Field notes · Depletion data

What is depletion data, really?

By the Sightglass team6 min read5 June 2026

A depletion is one case leaving a distributor's warehouse for a retail account — not a sale to a consumer. It is an inventory movement recorded on an invoice: the distributor sold product on to a bar, restaurant, or store, and the case came off their floor. Whether anyone poured it, bought it, or took it home is a separate question the file has no opinion about. Most of the confusion around depletion reporting starts right here, and most of the bad decisions follow from it.

The distinction matters because the whole file inherits it. Everything depletion data is genuinely good at follows from what a depletion is; everything it can't answer follows from what it isn't. What follows is the plain version — who reports what to whom, why the numbers arrive late, and where the honest edge of the data sits.

Shipments, depletions, and the gap between them

Two different numbers get called "sales" in this trade, and people mix them constantly. Shipments are what you sold the distributor — product leaving your warehouse or the port, invoiced to the house. Depletions are what the distributor sold on, case by case, to retail accounts. Most beverage alcohol in the U.S. moves supplier to distributor to account this way, though who may sell to whom, and whether the state itself sits in the chain, varies by state.

Both numbers are real. They measure different things and can move in opposite directions for a whole quarter. The difference between them is sitting in the distributor's warehouse: a heavy shipment month followed by a flat depletion month means inventory, not demand, and that inventory has to clear before the next order comes. Depletion is the closer of the two to what the market actually wants. It still isn't demand.

Check what a "case" means in your file before you trend anything. Some feeds report physical cases, some report equivalent cases normalized to a standard volume. Mixing the two across pack sizes will quietly bend a year-over-year comparison, and nothing in the file will warn you.

Who reports what, to whom

A depletion is a record of an invoice. It is worth knowing the path that invoice takes before it reaches your screen, because every step in it explains something you'll later wonder about.

01An account places an order — a rep writes it on a visit, the buyer calls it in, or it comes through the distributor's ordering portal.
02The distributor enters the order, picks it, and invoices the account. That invoice is the depletion. Its date is usually the date you'll see in the file — not the day the product reached the bar.
03The distributor's system rolls the day's or week's invoice lines into a supplier-facing feed. In this trade that feed commonly runs through VIP.
04Lines are mapped to your products, your account numbers, and your territories — which is where the same bar carried under three slightly different names starts to matter.
05You pull the export, or it lands with you on a schedule.

Cadence is set by the distributor, not by you. Some houses report daily, some weekly, some effectively monthly at close. One export can therefore hold accounts current to yesterday sitting next to accounts whose latest activity is three weeks stale, and the row itself rarely tells you which is which. If some of your houses report slowly, learn which ones — it changes how you read a quiet month.

Why it lags

In the ordinary case the data is a day or two behind reality, and longer around a period close. That isn't anyone failing at anything. It's the cost of a number that has to pass through several systems before it is yours.

Order entry. An order written on Tuesday afternoon may not be keyed and invoiced until Wednesday.
Batching. Feeds are compiled on a schedule, not the instant an invoice posts.
Mapping and aggregation. Lines from dozens of houses have to be reconciled to one set of products and accounts before the file is usable.
Period close. Credits, returns, and adjustments land after the fact, so a closed month can restate slightly. The figure you quoted on the 2nd may not be the figure on the 10th.

The practical rule: don't read the last few days of any export as signal, and don't call an account lost on one thin period. Wait for the second one. An account that ordered every month for a year and then goes quiet twice is telling you something. One quiet period is usually a calendar.

What it is genuinely good for

Within its limits the file is very good — accurate, granular, and account-level, which is more than most industries get. The cuts that reliably earn their keep:

Account-level trend. One account, cases by month, twelve months across. Growing, flat, or fading, in a number you can put in front of a buyer or a distributor without anyone arguing about it.
Distributor activity. Which houses are moving cases, which have gone quiet, and how long since the last sale by house. A distributor with no depletion in ninety days is a fact, not a feeling.
Brand and market performance. The same beer behaves differently in two states with two different houses. The file separates them cleanly.
Distribution breadth. How many accounts took a given SKU this period versus last — pickups and losses, held separate from volume. Volume can hold up while the account count quietly erodes underneath it.
Territory review. Every rep's book side by side, on numbers nobody has to self-report.
Program verification. Whether a promotion actually reached accounts, counted in cases rather than in a summary of intentions.

What it can never tell you

The limit
The file records what moved, in cases, to which account, on what date. It contains no reason. Every why in this business is still a phone call — the data's only job is to tell you which call to make first.
What's on the shelf right now. A depletion is the moment product left the distributor's warehouse for the account. Whether it's in the cooler, the back room, or on tap is not a depletion — it's a stock question, and a different report answers it.
What the consumer bought. A keg depleted on the 3rd may pour for six weeks. A pallet delivered before a holiday may clear in two days. Depletion timing and consumption timing are only loosely related.
Whether a big month was demand or a load-in. First orders, resets, and program buys all look like growth until the following month declines to repeat them.
Why an account stopped. Closed, sold, new buyer, lost the handle, moved to a competitor, or the distributor was simply out of stock — all identical in the file. All different conversations.
Anything at all about accounts that ordered nothing. A transaction file can only list transactions. An account that stops entirely doesn't post a zero; it stops appearing.

That last one is worth sitting with, because it is the one that costs money. We counted the accounts in one importer's export:

13,109
accounts listed in a single export
5,085
accounts absent from it that had ordered the prior year

No row, no zero, no flag. The only way those accounts surface at all is by comparing the current file against older files somebody remembered to keep.

Reading it every night

All of this can be done by hand, and plenty of good sales operations do it. The obstacles are volume and memory: around thirty thousand rows in a single export, and the fact that any one export is a snapshot of a single period — so a year-over-year comparison depends on an archive somebody has to keep.

That's the part Sightglass automates. It reads the same VIP feed you already pay for, every night, and keeps each snapshot so the history outlives any single export window. It flags the accounts slipping, losing SKUs, or gone quiet — and the ones winning — and drafts the outreach in your voice, one message to the account and one to the distributor. Nothing sends without a person approving it, and approved mail goes from the rep's own mailbox, so replies come back to them. It arrives in Slack, in an email digest, or on the dashboard. Your VIP subscription and your reporting stay exactly where they are.

And because the data lags, so does the system. If the file can't yet support a conclusion, it says nothing rather than flagging a gap that may not be there.

Want to see it on your book? Twenty minutes, screen shared. No file required. Book a 1-1 demo