AI FOR DEPLETION DATA

Can ChatGPT analyze a depletion report?

Yes. ChatGPT can read a CSV or Excel depletion export and run code on it to do the math. The hard part for any tool is the beverage detail: beer case equivalents versus wine and spirits 9-liter cases, keg sizes, pack notation, returns and deposits, and distributors that define a depletion differently. Tell it your standards, then check the totals.

Last updated September 29, 2026

What can ChatGPT do with a depletion report?

ChatGPT works well for a one-off look at a single file. For data-analysis tasks it can write and run Python code on the file you upload, so it can filter, total and compare the rows.

For repeat work, a ChatGPT project keeps files and instructions together, so a monthly report can reuse the same setup. OpenAI’s guide to projects suggests using one to “return each week to generate updates without starting over.”

Some people in the trade already do this. In a September 24, 2026 Wine Industry Advisor editorial, Paul C. Ringgold writes that Greg Shepard, a Certified Sommelier and wine sales executive, uses AI to “analyze distributor sales data” and “prepare reports.”

How to analyze a depletion report in ChatGPT

1. Export the report as .xlsx or CSV. Choose the detail layout, one line per account, product and date, if your system offers it. A layout without subtotals is easier to check.
2. Upload the file and paste the standards block below before your first question. In a ChatGPT project, you can save the block as the project’s instructions instead.
3. Ask one question at a time. Start with the prompts below and replace the bracketed parts.
4. Check the answer before you share it. The four checks are in the next section.

Standards block to paste

Use these rules for every answer about this depletion report.

1. Report beer in case equivalents (CE): 1 CE = 24 × 12 oz = 288 fl oz = 2.25 US gallons. A US barrel is 31 gallons. A keg’s CE is its gallons ÷ 2.25.

2. Report wine and spirits in 9-liter cases: 1 case = 12 × 750 mL = 9 liters. Keep each category in its own unit.

3. Treat a column labeled Cases or CS as physical cases and convert it from the package. Treat a column labeled CE, Eq Cases, 9L or 9LE as already converted, but check three rows against units × package size and tell me if they don’t match.

4. Keep returns and credits negative. Never use absolute values.

5. Exclude keg and bottle deposits from volume. They are money, not liquid.

6. List breakage, samples and transfers separately, and ask me whether my distributor counts them as depletions.

7. Drop total and subtotal rows before adding anything up.

8. Show every conversion you make next to the original product and package text. If a package is unclear, say so instead of guessing.

9. Calculate every total in Python, and give totals by distributor and by package.

The unit tables behind these rules are on Case equivalents. The full reading checklist is in How to read a depletion report.

Five prompts to start with

1. Accounts that stopped ordering

   The report runs through [end date]. List accounts that bought in the 12 months before the last 90 days but have no retail sales in the last 90 days. For each, show the distributor, last order date, last product and package bought, and volume over those 12 months in CE or 9-liter cases. Tell me how you matched account names.

2. Year over year by distributor and package

   Compare [months] this year with the same months last year, by distributor and package. Show beer in CE and wine and spirits in 9-liter cases, with the change and percent change. Put any distributor or package with volume in only one year at the bottom, and say which year it is missing from.

3. Convert to case equivalents and 9-liter cases

   Convert every line to CE (beer) or 9-liter cases (wine and spirits). Make a table with the original product and package text, quantity, unit, the package you read, the conversion factor and the result. Mark any package you could not read instead of guessing.

4. Returns, credits, samples and breakage

   List returns, credits, samples, breakage, transfers and deposit lines separately from retail sales. For each type, show the number of lines, the volume, the dollar amount and the column or text you used to identify it.

5. Top accounts by change

   Compare [this period] with [prior period]. Rank the 20 accounts with the largest volume increase and the 20 with the largest decrease, using net volume after returns. Show each account’s distributor, both periods’ volume and the change.

How do you check ChatGPT’s depletion numbers?

Run these four checks before you share a number.

1. Ask it to show every conversion. Scan the list for misread packages, such as a 1/6 BBL sixtel read as a six-pack or a 50 L keg read as a half barrel. See the notation traps.
2. Compare the grand total with the report footer. Ask for the total of the raw quantity column with subtotal rows removed, and compare it with the report’s grand total. If they differ, find out why before you trust anything converted from it. The worked example shows how subtotals and returns throw a total off.
3. Hand-check three rows. Pick a keg line, a case line and a return, and work each one out with the case equivalent calculator.
4. Re-sort and rerun. Sort the file a different way, for example by account instead of by date, upload it to a new chat and ask the same question. The totals should match. If they don’t, ask it to explain the difference.

