[wp-meta] [Making WordPress.org] #8166: Plugin Search Algorithm Creates Barriers for New Plugin Discoverability

[Making WordPress.org]

#8166: Plugin Search Algorithm Creates Barriers for New Plugin Discoverability
------------------------------+------------------------------
 Reporter:  webstepper        |       Owner:  (none)
     Type:  enhancement       |      Status:  new
 Priority:  normal            |   Milestone:  Improved Search
Component:  Plugin Directory  |  Resolution:
 Keywords:                    |
------------------------------+------------------------------

Comment (by webstepper):

 == Search vs Discovery ==

 Thanks for the thoughtful reply. I agree with the core distinction between
 search and discovery, and I don’t think search results should be used as a
 marketing channel for untested plugins.

 Where we still seem to disagree is not on whether popularity and trust
 signals should matter — they absolutely should — but on how strongly the
 current algorithm converts early advantage into long-term dominance, and
 what that means for users and the wider plugin ecosystem.

 == Search Can Be Correct in Intent but Still Flawed in Outcome ==

 Search should help users find the most relevant and reliable result for a
 known need. However, the current scoring model doesn’t just prefer proven
 plugins — it effectively collapses the result set around incumbents.

 Two mechanics contribute most to this behavior.

 === Exponential Active Install Weighting ===

 Using one million installs as the baseline creates a steep decay curve. A
 plugin with thousands of installs is not merely ranked lower; it is
 functionally invisible compared to plugins that crossed the early
 popularity threshold.

 This goes beyond ranking by trust and creates a strong reinforcement loop.
 Higher-ranked plugins receive more installs, those installs further
 improve ranking, and over time rank becomes a proxy for “got there first”
 rather than “best available option.” In this context, time is not just a
 filter — it is a multiplier.

 === Penalizing Missing Signals Instead of Negative Signals ===

 New plugins do not yet have support threads, resolutions, or ratings.
 Treating missing data as a negative value penalizes plugins before users
 have had any opportunity to evaluate them.

 There is an important distinction between a plugin with unresolved support
 issues and a plugin that is simply new. The current model does not clearly
 separate the two.

 == This Is Not About Helping Authors, but About Avoiding Monoculture ==

 One visible outcome of the current incentives is plugin name convergence.
 Authors quickly learn that unique or branded names are disadvantaged
 compared to exact-match, keyword-heavy titles.

 From a user perspective, this results in search results filled with nearly
 identical plugin names. This makes alternatives harder to distinguish,
 recommendations more ambiguous, and experimentation with different
 approaches less likely. While this behavior is a rational response to
 incentives, it is not a healthy search outcome.

 == This Is Not a Call to Weaken Search ==

 To be clear, this is not an argument for randomly promoting untested
 plugins, replacing relevance with novelty, or ignoring installs, ratings,
 or support history.

 The goal is not to make search less reliable, but to avoid permanent
 invisibility for new, high-quality plugins that users never get a chance
 to evaluate.

 == Containing Search and Supporting Discovery ==

 Search and discovery can coexist without undermining each other. Discovery
 does not need to be injected directly into core search results to be
 effective.

 Dedicated discovery surfaces such as “New & Notable” or “Rising Plugins,”
 clearly labeled emerging placements, optional user-controlled sorting like
 “Newest” or “Trending,” and short grace periods where missing support or
 rating data is treated neutrally would all help meaningful alternatives
 surface over time. Reducing title-match dominance would further ensure
 that relevance is not tied primarily to name cloning.

 These approaches preserve user trust while allowing room for innovation.

 == Why This Matters for WordPress Users ==

 WordPress has historically thrived because new ideas could gain traction
 without first copying what already existed.

 If the only viable path to visibility is cloning names and feature sets,
 the ecosystem slowly converges. Over time, this leaves users with fewer
 real choices, less innovation, and older or abandoned plugins continuing
 to rank highly due to past momentum.

 Search can continue to answer “show me the most proven option,” while the
 platform also answers “what else exists.” Supporting both does not weaken
 WordPress — it strengthens it.

 I’m happy to continue the discussion and explore ways to improve discovery
 without compromising the reliability users expect from search.

-- 
Ticket URL: <https://meta.trac.wordpress.org/ticket/8166#comment:3>
Making WordPress.org <https://meta.trac.wordpress.org/>
Making WordPress.org