Japan Steam Radar — Sample Report (Full Version v0.2)
Title: PEAK (AppID 3527290) — Team PEAK / Landfall + Aggro Crab Type: Full sample report. Built to meet the six v0.2 acceptance conditions set by the independent red-team review of v0.1 (2026-11-18). Date of this sample: 2026-11-19 Price point this sample is testing: $49/month (Solo, 1 title) / $99/month (Studio, up to 5 titles) Research note: Research-phase sample. No sales claim is made. Every figure below cites its source and snapshot date. Where a figure is an estimate, the method and its limits are stated next to it. Sections that could not be verified are labelled as such, rather than filled in.
0. What changed from v0.1 — and a correction to the pilot
v0.1 could not publish a direct classification of Japanese reviews. v0.2 is built on a full census, not a sample: all 1,113 Japanese-language Steam reviews for PEAK were retrieved on 2026-11-19 through Valve's official Web API (IUserReviewsService/GetAppReviews, keyless, ¥0) and classified.
Correction that matters (read this before quoting the pilot):
The earlier 100-review pilot (2026-11-19 04:24) reported difficulty mentions at 40/100 and Deep Forest mentions at 14/100. That pilot used the first page in API return order. Re-checking against the full census shows that first page is not a chronological top-100 and not a random sample — its rates do not generalise.
Full-census rates are lower: difficulty (explicit) 158/1,113 (14.2%), Deep Forest 27/1,113 (2.4%).
The pilot's directional finding survives in stratified form (see §2): Deep Forest mentions are concentrated in the recent window, bug/performance mentions are concentrated in the early window, and Japanese-language mentions disappear in the recent window. Its percentages must not be quoted. This report replaces them.
Method note for production:day_range is capped at 365 in this API, except that day_range=0 means "no limit". A full-history pull therefore requires day_range=0 plus cursor pagination (13 calls for PEAK: 12 pages + 1 empty terminator page). Using 365 silently truncates the launch period — for PEAK that would have dropped 689 of 1,113 reviews.
1. At a glance — Japan in the global picture (same-day, same-method snapshot)
All rows below were pulled on 2026-11-19 via the same official API, purchase_type default (Steam purchases), so the rows are comparable.
1,113 (984 positive / 129 negative in query_summary; 983 / 130 in the fetched census — a 1-review difference, likely a vote changed between calls)
Same, 2026-11-19
Japanese share of reviews
0.356% (1,113 / 313,072)
Calculated from the two rows above
Japanese positive rate
88.4% vs 94.6% all-language and 95.6% English (142,425 / 148,912)
Same, 2026-11-19
Release / price
16 June 2025 / US $7.99; Japan Steam price ¥880; console versions (Switch 2 / PS5 / Xbox Series) released 22–23 Oct 2026 at ¥1,200 (Switch 2) / ¥1,210 (PS5) in Japan
Steam store; Japanese guide Game Guide Scout, published 2026-10-29, updated 2026-11-13 (see §4)
Japanese review volume by month (census)
Jun 2025: 74 / Jul: 96 / Aug 2025: 336 / Sep: 99 / Oct: 49 / Nov 2025: 145 / Dec: 34 / Jan–Nov 2026: 13–44 per month
Full census, §2
Reading: Japan is 0.356% of review volume, and Japanese reviewers rate PEAK 6.2 points lower than the global average on the same day, by the same method. That gap is now free to see since Valve added language-specific scores in August 2025 — what is not free is why the gap exists, whether it is moving, and what to do about it. That is what §2 answers. The August 2025 spike (336 reviews, 30% of all Japanese reviews ever) coincides with the Mountain Day period covered in v0.1; causation is not claimed.
Category comparison (same method, same day) — is PEAK's Japan share unusual?
Title (AppID)
Total reviews
Japanese reviews
JP share
JP positive
All-language positive
PEAK (3527290)
313,072
1,113
0.356%
88.4%
94.6%
R.E.P.O. (3241660)
350,798
1,804
0.514%
92.6%
96.1%
Lethal Company (1966720)
412,289
616
0.149%
91.1%
97.0%
Content Warning (2881650)
73,313
170
0.232%
88.2%
93.0%
Reading: PEAK sits in the middle of its co-op category on Japanese share. R.E.P.O. — which only added official Japanese in v0.5.0 on 29 Oct 2026 (see the R.E.P.O. short sample) — already shows a higher Japanese share and a smaller positive-rate gap. A monthly version of this table is the category watch a global dashboard does not sell by language.
2. What Japanese players are actually saying — full census classification
Population: 1,113 Japanese reviews, posted 2025-06-17 to 2026-11-17 (UTC). Stratification: Early window = 2025-06-17 to 2025-11-17 (689 reviews, 88.7% positive in the fetched census). Recent window = 2025-11-18 to 2026-11-17 (424 reviews, 87.7% positive). The split date is the pilot's window boundary, kept so pilot and census can be compared honestly. Counting: Two layers, both keyword-based and reproducible. Strict counts require an explicit word (e.g. 難易度/難しい). Broad counts use a wider keyword set per theme and overlap (one review can count in several themes). Keyword lists are in the production log. These are mention rates, not sentiment scores — a difficulty mention can be praise.
