Note count: 2000
Images taken from picsum.photos
Mardkown content taken from Lorem Markdownum
Notes are initially generated and stored in the Markdown format.
Here is an example of a generated markdown note hosted on Amplenote.
For each note:
Prefix each note with the corresponding inline markdown images
Then append the corresponding number of generated markdown blocks
Joplin, Amplenote, Obsidian, Bear:
Notes imported directly from the generated markdown files
Evernote:
Generated markdown files are converted to the ENEX
format before importing into Evernote
Notejoy, Nimbus Notes, Apple Notes, Onenote:
Notes imported from previously generated ENEX
file
Notion:
Generated markdown files are converted to .docx
before importing into Notion
Standard Notes:
For every initially generated markdown note, all links to local images are converted into URLs to web images.
The modified markdown files are imported into SN.
Roam Research:
Just like with Standard Notes, for every initially generated markdown note, all links to local images are converted into URLs to web images.
The modified markdown files are converted to Roam's json format and then imported to Roam.
Each of the 5 tests described in the section above were performed for the following types of clients (when available):
Tests performed on mobile browsers (for apps not providing native mobile clients)
For each combination of test, platform and app, the average of 3 separate test runs was computed and added to the NoteApps database.
We chose the benchmarking thresholds to try to approximate 50/50 split between apps that did and didn't qualify as "faster than the threshold." You should find that about half of the note apps we feature will qualify as "under the threshold" for each feature that we benchmark. In the future, we will periodically re-adjust our thresholds if we find that the existing thresholds yield more than 80% inclusion or less than 20%.
Some apps do not support importing notes with bundled images, which is why we resorted to hosting images in a separate location and linking to them from affected apps. This workaround is expected to slightly favor these apps, with benchmarking times being potentially smaller due to images not being synced to the app's storage servers.
The set of notes generated to conduct these measurements had the following attributes:
2,000 markdown notes
Images interspersed
Variable length (sample of small note, sample of large note)
2,000 notes seems like a reasonable proxy for "a power note user" or "a normal user who accumulates notes over many years." It is difficult to model this generative use case in database-like/outliner apps, since there is no obvious equivalence between structured (markdown) documents and hierarchical blocks.
Therefore, for apps such as Roam Research and Workflowy, the benchmark results are not an apples-to-apples comparison relative to traditional note-taking apps. If any reader thinks they know of a better way to model a 2,000 note data set efficiently in a block-based architecture, drop us a line at hello@noteapps.info.