How Inaccurate Forecasts Drain Small CPG Brand Cash Flow

The distributor's purchase order came in at 2,400 cases. The brand built 2,400 cases. By March — three months after the promotional period closed — 900 of those cases were still sitting in the DC. Consumer scans at retail had been flat throughout.

The brand read "distributor not reordering" as "demand fell." What had actually happened: the distributor stocked forward before the promotion, ran the inventory down over eight weeks, and would not need another order for six more. The brand's demand view was running off shipment records in the ERP — the number of cases shipped to the distributor's DC — not off retail scan data. Those are two different signals, and they told two different stories.

For a brand doing $10M in retail revenue, that gap has a direct cost: excess inventory carrying 20–30% of its value annually in holding charges, or a short production run that misses fill rate and generates a compliance fine. IHL Group's tracking of global inventory distortion puts the combined annual cost of overstocks and out-of-stocks at $1.77 trillion. The dynamic starts at the forecast, not at the DC.

Distributor orders are not demand

A brand's ERP records one thing cleanly: how many cases left its warehouse headed to a distributor's DC. That number is accurate. It is also the wrong signal for forecasting consumer demand.

Distributors manage their own inventory positions. Before a promotional period, a distributor stocks forward — placing a larger-than-usual order to cover anticipated retail demand and protect against supply disruption. After the promotion runs, the distributor draws down that forward inventory before reordering. A brand reading the drawdown period as reduced consumer demand will cut production at precisely the moment the next promotional cycle is approaching.

This is the bullwhip effect: small fluctuations at the retail shelf create amplified swings in distributor order volume as the signal moves upstream. For a large brand with thousands of retail doors across dozens of distributors, the swings average out. For a brand with one or two major distribution relationships, a single distributor's inventory timing decision can resemble a quarter-over-quarter demand collapse. The signal is not wrong. The interpretation is.

The actual demand signal lives in the portals. Retail Link — Walmart's supplier portal — provides weekly sell-through data by item and store. UNFI Connect publishes velocity reports for UNFI-distributed items. KeHE CONNECT has equivalent data. Each shows what consumers actually purchased, not what a distributor decided to order.

The promotional period distorts both signals

Trade promotions create two distinct buying events. The first is the distributor buy-in: a large forward order placed ahead of the promotional window. The second is the consumer buy-through: retail scans rising during the promotion itself. These two events happen on different timelines and produce different volume curves.

A brand that builds its next production run off the distributor buy-in has confused the first event for the second. Distributor shipments during a promotional build can spike sharply; the actual scan lift at retail may be considerably more modest. The brand built to the shipment spike, not the scan lift, and the excess inventory sat in the DC until the next promotional period arrived or the write-off window closed — whichever came first.

The promotional period also generates the reverse error. After the promotion closes, the distributor stops reordering while it draws down forward inventory. A brand that reads the absence of a purchase order as a demand signal will underforecast the next cycle — short production run, short shipment, short shelf.

That timing gap runs through the trade spend accounting as well. Promotional commitments are agreed to before demand plays out, and deductions settle weeks or months after the event — the same lag that makes the gross-to-net bridge hard to build. A forecast built on scan data would at least align the production view with what is selling at retail, rather than with what the distributor bought in anticipation of it.

Forecast error costs money in both directions

The overforecast leaves excess inventory in the DC. Inventory carrying costs run 20–30% of inventory value per year, accounting for storage, insurance, opportunity cost on tied-up capital, and the labor of managing aging stock. For specialty food with expiration windows, the exposure extends further: according to Spoiler Alert's analysis of CPG liquidation programs, only 47% of excess inventory is sold through normal channels. The rest is discounted, donated, or written off. The brand that overforecast by 30% did not merely tie up cash in pallets — it eventually moved that excess at a margin the channel model never assumed.

The underforecast is harder to see until it has already cost something. A production run that falls short means the brand cannot fill a distributor purchase order at the required fill rate. When fill rate drops below threshold, that is an OTIF event that generates a compliance deduction. The short-shipped product also produces a stockout at the retail shelf. Out-of-stocks cost the global retail economy $1.2 trillion in annual lost sales — and in the natural and specialty channel, a stockout registers on the velocity number buyers use in line reviews. A few weeks of phantom availability drops units per store per week below the threshold that triggers a facing reduction or delist.

The overforecast is visible — inventory accumulates and the operations team can see it. The underforecast is invisible until the chargeback arrives or the buyer calls. Both cost the same money. Only one announces itself.

Scan data is already in your distributor portals

A SPINS subscription — the syndicated data service aggregating point-of-sale data across natural, specialty, and conventional retail — runs from a few thousand dollars for basic access to $30,000 or more for meaningful channel and panel coverage. For a brand at $8M–$15M in revenue, that subscription is worth evaluating. It is not a prerequisite.

The portals are already available. Walmart Retail Link is free to Walmart suppliers and provides weekly sell-through by item and store — actual consumer scan data. UNFI Connect publishes velocity reporting within the distributor relationship. KeHE CONNECT has equivalent data. A brand with distribution at Walmart or through UNFI or KeHE has access to at least one of these sources. Most are not using them as the primary demand input. The data exists. The workflow does not.

The forecast built from portal data is not perfect. Distributor velocity reports carry lag. Not every retail partner publishes self-service reporting. But it is categorically more accurate than a forecast built from the brand's own shipment records, because it tracks what consumers bought — not what the distributor decided to order.

Build the demand model before the production run

Lailara builds demand signal reconciliation for brands whose forecast is running off shipment data — distributor orders mapped against scan velocity from UNFI Connect, KeHE CONNECT, and Retail Link, with promotional lift separated from baseline. The deliverable is a clean demand view by SKU and channel, with working capital exposure quantified against the current inventory position. Book a 30-minute scoping call.

See the methodology behind this post. The worked example — 95% internal vs. 86% retailer-scored fill rate, $433,000 in annual OTIF exposure, root-cause attribution across on-time and in-full failures — is a live demo you can open and explore. Fulfillment & OTIF Diagnostic →

The Ten Decisions is the map behind this post. Every data problem a $25M specialty food brand runs into — chargebacks, deductions, launch economics, OTIF gaps — maps to one of ten decisions being made without adequate information. See the full picture →