# Methodology and maintenance Where the figures in these writing guides come from, what they can and can't tell you and how the guides are kept up to date. > **About this guidance:** this page explains how the guides were compiled. The guides summarize guidance maintained elsewhere, for reference; the [sources](README.md#sources) take precedence. Found a problem? Report it in the [Mangrove issue queue](https://github.com/PreventionWeb/undrr-mangrove/issues). - **Last reviewed:** 29 September 2026 ## The December 2025 baseline The "baseline" figures quoted in the [news](news.md), [events](events.md) and [publications](publications.md) guides come from a one-off analysis of existing content, run in December 2025 as part of drafting the first versions of these guides (undrr/drupal-base!88, a working branch that is not merged). ### How the examples were selected A Drupal command extracted published items that met all of these conditions: - Created on or after 1 January 2024 (for events, an event date on or after that day) - Rated highly by editors: an Interestingness value of 75 or more, or a Longevity value of 80 or more - Assigned to both undrr.org and PreventionWeb - Tagged with at least one theme, hazard or country This produced 122 items: 50 publications, 41 news items and 31 events. Blogs were not included. ### What the selection measures The criteria select content that editors chose to promote widely and keep visible for a long time. That is a measure of **prominence, not writing quality**. A prominent item can be badly structured, and a well-written one can sit unpromoted. So the figures describe what prominent content looked like at the end of 2025. They are useful for a sense of typical length and structure, and the guides use them that way. They are not targets, and none of the guidance should be read as "do this because 69% of items did". Where the baseline conflicts with guidance from the UNDRR Communications team, the Communications guidance wins. The main case is publication summaries: many baseline summaries were written as promotional introductions with bulleted lists, while the Communications guidance asks for an objective two-paragraph outline (see [publications](publications.md)). ### Baseline figures | | Publications | News | Events | |---|---|---|---| | Items analysed | 50 | 41 | 31 | | Body length, average (range) | 169 words (43–316) | 782 words (302–1,326) | 427 words (100–1,496) | | Body length, median | 163 words | 780 words | 356 words | | Paragraphs, average | 2 | 14.1 | 8.8 | | Items using headings | 0% | 69% | 97% | | Headings per item, average | 0 | 2.9 | 4.6 | | Items using lists | 44% | 36% | 65% | | Internal links per item, average | 0.2 | 4.2 | 2.5 | | Title length, average (range) | 82 characters (29–186) | 82 characters (45–148) | 88 characters (45–135) | | Titles with a colon | 56% | 31% | 65% | | Titles with a question | 0% | 2% | 0% | Other observations: - Every baseline publication had a PDF attachment, and none had a short description filled in. - 98% of items had themes (about two per item) and 98% had an image; 16% had a hazard tag. - News types: Update 59%, Op Ed 22%, Press release 10%, others 2–5% each. Event types: Meetings and conferences 61%, Training event 39%. Publication types: Documents and publications 98%, Statements 2%. ### Known limitations - **Small, single snapshot.** 122 items from one date. Differences of a few points between content types mean little. - **Rough text measures.** Paragraph and heading counts came from the stored HTML, so empty paragraphs and headings containing only a non-breaking space were counted. Some metrics were miscounted outright (a "related links" measure exceeded 100%); those are not used in the guides. - **Titles and headings were not checked for style.** Several baseline titles and headings use title case or trailing colons; the guides do not follow them. - **Hand-maintained term lists drifted.** The December 2025 drafts listed taxonomy term IDs by hand. By September 2026 several were wrong (for example, the ID given for France belonged to another country) and several theme and hazard names had changed. The guides now name terms without IDs, and term IDs are in the [taxonomy reference](taxonomy.md), which is generated from a copy of the production database with the queries it lists rather than edited by hand. ## Examples The full December 2025 extracts are not published. They were selected for prominence, so they would present unreviewed content as a model. The examples in each guide come from a separate pass in September 2026, using a copy of the production database taken on 2 September 2026: - **Candidates:** 138 items. News, publications and events published since 1 January 2025 were chosen with the same prominence criteria as the baseline. Blogs have no Interestingness or Longevity values, so the candidates were the 50 newest English blogs on PreventionWeb with a theme, hazard or country tag. - **Selection:** each candidate was read in full and judged against the Communications team's guidance (abstracts and blog editing), the Mangrove editorial manual and the structure guidance in these guides. Most candidates did not qualify; 19 were picked: five publications, five news items, four events and five blogs. - **Annotation:** each pick says what it does well and what an editor would still fix, so the guides can show a strength without endorsing the slips. The recurring problems seen in the candidates that were not picked are folded into the "Avoid" guidance in each guide, without naming items. The extraction is used only to find candidates; the guides and baseline figures are not regenerated from it. To propose a better example, or to replace one that has changed, edit the guide through a merge request. ## Maintaining the guides ### One source These Markdown files in the [shared-web-assets repository](https://gitlab.com/undrr/common/shared-web-assets/-/tree/main/docs/editorial-guides) are the single source for the content-type writing guides. Change them through a merge request. Any other copy is generated from them and says so at the top. Don't edit a copy; edit the source and generate it again. ### What goes where | Guidance | Home | |---|---| | Mechanics: spelling, capitalization, punctuation, numbers, terminology, inclusive language | [Mangrove editorial manual](https://mangrove.undrr.org/llms-editorial-manual.txt) ([rendered version](https://mangrove.undrr.org/?path=/docs/contributing-editorial-manual--docs)) | | Writing for each content type, and metadata | These guides | | Term names and IDs | The [taxonomy reference](taxonomy.md), generated from the production taxonomy | | Page building with Gutenberg | The Mangrove [page building guide](https://mangrove.undrr.org/llms-page-building.txt) (rendered versions: [Building landing pages](https://mangrove.undrr.org/?path=/docs/patterns-building-landing-pages--docs) and [Drupal Gutenberg integration](https://mangrove.undrr.org/?path=/docs/getting-started-integration-drupal-gutenberg--docs)) | | Internal UNDRR style, editorial workflow | SharePoint (UNDRR staff only); see the [index](README.md#for-undrr-staff) | Link to the manual instead of restating its rules. If a guide needs a mechanics rule the manual doesn't have, propose it for the manual. ### When to review - When the Mangrove editorial manual changes, check that the guides still link to the right sections and don't contradict it. - When the Communications team updates its guidance. - When a content type, field or vocabulary changes on the platform: new news or publication types, renamed themes or hazards. - When editors report that a guide is unclear or wrong. After a review, update the "Last reviewed" date in each guide you checked and add a row to the changelog in the [index](README.md#changelog). The baseline figures are not refreshed on a schedule. If a new analysis is ever run, label its figures with their date alongside the December 2025 ones rather than silently replacing them.