In short: A principal forecasting from its own invoices is forecasting shipments into a distributor rather than demand out of one, and the difference between the two is the distributor's inventory swing. Usable reporting needs sell-out by item and period, closing stock at a stated date, and a mapping from distributor item codes to yours, all in consistent units. The inventory identity, opening stock plus purchases minus sales equals closing stock, catches most unusable files before anyone builds a forecast on them. Distributors have weak incentive to report well unless the principal gives something back, such as a forecast they can order against or funding for the system that produces the data.
A regional demand planner has clean shipment history for fourteen markets and secondary sales for three of them. In the other eleven she is forecasting her own invoices. Once a month a spreadsheet arrives from each distributor with a stock figure as at the last day of the month, which lands somewhere around the eighteenth of the following month. The item codes are the distributor's own. Two markets report in cases, one reports in bottles, and one has been sending the same closing stock number, plus or minus a rounding, for five months.
She has asked for better. What came back, in three of the eleven, was a new tab on the same spreadsheet.
The gap the principal is trying to close
The structure of the relationship creates the problem. The principal sees what went into the distributor and would like to see what came out. Between those two events sit the distributor's own inventory decision, their credit position, their working capital priorities across every principal they represent, and their sales force's judgement about which brand to push this month.
That means the shipment series is a record of the distributor's purchasing behaviour rather than of market demand, and it carries the fingerprints of the principal's own quarter end incentives as well. The consequence for planning is that the principal is forecasting a series that partly measures their own commercial pressure. Modelling around that gap, once you have both series, is a separate method with its own post on this site. This one is about the prior question, which is how you get the second series at all and keep getting it.
Worth being clear about where the value sits, because it is easy to sell this internally on the wrong benefit. Forecast accuracy in a reporting market does improve, usually modestly. The larger return is diagnostic: when orders fall, you can tell whether consumers stopped buying or the distributor is working stock down, and those two situations call for opposite responses at a point in the year when you have about two weeks to choose.
What the reporting set has to contain
A workable set is shorter than most principals ask for and longer than most distributors send.
Secondary sales, meaning what the distributor sold on, by item, by period, split by customer or channel at whatever level they genuinely maintain.
Closing inventory by item and by physical location, with an explicit as-at date rather than an implied one.
Goods in transit and goods received, so you can reconcile their receipts against your shipments and find out whether a discrepancy is timing or something else.
Sub-distributor or wholesaler splits where the distributor sells on to a second tier, because a sale into a wholesaler is a shipment rather than a sale, and treating it as consumption reintroduces the exact problem you were trying to solve one level down.
Batch or expiry position where the category has dated stock, since the closing quantity means something different when a third of it expires in six weeks.
On frequency, the honest answer is that the calendar discipline matters more than the interval. Weekly stock and sales is right for fast moving categories with short shelf life, monthly is the realistic floor everywhere else, and a monthly file that always arrives on working day three is more useful than a weekly file that arrives whenever someone remembers. Set a latency target and treat it as the primary service measure: a complete and accurate report that lands on the eighteenth is a historical document, because the decisions it would have informed were made on the fifth.
Take completeness over granularity at the start. Every item at monthly, covering the whole portfolio, supports a plan. Four hero items at weekly does not, and the effort to get the four is most of the effort to get all of them.
Why the distributor's incentive is weaker than you assume
Principals tend to treat reporting as an administrative task that a reasonable partner would simply do. From the other side it costs staff time in a business that runs lean, and it gives away three things that have commercial value.
Their true stock position, which affects how hard you push them to order and how much negotiating room they have at quarter end. Their customer list and channel mix, which is the asset that makes them expensive to replace. And their pricing and margin structure by channel, which is visible in the shape of the data even when the price fields are left out.
There is also a direct conflict at specific moments. A distributor sitting on high stock who expects a push to take more before your quarter closes has an immediate reason to report a softer inventory number. Nothing about that requires bad faith, just a fairly ordinary commercial instinct, and it explains why inventory accuracy tends to degrade at exactly the points in the year when you most need it.
What changes the behaviour is making the report an input to something they want. Four mechanisms work, and the second is the strongest.
Allocation priority during shortage, based on reported sell-out. When supply is tight, markets with credible sell-out data get served first because the principal can defend the decision. Announce it in advance and apply it once, and the reporting quality across the network improves within two cycles.
