Jump to content

Analytics/Limn/Roadmap

From mediawiki.org

Limn Roadmap

One of the questions I want to seek a collaborative answer to in the coming weeks is how does Limn fit into the overall product offering of the Analytics team and what feature set should it support and what feature set should it not support.

To kickstart this discussion I want to propose some boundaries to Limn's features just to gauge response and get feedback on what people think Limn should be. So here we go (remember this is for discussion purposes ATM).

Limn Vision

Limn supports people in data-based storytelling by supporting people to create charts on the web and annotate them with notes, embed the charts in wikis and link to the (meta) data that created the chart. Limn is a GUI-based visualization tool which means that data analysts and product folks should be able to create charts from datasets that are published on the web without any programming support.

Features that Limn should support

If we take that as Limn's vision, what are the consequences for the feature set?

Broader community support

Limn is now rather limited in scope and mostly used by (usable by?) WMF staff. For many key metrics (together grouped into Report Card) it can only show top 10 or so Wikipedias, and only last 3 years. To fulfill its original promise and evolve into a tool widely usable also by community it should be possible to

  • Access Limn in other languages than English (starting with generated charts/legend/annotations, later maybe also UI)
  • Show data for any language from any project, and select these languages out of 280+ in a friendly manner (using a language selector box?). E.g. show data for all Wiktionaries in Indian languages (+/- 22)
  • Select any time range.

MediaWiki support

Currently it is not possible to embed a Limn chart in another environment; most of our data analysis is currently presented on a wiki and hence if we are serious about supporting people in data-based story telling then we should introduce MediaWiki support in Limn.

Edit UI

Currently, there is no UI to edit the dashboard.json, datasource.yaml, and graph.json files which breaks the GUI promise of Limn.

Annotations

Just a line on a chart is often not enough to tell the entire story. We need annotations to explain gaps, spikes and other anomalies.

Transformation of Data

Not all the data has the same range and this can easily distort a chart. Transforming the data (indexed, log) should be a feature of Limn.

Discoverable

Right now it is very hard to find a Limn chart or dashboard. We need to add support of publishing a chart / dashboard to a Data Registry so people can easily find charts. Another short term solution is to make Limn charts indexable by search engines.

Features that Limn should not support

Type of charts

Primary focus is on charts with either time on the x-axis or geomaps; we should not support pie charts, scatter plots, 3d diagrams, etc.

Analysis support

Limn should not offer any analysis support as there are many tools that are much better suited for that, think of R, Excel, Google Spreadsheet.

Pivot support

A recurring feature request over the last year has been to add pivot support in Limn. Pivot support is the ability to aggregate dimensions in a multidimensional dataset. Every spreadsheet program offers this feature and hence we should not offer this feature in Limn.

Native versioning support of charts

Limn is a GUI to create charts, the problem of version management of charts should be solved outside of Limn, possibly in MediaWiki.

Discussing a chart

Limn should not support discussing a chart natively, if we create MediaWiki support then we can piggybag on MediaWiki's discussion features.

References