"Your top fan list" is theater - what a measured tier actually shows

"Your top fans" is one of the most seductive interfaces in social software:
a short list of names that flatters you, ranks people you know against each
other, and invites the next move - check profiles, send messages, post again.
It is also, at every layer, an artifact: a particular window, a particular
definition of "engagement", a particular purpose (yours is to keep you
engaged), and no way for the person being ranked to inspect any of it.

Theatre is not the same as lying. The names on such a list may well be people
who interacted with you. The performance is the authority the list projects -
a verdict that looks measured while being unauditable.

What any fan list is made of

Strip the label and every version of this list has four hidden choices:

1. The window. Last 30 days? Last engagement ever? Since you opened the

   business dashboard? Change the window and the list changes; the interface
   rarely tells you which one produced today's ranking.

2. The currency. Reactions weighted more than views? Comments more than

   taps? Story replies? Each platform (and each third-party tool) picks its
   own exchange rate, and none of them publish it.

3. The denominator. Top of whom? Followers-only? Everyone who saw you?

   Lists measured against tiny pools look dramatic.

4. The purpose. A list that made you feel finished would be a failure by

   its designer's metrics. These lists are built to prompt the next action,
   which means they are built to highlight what's missing as much as what's
   there.

Psychology supplies the hook: people are drawn to social ranking because
rankings compress identity into one legible number (or short list), and
legibility feels like knowledge.¹ The discomfort when your name
isn't on someone's list - or when a low-rank reading appears - shows the same
mechanism running in reverse. Any system that hands out flattery by ranking
is also handing out anxiety by ranking.

The measured alternative: a tier you can audit

What your own archive supports is structurally different. Not nicer feelings -
inspectable arithmetic:

• Inputs are visible rows. The engagement tier per person is computed from

  message exchange (sender tags, timestamps), graph sets (who follows whom,
  close friends, favorites) and activity presence - every input a file you
  can open in a text editor.²

• **Thresholds are published, in the code and in the post that documents

  them.* core >= 60, active >= 35, casual >= 15, otherwise dormant;
  at-risk when the score is at least 35 and their recent reply rate fell
  below 30% of their prior rate and* it has been over 21 days since their
  last reply.² No secret currency, no unannounced window: 60 and 35 are
  numbers you can argue with.

• The window is stated, not implied. Recent-versus-prior rate comparisons

  use declared windows (the last 60 days against the 120 before them); the
  strength side uses declared staleness cutoffs of 180 and 365 days.³

• Ranking is your choice, not the interface's. The accounts directory

  sorts by any column - searches, messages, last exchange, tier - and filters
  by relationship state. The default order is a reading order, not a
  leaderboard, and nothing rolls a person's rows into a public-facing badge.

• It never leaves the device. The tier is computed inside your browser

  from your ZIP; there is no server that knows who your core people are,
  and no counter that exists to tempt you into checking.⁴

Why flattery-by-list fails even on its own terms

• It mistakes volume for regard. The loudest interactors are not

  necessarily the most attached - spam-like engagement patterns and genuine
  intimacy look identical in a window short enough.²

• It collapses direction. Someone who watches every story and never

  replies is a fan by one currency and invisible by another; the
  parasocial post covers why direction has to be shown as two columns, never
  summed into one.⁶

• It is unfalsifiable from outside. You can never check the window, the

  weights or the denominator - which is precisely why the list's authority
  can't be audited, and why the anxious part of the mind treats each reshuffle
  as meaningful news.

The archive's version fails differently and therefore better: if you disagree
with your tier assignment, you can open the rows and see exactly which inputs
produced it. Disagreement with a fan list can only be felt; disagreement with
a threshold can be demonstrated.

Related reading

The counts post for how every metric in the product is derived and disclosed;
the follow-doesn't-mean-care post for the graph-versus-evidence layering that
underlies any tier; the quantified-self post for why no score of you ever
appears next to these labels.

¹: Scott, J. - Seeing Like a State (1998) on legibility as a
prerequisite for institutional control of complex social facts; cited here as
framing for why ranking compresses identity into readable outputs.
²: packages/shared/src/analytics/engagement-intel.ts:149-176 -
recent (60d) vs prior (120d) reply-rate windows, balance from sender-tagged
counts, tier thresholds core >= 60, active >= 35, casual >= 15, else
dormant, and the three-part at-risk rule.
³: packages/shared/src/analytics/relationship-insights.ts:10-12 -
STALE_DAYS = 180, VERY_STALE_DAYS = 365.
⁵: The decision ledger D5 ban on composite scores, documented in
content/blogs/2026-10-07-the-parts-we-refused-to-build.md section 5.
⁴: SECURITY.md section 1 - analysis engine makes zero network calls;
packages/shared/src contains no fetch, XMLHttpRequest or sockets.
⁶: content/blogs/2026-10-08-parasocial-imbalance-who-watches-back.md -
directional attention shown as two columns, never summed.
⁷: packages/shared/src/analytics/relationship-insights.ts:39 -
if (!following && followed) return 'silent-fan'; population collected at
relationship-insights.ts:312.