Planning support that reduces their working capital. A suggested order that holds their service level while taking a week of cover out of their warehouse is worth real money to a distributor financing their own stock, and it can only be produced from the data you are asking for.
Trade investment allocated on reported data. Promotional funds and marketing support distributed against what markets can evidence rather than what they assert.
A rebate or scorecard component on reporting itself, measured on completeness and timeliness rather than accuracy. This one has a trap in it, discussed below, and it works only alongside the others.
Standardising across markets that all count differently
A network of twenty or thirty distributors will have twenty or thirty item coding schemes, several calendars, and no shared definition of the word stock.
The mistake is to solve it in their systems. Asking thirty independent companies to change their ERP configuration to suit a principal produces a two year programme, a lot of resentment, and partial compliance. Map at your boundary instead. Publish a fixed file layout, accept their own codes in it, and maintain the translation yourself: their item code to your GTIN, their location to your GLN, their calendar to yours, their unit of measure to yours with the conversion factor held on your side and versioned.
Reuse the definitions that already exist rather than inventing them. The X12 852 product activity transaction and the EDIFACT SLSRPT and INVRPT messages have settled definitions for exactly these fields, built over decades of arguments between trading partners. Even where a distributor cannot send structured messages and will be emailing a spreadsheet for the foreseeable future, borrowing the field semantics from those standards saves you from relitigating what counts as a sale.
Write the definitions down once and circulate them with the template. Does stock include goods in transit and goods in bonded warehouse. Is a return a negative sales line or a separate movement. Is a sale to a related company a sale. Does the period end at the calendar month end or at the end of the distributor's fiscal week. Most disputes about data quality in these networks turn out to be definitional, and they persist because nobody wrote a page of definitions at the start.
Then reconcile, monthly and mechanically. Your shipments against their reported receipts, allowing for transit time. A stable gap is a definition problem. A growing gap is something else, and both are worth knowing before the annual stock count finds them.
What the principal has to give back
Reporting that survives is reciprocal, and the principal usually has more to give than they realise, because they can see thirty markets and the distributor can see one.
Category insight is the standard currency. What the category is doing in comparable markets, which pack formats are growing where, what a promotional mechanic delivered in a market at a similar stage, how their own performance compares to the network on measures they cannot compute alone. A distributor can use that in conversations with their own retail customers, which is where it converts into something they value.
Planning support is the more concrete gift. If you are receiving stock and sell-out data, you can produce a suggested order and a cover recommendation, and hand back a working replenishment proposal each cycle. Handing over the replenishment decision entirely is a different arrangement with its own economics, covered elsewhere on this site; the version here stays advisory, and it earns its keep by taking cover out of the distributor's warehouse without hurting availability.
Two further things cost you nothing. Earlier visibility of your own supply constraints, so a distributor can plan around a shortage instead of discovering it. And a fixed slot in the calendar. If the data comes monthly, the insight goes back monthly, in the same meeting, with the same person accountable for both halves.
The limit
Reporting quality tracks the commercial relationship, and a mandate with no benefit attached produces compliance shaped data.
This is the failure mode to watch for, because it defeats every automated validation you would think to build. Files arrive on the agreed working day, every field is populated, nothing is missing, and the data was derived rather than measured. Closing stock is last month's closing plus receipts minus a smoothed sales figure. The channel split is a fixed percentage applied to the total. The arithmetic reconciles perfectly because it was constructed to. That data is worse than an empty file, because an empty file starts a conversation and this one passes your checks and enters your planning model.
Three tests find it. Run the inventory identity on each market, opening plus receipts minus reported sales against reported closing, and look for the markets where it ties exactly, month after month, to the unit. Real physical stock does not do that, and a perfect reconciliation over six periods means the closing figure was calculated rather than counted. Look at the variance of the channel split over time, because a real split moves with the season and a derived one sits within a point of itself all year. And ask whether the data has ever surprised you, since a market that has never reported anything you did not expect is not reporting.
Two honest qualifications. Some distributors genuinely cannot report at the standard you want, because a small operation running on a paper ledger and one bookkeeper has a capability gap rather than an attitude, and the answer there is a simple tool, a lower bar, or funding the headcount. And in a market where the distributor is the only viable route and knows it, the balance of power runs the other way, so buying the data by funding a distributor management system is cheaper than issuing a mandate that will be satisfied on paper and nowhere else.
Take last month's file from your five largest markets, run the inventory identity on each, and start with the markets where it reconciles exactly.