How Sourcing Agents Keep Production Visible Across Brands and Factories

What production visibility means for a sourcing agent

For a sourcing agent or buying office, production visibility is the ability to answer — for any client brand, any factory, and any live purchase order — where the order stands against its committed delivery date, at size-level detail, without picking up the phone. It is the difference between telling a brand "we're checking with the factory" and telling them "sew completed Tuesday, final QC is Friday, ex-factory holds, and here is the size breakdown."

This guide is written for the agency side of the table: the firm coordinating production across multiple client brands and multiple factories at once. A brand's own in-house teams face a related but structurally different problem — one brand, many vendors — and that story is covered separately for sourcing teams and production teams. What follows is about the practices themselves, not any particular software.

The grid problem: N brands by M factories

A brand tracking its vendors has a list. An agent has a grid — every client brand crossed with every factory that produces for it — and the grid compounds on both axes. Add a client and you inherit their calendar, their size curves, their delivery terms, and their expectations about what a status update looks like. Add a factory and you inherit its T&A conventions, its language for the same milestones, and its particular way of being late.

The grid has properties a single-brand tracker does not:

• →The same factory serves multiple clients. A capacity problem at one factory is simultaneously a delivery risk for three brands, but each brand sees only its own orders. The agent is the one party positioned to see the collision coming.
• →The same week matters differently to different clients. One brand's ex-factory slip of five days lands inside its consolidation window; another's misses a vessel and turns into air freight or a cancelled delivery.
• →Priority conflicts are settled by whoever is watching. When a sewing line falls behind, which client's order slips is decided on the factory floor — unless the agent has the visibility to negotiate it deliberately.

None of this is exotic. It is the ordinary operating condition of a buying office. The question is whether the information structure matches it, and in most agencies it does not: the information is organized the way the clients ask for it, which is the root of the next problem.

The one-format-per-brand reporting trap

Every client brand has a WIP report format it expects. One wants a weekly Excel grid by PO with a traffic-light column. Another wants status by delivery, rolled up to the drop. A third has a portal that wants dates keyed in one order at a time. The path of least resistance is obvious and nearly universal: maintain one tracker per brand, in that brand's format, and update each one as news arrives from the factories.

This is the trap. Not because any single template is unreasonable, but because of what the arrangement does to the data:

• →The report becomes the record. There is no version of the truth other than the client-facing documents, so the agency's own operating picture is scattered across N files shaped by other people's preferences.
• →Every fact is keyed N times. A factory reports the cut milestone on twelve POs spanning three clients; that update is now three data-entry tasks in three formats, each an opportunity for a transposed date or a missed line.
• →Cross-client questions become research projects. "How loaded is this factory in October across all our clients?" has no file to open. Someone builds it from scratch, and it is stale by the time it circulates.
• →Friday disappears. Merchandisers spend the end of every week reformatting the same facts for different audiences — hours that produce no new information.

The structural fix is not better templates. It is inverting the relationship: one internal record, many generated views. The record is the order book; the client formats become outputs.

What a single order book looks like

The order book is the agency's consolidated record of every live PO across all clients and factories. Its exact shape varies, but the versions that work share five layers:

Two of these layers do most of the work and are most often missing.

Size-level detail is what separates a tracker from a reassurance machine. An order can be on time at the unit level and broken at the size level — the factory cut short on the tails of the curve, packed incomplete runs, and the tracker's quantity column shows green all the way to the brand's DC. Agencies that carry size curves in the order book catch this at packing; agencies that track units find out from a claim. The downstream arithmetic of what an incomplete size run does to a wholesale order is covered in order fill rate in wholesale apparel.

Computed exception flags are what make the order book usable at scale. A merchandiser responsible for two hundred live POs cannot re-read two hundred rows every morning. If flags are raised by the data — a projected date moving past plan by more than the threshold — the daily working view is the exception list, and quiet orders stay quiet. If flags are typed in by hand, they reflect who was diligent last week, which is a different thing from where the risk is.

Standardizing WIP milestones across factories

The order book only stays comparable if every PO reports against the same milestone set — and factories do not naturally cooperate with that. One factory's T&A has thirty lines including lab dips and trim approvals; another reports "in production" until the week it ships. Asking every factory to adopt the agency's format has roughly the success rate of asking every client to drop theirs.

