When content capacity tops out: make the shutdown a formal state change, not a quiet reduction

One-liner: When content output can no longer keep pace with how fast players consume it, toughing it out upgrades the problem from "less content" to "quality collapse plus trust collapse." Announcing a formal downgrade — stating what stops, what continues, what is compensated, and whether servers stay up — is the cheapest path currently visible. Baijing Corridor and Soul Tide ran the same playbook: stop new production, keep servers up, convert limited-time content to permanent, buy forgiveness with scarce and certain compensation. One rule sits underneath all of it: stop adding, never reclaim what players already have.

When to use it

Triggers have to be observable, not felt. If any two of these appear, put "downgrade" on the table instead of scheduling more overtime:

1. The roadmap shows its first version whose entire window is filled by reruns. The whole window, not a week or two inserted between new content.
2. Finished, unshipped content covers less than one full version cycle. This is the hardest number, and it takes one afternoon to measure.
3. The same version has slipped twice or more. Marathon's Season 3 moved from 22 September to 8 December; Bungie then abandoned a strict seasonal cadence outright. Cancelling the cadence after two slips is two halves of one move.
4. Your own external language starts saying "we cannot guarantee stable output." The Baijing Corridor notice says exactly that.

When not to use it: if the problem is on the demand side rather than supply — players cannot finish what already exists, events run with nobody participating — that is a cadence-length problem, not a capacity problem. The fix is to compress durations and raise content density, not to downgrade. (For the other end of this mechanism see cards/2026-09-20-mergic-shorter-events-iap.md: halving a 30-day event to 14 days lifted IAP roughly 50%, because half of the original window was dead time.)

Step 1: measure three numbers before deciding

Do not decide on a feeling. All three are measurable today:

Decision gate: if inventory-in-days ÷ one version cycle < 2, and restoring capacity takes more than two version cycles, take the downgrade path. Do not tough it out until quality slips. The cost of toughing it out is not lost revenue — it is turning a controlled downgrade into a forced crash.

Step 2: the notice must state what stops and what continues

Both notices use one structure worth copying directly: stops first, then continuations, then compensation.

• Baijing Corridor (15 Jan 2026): from 26 February, periodic new-version iteration suspends and operations move to permanent reruns plus rotating cyclical content; servers stay open; former limited-time events become permanent reruns; core modes keep rotating on a new cycle.
• Soul Tide (12 Mar 2026): main story and new-character production stop; effort redirects to deepening existing characters (new chapters, date stories, memory stories, outfits); existing events continue as reruns; servers maintained.

The critical part is that the continuation list must be long enough. The player's first question is "what can I still play"; compensation is the second sentence, not the first. A notice listing only stoppages reads as a shutdown warning.

Step 3: compensate with scarcity and certainty, not randomness and promises

The two compensation packages work as reference tiers:

• Baijing: 80 Synchro Clues + 50 Flashback attempts + a self-select six-star character agreement
• Soul Tide: a Vow Ring + fashion tokens, with three outfits entering the exchange shop on 26 March

Three shared properties: a scarce item (self-select six-star, Vow Ring), certainty (self-select, not a random draw), and a calendar (the exact date things go live). Compensation is not there to offset the loss — at a 1.5–2.5x loss-aversion ratio, offsetting one unit of loss takes 1.5–2.5 units of gain. Its job is to prove you are still operating. So prefer small and certain over large and vague.

Step 4: convert limited-time content to permanent — the one move that offsets loss aversion

This is the most important step on the card.

A slowdown takes away "new content that was coming." That is a disappeared gain. Vaulting or sunsetting takes away "content I already have." That is an incurred loss. Loss aversion runs at 1.5–2.5x, so these two are not the same order of magnitude on the player side. Control this, and you have controlled half the cost of the downgrade.

In practice: walk every limited-time rotating event and assess whether it can become permanent (both titles did this). The test is whether permanent availability breaks economic output or the monetisation structure. If it does, change the rewards rather than the content — event body permanent, reward access time-limited, a pattern already shown to work. If it does not, make it permanent.

The counter-example: Bungie's Destiny Content Vault removed campaigns and destinations players had already purchased. Six years later it has to restore all of it, free. It saved development resources and paid with a six-year trust ledger (see cards/2026-09-23-destiny2-content-vault-reversal.md).

Step 5: put a time boundary on every promise

The most dangerous sentence in a downgrade notice is "it may return in some form." It is the most useful sentence at the time and the most expensive later: it opens an account with no maturity date. Players keep waiting, and the balance grows every year.

The rule is simple: any "it will come back" must carry a date or a condition. No date? Write the condition. Neither? Then do not write "it will come back" — write "not for now."

Common failure modes

1. Quiet reduction (the most common). Event durations cut, rewards trimmed, and not a word in the patch notes. Players read it as concealment, and concealment costs more trust than the reduction itself — this is where "shadow nerf" accusations come from. How to detect it: read your last three patch notes. Any numeric change with no explanation counts.
2. Apologising without saying what is still playable. Compensation is one-off; availability is continuous. Hand out compensation, and if players log in next day with nothing to do, it is forgotten within two weeks.
3. Promising a return with no timeline. The consequence is not disappointment — it is a permanent discount on everything you say afterwards.
4. Using a slowdown as a cost tool without naming the trade-off. Soul Tide's notice said "due to practical constraints," and players read candour. Vagueness reads as "they want to keep the revenue and string us along." Same facts, two readings, and the only difference is whether the reason was stated.
5. Repeat downgrades. Every additional cut re-prices trust. Once a downgrade notice ships, each later contraction needs to be treated as a new announcement, not as "we already downgraded."
6. Marketing reruns as new content. Reruns are cheap, but they generate no new decisions. Players do the maths fast, and a third run of the same event will post participation numbers you would rather not see.

