Forecast, fulfilment, and the chargebacks nobody warns you about.

System 03 of six · Read time about nine minutes

A retailer is a logistics business that happens to sell things. Every requirement in a vendor manual exists because at some point a supplier cost them money by not doing it. When you fail one, the response is not a conversation — it is an automated deduction, applied without discussion, and the second one is more expensive than the first.

The brands that struggle here are almost never careless. They are brands whose operations were built to do one thing extremely well — ship parcels to consumers — and who assumed that shipping to a distribution centre was the same activity at a larger size. It is a different activity with a different failure mode.

The forecast has to have an owner with a name

Start here, because everything downstream depends on it. A forecast that lives in a spreadsheet nobody owns is not a forecast; it is a record of what someone thought in March.

A working forecast has four properties: a named owner, a stated cadence at which it is revised, a documented method, and a history you can look back at to see how wrong you were and in which direction. That last one matters more than accuracy. A forecast that is consistently 20% high is far more useful than one that is randomly right, because you can correct for a bias and you cannot correct for noise.

At retail the forecast also changes shape. DTC forecasting is site-level and demand-driven. Retail forecasting is door-level and allocation-driven: units per store per week, multiplied by doors, plus pipeline fill for the initial set, plus promotional lift, minus the returns you will actually receive. If your planning process cannot produce that shape, it will not survive the first reset.

Read the vendor manual before you sign anything

Retailers publish their requirements. The document is long, dry, and almost never read cover to cover by the supplier who agreed to comply with it. Inside it, in plain language, is every deduction you are about to incur.

Extract at minimum:

• Routing instructions — who books freight, at what threshold, through which portal, with how much notice
• Labelling standards — carton labels, pallet labels, GS1 barcode requirements, placement tolerances
• Pallet specification — type, height, overhang, tie and high, shrink wrap
• ASN requirements — format, timing, and what counts as an inaccuracy
• Appointment scheduling — lead time, window tolerances, penalty for missing one
• On-time and in-full expectations, and the measurement window used
• The deduction schedule itself, by violation type

Turn that into a checklist with a named owner per line. This single exercise prevents more money leaving the business than any negotiation you will have.

EDI is a translation layer, not a technology project

Electronic data interchange sounds like an IT problem, which is why it is usually handed to whoever seems most technical. It is really a data-quality problem. The documents themselves are stable and well specified: purchase order in, acknowledgement out, advance ship notice, invoice, and typically some form of sales or inventory reporting back to you.

Where it breaks is at the joins. Item numbers that do not match between your system and theirs. A unit of measure that means something different on each side. A ship-to location that was set up twice. Costs that were updated in one system and not the other. None of these are complicated; all of them generate deductions.

Two practical decisions:

1. Full-service versus self-service. A managed provider costs more per month and absorbs the mapping work. A modern API-first provider costs less and expects you to have someone who owns it. If nobody at your company owns it, the cheaper option is the more expensive option.
2. Test with real data before go-live. Not sample data. Real item numbers, real costs, real ship-to locations, real pack configurations. Almost every launch-week failure is something that would have shown up in a real-data test.

Fill rate is a reputation, not a metric

A buyer forgives a slow start. They do not forgive unreliability, because unreliability makes them look bad to their own organisation. Below roughly 95% fill you stop being a supplier with a problem and start being a supplier who is a problem — and the distinction between “your product experienced stockouts” and “you caused stockouts” is one the retailer does not draw.

This matters especially when a distributor sits between you and the retailer. Their fill-rate history becomes your fill-rate history. If you have not asked them what it is for this specific retailer, you are carrying a risk you have not priced.

Build the exception path before you have an exception

Something will go wrong. A container will be late, a lot will fail inspection, a promotion will oversell, a deduction will be applied that you believe is incorrect. What separates brands that recover from brands that spiral is whether the path was defined before it was needed.

A defined exception path answers, in advance:

• Who at your company is the single point of contact for this retailer, by name
• Who at the retailer they contact, and through which system rather than which inbox
• What the internal decision rule is for expediting freight — at what cost, approved by whom
• Who reviews deductions, on what cadence, and what the threshold is for disputing one
• How the customer service or account team is told, so nobody promises something that cannot be delivered

The deduction review is worth calling out. Deductions frequently run at 3–8% of gross revenue in a first retail year, and a meaningful share of them are disputable within a window that closes. Brands that never review them are effectively funding a permanent discount they did not agree to.

Prove it at one account first

The single most useful thing a brand can do before a major placement is run the whole operational chain once, at small scale, at one account. Not a pilot in the marketing sense — a real shipment, on a real routing guide, with real labels and a real ASN.

You will find three or four problems. All of them would also have occurred at four accounts simultaneously, at ten times the volume, in the eight-week window where a buyer is deciding whether to keep you. Finding them early costs a few hundred dollars in freight. Finding them late costs the placement.

What good looks like

A brand with a working operations system can produce, on request: a door-level forecast with an owner and a revision history; a vendor-manual checklist with names against each requirement; evidence of a successful end-to-end test shipment; a fill-rate figure they can defend, including their distributor’s; and a one-page exception path that somebody other than the founder could execute.

None of that is impressive. All of it is rare, and every buyer notices its absence.

Next in the series: System 04 — vendor requirements, product compliance, and naming an owner for each.