Short answer

QGIS and ArcGIS Online are not the same kind of thing, and comparing them directly is the first mistake. QGIS is a free, open-source desktop application for spatial analysis, editing, terrain work, and cartography. ArcGIS Online (AGOL) is Esri's hosted SaaS platform for storing, sharing, and publishing maps — hosted feature layers, web map viewers, dashboards, StoryMaps, and field apps delivered to a managed set of named users. The desktop counterpart to QGIS in the Esri world is ArcGIS Pro, not AGOL. So the real question is usually: do you want an open desktop workbench with self-managed publishing, or a managed cloud platform that handles hosting, accounts, and ready-made apps for you?

What each product actually does

QGIS gives you the full analytical surface on your own machine: the Processing toolbox (hundreds of native, GDAL, GRASS, and SAGA algorithms), live access to PostGIS, raster terrain analysis, the print layout composer, and an enormous plugin ecosystem. It costs nothing to run on any number of seats, and its file and database formats are open.

ArcGIS Online sits at the other end of the lifecycle. It is where you publish and share. Its strengths are operational: organisational subscriptions with role-based permissions, hosted feature and tile layers backed by Esri's cloud, the Map Viewer for browser editing, and turnkey apps — Dashboards, Experience Builder, StoryMaps, Field Maps, Survey123. You do not run servers; Esri does. The trade-offs are a per-user subscription with credit-based consumption (for storage, tile generation, geocoding, and analysis) and tight coupling to the Esri ecosystem.

If your comparison is really "desktop vs desktop," compare QGIS vs ArcGIS Pro. If it is "how do I get maps to other people," compare QGIS Server / vector tiles vs ArcGIS Online. Most teams need both halves.

Can QGIS do what ArcGIS Online does?

Largely yes, but you assemble it rather than buy it:

  • OGC services: QGIS Server publishes WMS, WFS, WMTS, and OGC API – Features straight from a QGIS project file. You host it (Apache/Nginx or a container).
  • Vector tiles + web client: export tiles from PostGIS/QGIS and serve them to a MapLibre GL front end for a fast, fully custom web map.
  • Dashboards/apps: there is no single drop-in equivalent of Esri Dashboards; you build with web frameworks or use third-party tools. This is where AGOL's managed apps save the most time.

So QGIS can publish to the web, but you own the hosting, security, and uptime. ArcGIS Online trades control and cost for a managed, low-effort path to shared maps and apps.

A concrete decision walkthrough

Say a geology team must: (1) run slope and contact buffering analysis, (2) maintain an authoritative editable dataset, and (3) give a regional office an interactive dashboard of active sites.

  • Analysis (1): QGIS handles it for free — Processing > Raster terrain analysis > Slope, then vector buffering, with results reproducible via the model designer or PyQGIS. ArcGIS Pro does the same but is licensed per seat.
  • Authoritative store (2): both can sit on PostGIS; AGOL can instead use a hosted feature layer with built-in versioning-lite and sync for field edits.
  • Dashboard (3): ArcGIS Online wins on speed here — a hosted feature layer plus a Dashboard is an afternoon. The QGIS path means QGIS Server or vector tiles plus a custom or third-party dashboard, which is more flexible but more build effort.

A pragmatic outcome: QGIS for analysis and cartography, PostGIS as the store, and either AGOL for the managed dashboard, or a self-hosted MapLibre app if licence cost, data sovereignty, or full control dominate.

Cost, lock-in, and reproducibility

QGIS has no licence cost and uses open formats, so a project can be reopened years later without a subscription. ArcGIS Online is a recurring subscription with a credit model: certain operations (tile generation, hosted analysis, geocoding, storage above quota) consume credits, so total cost depends on usage, not just seat count — budget for it. Lock-in is a real factor: AGOL apps, hosted layers, and StoryMaps live inside Esri's platform, while a QGIS + PostGIS + GDAL stack is portable and scriptable end to end. Neither is "better" in the abstract; the right call depends on whether managed convenience or open control matters more for the work.

Common pitfalls and why they happen

  • Comparing QGIS to AGOL as if both were desktop tools. Cause: conflating Esri's products. Map QGIS to ArcGIS Pro and AGOL to a self-hosted publishing stack.
  • Underestimating AGOL credits. Cause: treating SaaS as flat-rate. Tile generation and hosted analysis draw credits; model the consumption before committing.
  • Assuming QGIS "can't do web." Cause: not knowing QGIS Server / vector tiles. It can; you just host it yourself.
  • Choosing the platform before the data model. Cause: picking a brand first. Decide the store (e.g. PostGIS), the CRS discipline, and the publish target, then pick tools.
  • Letting one analyst's desktop be the system of record. Cause: no shared store. Put authoritative data in a database both tools can reach.

QA and validation

Whichever stack you choose, validate the same things: confirm layer CRS and units before any measurement; check that a published web layer renders identical geometry and extent to the desktop source (a wrong reprojection on publish is common); verify permissions actually restrict who can edit; and confirm the workflow can be rebuilt — a saved QGIS model or PyQGIS script, or a documented AGOL item with its source service — so it does not depend on one person's memory.

Bathyl perspective

Bathyl is genuinely tool-agnostic: we judge a stack by whether it produces results others can inspect, reproduce, and extend. In practice that often means open desktop analysis (QGIS) over a PostGIS store, with the publishing layer chosen per project — ArcGIS Online when an organisation already lives in Esri and wants managed apps fast, a self-hosted MapLibre/QGIS Server path when control, cost, or data sovereignty lead.

Related reading

Sources