabeled totes, unverified packs — not careless workers, so error-proof the process instead of retraining people.

When should I switch from picking orders myself to a 3PL? When your own pick-and-pack cost — labor, space, and your time as a hidden line — approaches the all-in fee an outsourced partner quotes for your order profile, or when fulfillment volume is crowding out the work only you can do (buying, planning). Run the comparison on total cost per order at your real package mix, not headline pick fees.

What is a good pick accuracy rate? Establish your own baseline first — errors caught per hundred orders over a month — and improve against it rather than chasing a published benchmark you can't verify. The number's purpose is trend detection: a rising error rate points at a process change (new SKUs, new space, new picker) that needs a look, before customers start reporting it.

Where to go next

Run one honest week of measurement: pick time per order and errors caught per hundred orders, whatever your current method. Then red-team your own process against the five attacks above — two or three will land, and each one that lands is a fixable system gap, not a people problem. If the numbers say volume has outgrown the bench, start pricing partners; if they say the process is the problem, the SOP above is the fix.