Summary
Should our download packages (chart / data page / mdim) offer an Excel .xlsx file alongside the CSV+metadata+README zip?
Proposed in the mdim download-package review (#6668) as an xlsx with 4 tabs — Data / Metadata / Sources / Readme — and deferred in #6769 with "the Excel variant is on hold until we know whether anyone would use it".
Current recommendation: don't build it yet. Filing this so the evidence is written down and it can be triaged rather than re-litigated.
What we know about demand (checked Sept 2026)
Short version: there is no demand signal, in either direction.
- No external user has ever asked. An org-wide sweep of issues and comments for
excel / xlsx / spreadsheet finds zero external requests. Every mention is internal:
- Google search demand is negligible and mostly aimed elsewhere. Over 12 months, queries reaching ourworldindata.org mentioning Excel/XLSX/spreadsheet total 39,315 impressions and 2 clicks. The top ones are
maddison project database 2023 download excel, sipri military expenditure database excel download, "ghe2021_deaths_bycountry_2020.xlsx" — people hunting the producer's spreadsheet, not an OWID export. An Excel-native audience exists; nothing shows it wants Excel from us.
- Nobody is guessing at an xlsx endpoint.
/grapher/<slug>.xlsx doesn't exist and gets ~470 requests/month, all 404, overwhelmingly vulnerability scanners.
Important caveat: every instrument above measures behaviour on formats we already serve. You cannot measure revealed preference for an option that was never on screen, so the absence of a signal is not evidence that demand is absent. This question is not answerable from analytics — only by shipping something and measuring.
Why not now: the cost is the format, not the file
Generating an xlsx is nearly free — Table.to_excel(with_metadata=True) already exists in owid-catalog (lib/catalog/owid/catalog/core/tables.py) and writes a data sheet plus a metadata/codebook sheet. And the download modal already ships an Excel/Google-Sheets affordance today (=IMPORTDATA(csvUrl)).
The expensive part is getting the presentation right. The 4-tab shape means porting the README/citation/sources formatting into a spreadsheet layout and then keeping it in sync. etl/collection/download_package_format.py is already a hand-maintained Python port of grapher's own formatting logic; adding a second presentation surface doubles that sync burden for a format nobody has asked for.
Context on who actually reads downloads
Worth weighing: ~98% of requests to chart data-file endpoints are machines. Even on life-expectancy — the most-downloaded chart on the site — a realistic human data-download count is a few hundred per month (measured as in-browser clicks). Excel is a human-only format, so it competes for a small, though real, slice of the audience.
If we decide to do it
Ship the simplest version first: data sheet + metadata sheet from the existing writer, as a second button next to the zip. Then read the split after a month. Don't build the 4-tab format up front — that's the part with the maintenance cost, and we'd be building it on intuition.
Summary
Should our download packages (chart / data page / mdim) offer an Excel
.xlsxfile alongside the CSV+metadata+README zip?Proposed in the mdim download-package review (#6668) as an xlsx with 4 tabs — Data / Metadata / Sources / Readme — and deferred in #6769 with "the Excel variant is on hold until we know whether anyone would use it".
Current recommendation: don't build it yet. Filing this so the evidence is written down and it can be triaged rather than re-litigated.
What we know about demand (checked Sept 2026)
Short version: there is no demand signal, in either direction.
excel/xlsx/spreadsheetfinds zero external requests. Every mention is internal:.xlsx, but filed under "Can have". Box still unchecked, never built.owid_energystep and theenergy_datarepos #6668 / Rework the download-package readme, and stamp zip entries with the build date #6769 — the current proposal and its deferral.maddison project database 2023 download excel,sipri military expenditure database excel download,"ghe2021_deaths_bycountry_2020.xlsx"— people hunting the producer's spreadsheet, not an OWID export. An Excel-native audience exists; nothing shows it wants Excel from us./grapher/<slug>.xlsxdoesn't exist and gets ~470 requests/month, all 404, overwhelmingly vulnerability scanners.Important caveat: every instrument above measures behaviour on formats we already serve. You cannot measure revealed preference for an option that was never on screen, so the absence of a signal is not evidence that demand is absent. This question is not answerable from analytics — only by shipping something and measuring.
Why not now: the cost is the format, not the file
Generating an xlsx is nearly free —
Table.to_excel(with_metadata=True)already exists inowid-catalog(lib/catalog/owid/catalog/core/tables.py) and writes a data sheet plus a metadata/codebook sheet. And the download modal already ships an Excel/Google-Sheets affordance today (=IMPORTDATA(csvUrl)).The expensive part is getting the presentation right. The 4-tab shape means porting the README/citation/sources formatting into a spreadsheet layout and then keeping it in sync.
etl/collection/download_package_format.pyis already a hand-maintained Python port of grapher's own formatting logic; adding a second presentation surface doubles that sync burden for a format nobody has asked for.Context on who actually reads downloads
Worth weighing: ~98% of requests to chart data-file endpoints are machines. Even on
life-expectancy— the most-downloaded chart on the site — a realistic human data-download count is a few hundred per month (measured as in-browser clicks). Excel is a human-only format, so it competes for a small, though real, slice of the audience.If we decide to do it
Ship the simplest version first: data sheet + metadata sheet from the existing writer, as a second button next to the zip. Then read the split after a month. Don't build the 4-tab format up front — that's the part with the maintenance cost, and we'd be building it on intuition.