2a. Strict counts (explicit words only)
Indicator
All 1,113
Early 689
Recent 424
Negative reviews only
Difficulty explicit (難易度/難しい etc.)
158 (14.2%)
99 (14.4%)
59 (13.9%)
25 / 130
Deep Forest biome explicit (深き森/深い森)
27 (2.4%)
8 (1.2%)
19 (4.5%)
14 / 130
Bug explicit (バグ)
38 (3.4%)
30 (4.4%)
8 (1.9%)
11 / 130
Performance narrow (重い/最適化/フリーズ/クラッシュ)
22 (2.0%)
21 (3.0%)
1 (0.2%)
11 / 130
"Japanese" explicit (日本語)
10 (0.9%)
10 (1.5%)
0 (0.0%)
1 / 130
Pause / save / session explicit
28 (2.5%)
20 (2.9%)
8 (1.9%)
4 / 130
No public matchmaking explicit (野良/マッチング)
9 (0.8%)
6 (0.9%)
3 (0.7%)
3 / 130
Motion sickness explicit (酔い incl. 画面酔い)
18 (1.6%)
13 (1.9%)
5 (1.2%)
2 / 130
2b. Broad themes (overlapping)
Theme
All
Early
Recent
Co-op / friends mention
365 (32.8%)
223 (32.4%)
142 (33.5%)
Core praise words
363 (32.6%)
230 (33.4%)
133 (31.4%)
Solo mention
223 (20.0%)
149 (21.6%)
74 (17.5%)
Difficulty / biome / balance (broad)
201 (18.1%)
120 (17.4%)
81 (19.1%)
Network / matchmaking (broad)
139 (12.5%)
96 (13.9%)
43 (10.1%)
Session length / time
125 (11.2%)
74 (10.7%)
51 (12.0%)
Bug / performance (broad)
103 (9.3%)
78 (11.3%)
25 (5.9%)
Update / new content mention
94 (8.4%)
61 (8.9%)
33 (7.8%)
Language / UI / settings (broad)
35 (3.1%)
28 (4.1%)
7 (1.7%)
2c. What this proves — and what changed over time
v0.1's implied story was an early-window story. In the early window, bug mentions (4.4% strict, 11.3% broad), performance mentions (3.0% strict) and language mentions (all 10 "日本語" reviews in the whole census) are concentrated. A report written at launch would correctly have said "fix bugs, performance and language defaults first".
That is no longer the centre of Japanese dissatisfaction. In the recent window, strict performance mentions fall to 1 review (0.2%), language mentions fall to zero, and bug mentions halve. What rises is one specific biome: Deep Forest goes from 1.2% to 4.5% of all reviews, and accounts for 11 of 52 recent negative reviews (14 of 130 negatives overall). The Japanese complaint has narrowed from "the game is rough" to "Stage 2's Deep Forest is mis-tuned".
Difficulty itself is not rejection. Difficulty is mentioned at a stable ~14% in both windows, and most short difficulty mentions in positive reviews frame it as the appeal ("even the easiest difficulty is properly hard", §2d). The negative concentration is biome-specific, not difficulty-general. Treating "Japanese players find it too hard" as the takeaway would misdirect a balance patch.
Shared-with-global vs Japan-visible: No-save / session length and no public matchmaking appear in Japanese reviews (2.5% and 0.8% strict) but also dominate the global theme lists (Niklas Notes, §6) — they are common complaints, not Japan-only ones. The Japan-visible signal in this census is the time shift (early tech/language → recent single-biome balance) and the positive-rate gap (§1), neither of which a one-off global theme list shows.
2d. Short quoted examples (Japanese original excerpt + translation)
Quotes are short excerpts from the census, trimmed, with posting date and the reviewer's recommendation. They illustrate the counts above; they are not a substitute for them.
Deep Forest / Stage-2 difficulty (negative):
「深き森さえなければ協力プレイが楽しめる良ゲー。深き森が本当にゴミ、第二ステージのくせにどのステージよりも難易度が高い」(2026-05-11, not recommended) — "If it weren't for Deep Forest, this would be a great co-op game. Deep Forest is genuinely awful — for a second stage, it is harder than any other stage."
「深い森、あまりにもひどいマップ。ゾンビ、クモ、胞子の3つのギミックがとてもじゃないが難易度をあげすぎている」(2025-12-13, not recommended) — "Deep Forest is just too punishing a map. The zombie, spider and spore gimmicks together push the difficulty up far too much."
「新マップの深い森が本当に面白くないし、とても難しい。」(2025-08-11, not recommended) — "The new Deep Forest map really isn't fun, and it's very hard."
