# Limitations

Source: https://getanalog.io/docs/limitations/

Page types where structured output is currently least reliable.

Hi, Marcus here. I'm Analog's founder, and I work on structured
extraction every day. These are the gaps I know about:

- **Complex HTML tables.** We've put a lot of effort into high-fidelity
  table capture, but complex tables can still extract poorly. The
  [chest-to-belt throughput table on the Factorio
  Wiki](https://wiki.factorio.com/index.php?title=Inserters&oldid=219689#Chest_to_belt)
  is a good example.
- **Long social feeds and walls.** Only part of a very long feed may be
  captured; the result says how far it got.
- **Deep discussion threads.** Tables inside replies can sometimes
  become garbled.
- **Restaurant menus.** Menus that change structure from section to
  section can produce garbled or oddly split output.
- **Multi-page sweeps** (`--pages`). This is still experimental. Pages
  pool only where their sections share a template; fields only some
  pages carry are null elsewhere and named in the sweep note, and
  sections that don't line up stay separate.

## Expected boundaries

JavaScript-heavy pages are not a limitation by themselves: the
built-in browser renders them by default. A site that blocks automated
visitors is different from a weak extraction; Analog reports the block
as the result and stops. See [Fetching](https://getanalog.io/docs/fetching/).

## Report a result

A quality report is useful whether or not the page type appears above:
the [Feedback guide](https://getanalog.io/docs/feedback/) explains the symptom labels and
submission routes.