For Sustainability Consultants ·
What you'll accomplish
The utility and fuel export for your sites runs to hundreds of rows, collected from several different data owners. Somewhere in it a month is missing, a meter reads in MWh where the rest read in kWh, and one bill appears twice. Scrolling for them takes hours. You upload an anonymized copy to a regular ChatGPT chat and ask for a table of suspected gaps, outliers, unit mismatches and duplicates, each with its row reference and the reason.
ChatGPT flags. You decide what each flag means and fix the data in your own spreadsheet. It never recalculates emissions and never picks an emission factor.
This guide goes deeper than the Excel-based check in Level 2. It suits a large export, and you ask for an explanation of each flag.
What you'll need
What you should see: A clean table with one header, no merged cells and a Row column. Troubleshooting: If your workbook has several tabs, export each tab as its own file, or upload the one you want checked.
The attach control looks different between web and app versions, so look for a plus icon or paperclip near the message box.
What you should see: The file name shown with your message. Troubleshooting: If the upload is refused, check file size and your plan's upload limits. Split the file by site or by year.
Large exports go better in separate passes. Start with the structure.
I have attached an anonymized activity data export for [Company A] with one row per utility or fuel bill. Describe the columns you see, the number of rows, the date range, and the units that appear in the Unit column. Do not analyze yet.
Check this description against what you know. If the row count or the units differ from your expectation, stop and fix the upload before going further.
Check the file and give me one table with these columns: Row, Site, Check type (gap, outlier, unit mismatch, possible duplicate), What you found, Why you flagged it.
Checks:
1. Gaps: for each site and each fuel or utility, list months with no row between the first and last period shown.
2. Unit mismatches: rows where the Unit differs from the unit used for the same site and fuel in other months, or where the quantity looks like it is in a different unit.
3. Possible duplicates: rows with the same Site, fuel, period and quantity, or the same Source document ID.
4. Outliers: quantities far above or below the same site's other months. Say how you judged "far".
Rules: Flag only. Do not correct, delete or fill any value. Do not calculate emissions or apply any emission factor. If you are unsure, flag it and say why.
For the five flags you consider most likely to affect the totals, explain what could cause each one in utility or fuel data, and what document I should check to settle it.
Add a "Flag checked" column in your master file. For each flag, open the source bill or meter record. Decide what is wrong, or whether anything is. Fix the value in your own sheet and note the decision.
What you should see: A table with row references and reasons. Troubleshooting: If row numbers in ChatGPT's table do not match yours, use your Row column. If it skipped the whole file, ask it to confirm the number of rows it analyzed.
Check one site only:
Look only at Site [code]. List every month with a missing row, and any month where the quantity differs from the previous month by a large amount.
Cross-check units:
List every distinct Unit value in the file with the number of rows for each. Flag any that look like a typing variant of another unit.
Compare two exports:
I attached last year's and this year's exports. List rows or sites present in one and not the other.
Draft the data-owner query:
For the flags below, draft one short email to the data owner asking them to confirm the figure and send the source bill. Use site codes only.
[paste flags]