OpenAI’s data analysis guide gives similar advice: review the generated code, outputs and assumptions before relying on the result. If a result seems incomplete, it suggests asking ChatGPT to inspect specific sheets, rows or columns, or splitting the file. In Ringgold’s editorial, Daniel Webber, wine director at Wrigley Mansion in Phoenix, adds that “you still have to know the subject well enough to recognize when something doesn’t look right.”

These checks apply to any AI tool, including BevBeacon.

Can ChatGPT handle monthly reports from several distributors?

Yes, but it takes more setup than one report and one question. Three things add work.

• Several distributors. Each distributor assigns its own item codes, and what counts as a depletion varies between distributors. The same account can also be written differently from file to file.
• Next month’s file. You add it, rerun your prompts and keep track of which answer came from which file.
• Team definitions. Everyone who asks questions needs the same rules: what counts as a depletion, which accounts are a focus and how you rank opportunities.

ChatGPT projects help with the first and third: a project keeps files, including a mapping table, and instructions in one place, and can be shared with your team.

What a dedicated tool adds

BevBeacon’s analysis uses the same kind of large language model as general AI assistants. What it adds is the setup around the model.

• Unit instructions built in. The analysis is instructed to handle case equivalents, 9-liter cases, keg sizes and pack notation, and to list any unit corrections it makes in the answer’s details.
• Reports kept as versions. Every report is kept exactly as uploaded. Add next month’s export and BevBeacon keeps both versions. Each conversation remembers which version of each report it used, so earlier answers stay as they were.
• Team rules and opportunity score. Your team’s rules and opportunity score, a weighted ranking of accounts, are applied to each question by default. Signals your data can’t support are skipped and noted. Every version is kept, and only the people you choose can change them.
• Remembered account matches. When the same account appears differently across files, BevBeacon links it only on strong evidence, asks you when it’s unclear and remembers your answer. When both files carry a unique account ID, your answer can also carry over to next month’s version.

ChatGPT covers more ground in other ways; see BevBeacon vs. ChatGPT for a side-by-side.

For a side-by-side view, see BevBeacon vs. ChatGPT. For what BevBeacon does with depletion files, see AI depletion analytics.

Questions

Can ChatGPT convert kegs to case equivalents?

Yes. To keep results consistent, give it the standard you use: a US barrel is 31 gallons and a beer case equivalent is 2.25 gallons. A 1/2 BBL keg (15.5 gallons) is then 6.89 CE, and a 1/6 BBL sixtel (5.17 gallons) is 2.30 CE. Metric kegs differ: a 50 L keg is 13.21 gallons, or 5.87 CE. Check its figures against the keg table or the calculator.

Can ChatGPT combine reports from several distributors?

Yes, with a mapping. Give it a table that maps each distributor’s item codes to your products and a list of account names that refer to the same account, and ask it to list every row it could not match. Then check the joins by hand: confirm the combined total equals the sum of each file’s total, and follow two or three accounts that appear in more than one file.

Sources

1. OpenAI Help Center, Data analysis with ChatGPT (ChatGPT supports .xls, .xlsx and .csv files and, for some data-analysis tasks, writes and runs Python code; review the generated code, outputs and assumptions; read September 28, 2026).
2. OpenAI Help Center, Projects in ChatGPT (projects keep files and instructions, can be shared, and can be used to “return each week to generate updates without starting over”; read September 28, 2026).
3. Wine Industry Advisor, From Cellar to Table: How Wine Professionals Are Putting AI to Work (Expert Editorial by Paul C. Ringgold, September 24, 2026; Greg Shepard and Daniel Webber).
4. Dimensional Insight, Challenges with Distributor Data Reporting (what counts as a depletion varies between distributors; distributors assign their own product codes).
5. 27 CFR 25.11, Meaning of terms (barrel = 31 US gallons).
6. 27 CFR 25.156, Determination of tax on keg beer (1/2 and 1/6 barrel kegs; 50 L = 0.42608 barrel).
7. NBWA, Industry Fast Facts (one barrel = 31 gallons or 13.78 24/12 case equivalents).
8. American Craft Spirits Association, Proof Gallons to Cases (a case is 12 750 mL bottles, 9 liters).
9. Wine Spectator, When Wine Spectator lists a wine’s production volume, are these 6-bottle cases or 12-bottle cases? (9-liter units, the equivalent of 12 × 750 mL).
10. OpenAI Help Center, File Uploads FAQ (supported file types; read September 28, 2026).
11. OpenAI Help Center, Connected apps in ChatGPT (Google Drive, SharePoint, OneDrive and other apps; read September 28, 2026).