What you can do tomorrow

1. Build an inventory × consumption table: finished unshipped content on the left, median days-to-first-completion on the right, total days at the bottom, divided by your version cycle length. That gives you "how many versions left." It can be done today.
2. Mark the first all-rerun version on the roadmap as a decision point, not a scheduling outcome.
3. Pull the last three patch notes and list every unannounced reduction. Explain them in the next notice. Cheapest trust repair available.

Appendix

Evidence and sources

• 烛龙聚焦3A古剑？《白荆回廊》发布公告：调整更新节奏 — secondary reporting (industry media), 2026-01-15. Provides: the Baijing stops/continuations, compensation specifics, the studio's own statement on capacity
• 国产二游官宣"缓更"：主线停摆 只求"活下去" — syndicated media piece (originally 游民星空), 2026-03-12. Provides: the full Soul Tide notice, stops and continuations, compensation and launch calendar
• Bungie不瞎折腾了？官宣《命运2》复活,《马拉松》冷处理 — secondary reporting (industry media), 2026-09-23. Provides: Marathon Season 3 slipping 22 Sep → 8 Dec, the cadence abandonment, the DCV vault and restoration
• Bungie says it's "not done with Destiny"… — search summary (article body JS-rendered; headline and standfirst only), 2026-09-22
• 「20年続く運営型ゲーム」の売上が15年目にしてV字回復：“勘や経験に頼った運営”を廃し成功した「踊り子クリノッペ」の歩み【CEDEC2026】 — search summary (origin returned 403; headline and standfirst only: a 20-year-old live-service title whose revenue V-recovered in year 15 by abandoning gut-feel, experience-based operations). No figure from this source is used in this card

Evidence and sources (assessment)

• Source type: the two main cases are first-party notices (quoted verbatim through media syndication) and are the most reliable material here. The Marathon and DCV data come from Chinese industry media synthesis.
• Evidence tier A: a complete action sequence (what stops, what continues, what is compensated), concrete parameters (compensation specifics, notice structure), and both positive and negative cases. It is not S because there is no follow-up retention or revenue data for either title — "players stayed" is unverified — and no control group.
• Confidence: medium-high. The action list is directly readable from two notices rather than inferred. "This works" rests on media reports of player reaction (understanding, regret), which is indirect evidence.
• Measurement caveats: Baijing's 80 clues / 50 flashbacks / self-select six-star and Soul Tide's Vow Ring / fashion tokens are single-case compensation values, not an industry benchmark. Currency value structures do not convert across titles.
• What is inference: the gate "inventory ÷ cycle < 2 and capacity restoration > 2 cycles → downgrade" is this card's inference. No source supplies that threshold; it exists to turn a judgement into an operable gate, and must be recalibrated to your own delivery cycle. "Compensation proves you are still operating rather than offsetting the loss" is also this card's inference.
• One unused lead: a CEDEC2026 talk report claims a 20-year-old Japanese title V-recovered its revenue in year 15 by abandoning gut-feel operations (Gree's Odoriko Crinoppe). Only the headline and standfirst were retrievable (origin returned 403), so no figure or method from it is used here — it is recorded solely as the next search target for "how to operate after a downgrade."

Scope and limits

• Genre: applies to content-consumption titles that must keep producing (anime-style, narrative, level-based). Re-verify for PvP- or social-retention-driven titles — their retention does not depend mainly on new content supply, so the impact path differs.
• Stage: mainly mature and late-lifecycle products. Do not use it during a growth phase: pausing output cuts off the landing page for UA spend, shutting two engines at once.
• Region: both cases are Chinese anime-style mobile titles. Chinese players' forgiveness may partly reflect accumulated affection for legacy IPs (the Gujian IP; Soul Tide's long-tenured players). This was not separated out, and it is where this card is most likely to be misapplied. Acceptance of equivalent notices in Western markets is unknown.
• Scale: both notices came from small-to-mid-sized projects. For a high-DAU title, a downgrade notice triggers media, channel and competitor chain reactions; both pacing and wording need rework.
• Compliance: any promise about the availability of already-sold content must be checked against the user agreement, channel rules and local consumer protection law (see kb/OPS-03-合规即设计约束.md).

What would falsify this

1. If titles that announced a formal downgrade with compensation and permanent conversions show no better 30/90-day retention and revenue than titles that quietly reduced output, the method fails — it may only make the operator feel better, with no player-side difference.
2. If a title held a hard cadence (no slowdown, updates on schedule) and successfully rode out a capacity crisis, downgrading is not the only answer and this card must be rewritten as "one path among several."
3. If the forgiveness actually comes from accumulated IP affection rather than notice structure, the takeaway degrades to "only works if you have a beloved legacy IP," and is unusable for new products.

Weakest link: this card attributes "players forgave" to the notice structure (stops / continuations / compensation / server commitment), but controls for nothing — both cases share Chinese anime-game context, legacy-IP affection, and candid first-party wording. Which one is doing the work cannot be separated. Confirming it needs a title with no IP affection that used the same notice structure, to see whether the reaction is equally mild.

Unverified

• 30/90-day retention, revenue, and actual rerun participation rates after the Baijing and Soul Tide notices.
• The body of the CEDEC2026 Odoriko Crinoppe talk: what replaced gut-feel operations, and with what parameters (currently headline and standfirst only).
• Whether the "inventory ÷ cycle < 2" gate matches real failure cases — 3–5 cases would be needed to back out a threshold.
• Whether any high-DAU title has a comparable public downgrade case.