The workable approach is translation, not enforcement. Define a canonical milestone set for the order book — small enough to be reportable, rich enough to catch trouble early. The natural choice is the six standard milestones covered in WIP Reporting Between Factories and Brands:

1. →Fabric booked
2. →Cut
3. →Sew
4. →Finishing
5. →Final QC
6. →Ex-factory

Then map each factory's own stages onto the canonical set. The factory keeps its internal T&A exactly as it is; the agency's merchandiser translates at the point of update. A factory that reports "stitching 80%" has not yet reached sew; one that reports FRI booked is approaching final QC. The mapping is written down once per factory and applied consistently, so a milestone means the same thing in the order book regardless of which factory it came from.

Two disciplines make the standard hold. First, every milestone carries a date pair — planned and actual-or-projected — because a milestone without a plan date cannot slip, and a milestone that cannot slip cannot warn you. The plan dates come from the T&A built at order placement; constructing that calendar properly is its own subject, covered in depth in how to build a T&A calendar on retail-plan.com. Second, a fixed update cadence — twice a week is the common rhythm — because milestone data that arrives whenever the factory feels like it is not a tracking system, it is an anecdote collection.

Reporting to each brand from one source

With the order book in place, client reporting inverts. Each brand's format stops being a maintained document and becomes a filtered view generated from the record: this client's POs, their milestone status, their exceptions, arranged the way that client likes to read it. The facts are keyed once, when the factory reports; the formats are produced on demand.

The practical gains are larger than the saved Friday afternoons:

• →Consistency across audiences. The brand, the agency's management, and the factory follow-up list are reading the same underlying dates. Nobody is explaining why two reports disagree.
• →Exceptions lead. A generated report can open with what changed and what is at risk, because the flags are computed. Hand-maintained reports tend to open with everything, which trains clients to skim.
• →The conversation moves up a level. When the status data is assumed to be current and correct, the weekly call stops being a data-reading exercise and becomes what the agency is actually paid for: judgment about what to do with the delay, the short-cut fabric, the capacity squeeze.

There is also a quieter benefit. An agency that reports from a single record can answer the questions brands increasingly ask in reviews — on-time percentage by factory, slip causes, milestone variance trends — from evidence rather than recollection. That is the same data brands use to grade their vendors; keeping an internal vendor scorecard view of your own factory base means seeing yourself the way your clients do before they show you.

When spreadsheets stop scaling — and what to look for instead

A disciplined spreadsheet order book is a genuine system, and for a small agency it can run well; a free production WIP tracker template is a reasonable starting structure. The failure is not the spreadsheet format — it is what happens to any hand-maintained record past a certain scale. The signals that the point has been passed are consistent:

• →Version conflict as a routine event. Two merchandisers answer the same delay question differently because they have different files open.
• →Reporting hours exceed chasing hours. The team spends more time reformatting known facts than discovering new ones.
• →Size-level errors surface downstream. Broken size runs are caught at the brand's DC, not in the tracker, because size detail was the first thing dropped when the workload grew.
• →No reconstructable history. A client disputes a delivery date and nobody can show when it moved or who agreed.

Past that point, the question becomes what system to move the order book into, and the evaluation criteria follow directly from the practices above rather than from any vendor's feature list. Whatever the system, it should hold one record across all clients and factories with client-scoped views, so confidentiality between brands survives consolidation. It should support a canonical milestone set with per-factory date pairs, computed exception flags, and size-level quantities as first-class data rather than a comments column. Factory updates need a path that assumes no factory ERP on the other end — something a floor merchandiser can complete from a phone in minutes, or the cadence dies. Reporting should generate per-client outputs from the record instead of asking the team to keep maintaining templates. And every date change should leave an audit trail. A system that fails these criteria recreates the spreadsheet problem at a higher subscription price.

Related resources

• →Production Visibility — how WIP status connects back into a merchandising plan on the brand side
• →PO Coverage — the formula for how much of a plan is covered by placed orders
• →Lead Time — the glossary entry, including how lead-time bands shape order placement
• →Planning Against Fabric Minimums — the mill-minimum arithmetic that sits upstream of fabric booked dates