Positive counterweight, same biome: 「深き森?ってのが強風、虫、毒?が多くて難しい。でも、萎えることはない。友人と協力して、絶対にクリアするぞ!!!」(2026-03-03, recommended) — "Deep Forest has wind, bugs, poison — it's hard. But it doesn't crush us. We'll clear it together with friends!!!"
Difficulty as appeal (positive):
「とても面白かった。難易度『簡単』でも大変難しく、ヒリヒリしたプレイを楽しめた。」(2025-06-21, recommended) — "Really fun. Even on 'easy' difficulty it was very hard, and I enjoyed that tension."
Bug / backpack (early window):
「バックパックを床に置くときに床に貫通して消失するバグを何回も遭遇して萎えました。」(2025-07-22, recommended, excerpt) — "I kept hitting the bug where a backpack placed on the floor clips through and disappears. It killed my motivation."
Language default (early window — the complaint class that later disappears):
「初回起動時に英語か日本語以外の言語に設定されているゲームに久しぶりに遭遇した。」(2025-08-14, not recommended) — "First game in a long time that booted up set to a language that was neither English nor Japanese."
「日本語非対応なのは注意! 追記:日本語は追加されましたが、初期言語が韓国語になっている事があるのでゲーム内オプションから変更しましょう!」(2025-06-19, recommended, excerpt) — "Note: no Japanese support! Update: Japanese was added, but the default language is sometimes set to Korean — change it in the in-game options!"
Session length / no save:
「しいて言うならばシングルプレイ時はセーブ機能が欲しかった。一度始めたら成功失敗に限らず終わるまでやめられない」(2026-01-04, recommended, excerpt) — "If I had to pick one thing: I wanted a save function for single-player. Once you start, you can't stop until you succeed or fail."
3. The money side — what 0.356% is worth for this title (estimate, method stated)
Red-team condition: give at least one money-side number, with source and method, as the answer to "why pay to watch 0.35%?".
Unit-sales estimates for PEAK (global, all sources): PureGameStats estimates 6,192,048 Steam units ("high confidence", modelled from multiple signals; snapshot in search results 2026-11-19). A second, white-box Boxleiter estimate (indielist) gives a much wider 8.5M–19.8M units, median 14.1M. The two methods disagree by more than 2× — treat any derived figure as a range.
Derived Japanese-attributable units (illustrative, not measured sales): If Japanese buyers reviewed at the same rate as the global average, Japanese share of units ≈ Japanese share of reviews (0.356%). That gives ≈22,000 units on the PureGameStats base, or ≈50,000 units on the indielist median.
Derived gross at the Japanese Steam price (¥880):≈¥19 million to ≈¥44 million before Steam's cut, regional pricing effects and refunds.
Limits (stated, not buried): Review propensity is not equal across languages — Japanese reviewers are documented as more complaint-driven (§4), which could inflate or deflate this proxy in either direction. Steam does not publish country-level units for other studios' titles. This range answers only the scale question: for a hit of PEAK's size, even a 0.356% review share corresponds to tens of millions of yen of Japanese gross. For a median indie title the same share is far smaller in absolute yen — which is why Radar's pitch cannot rest on share alone, and why §5's moves are about timing and conversion, not share.
4. Japanese media and community — primary mentions
At least two primary Japanese-language sources, not English summaries of them:
Game Guide Scout (Japanese guide site) — PEAK console / crossplay guide. Published 2026-10-29, updated 2026-11-13. Reports, in Japanese and from the Japanese storefronts: console versions launched 22–23 Oct 2026 with crossplay via room codes, Steam ¥880 / Switch 2 ¥1,200 / PS5 ¥1,210, Japanese available on all platforms, online play requires each platform's subscription, and Steam progress does not carry to console. Why it matters: Japanese discovery for PEAK is now a console + crossplay story, not only a Steam story — a Radar issue that only tracks Steam reviews would miss the channel Japanese newcomers actually arrive through.
Japanese player攻略 on note.com — 東雲奈緒, "PEAK攻略のポイントと難関実績解説" (2025-09-19). A long-form Japanese community guide: six major biomes, Mesa as the then-new biome, daily terrain rotation at 02:00 JST, and achievement breakage after the Mesa update. Why it matters: detailed Japanese攻略 existed and aged quickly as updates landed — the audience already produces its own Japanese information layer; a paid report competes on classification, timing and cross-title comparison, not on basic攻略.
AUTOMATON / Shinjisan (Japanese primary utterance, via AUTOMATON WEST, 2025-08-21). Japanese streamer Shinjisan, quoted in Japanese in the article: 「どのゲームも日本のレビューがほぼダントツで低評価をしていました」 — "Whichever game I checked, Japanese reviews were by far the most negative." The article also cites the 2021–22 user analysis of 262 titles (Japanese negative share materially above the worldwide average). Why it matters: this is the cultural context for PEAK's 88.4% vs 94.6% gap in §1 — and Japanese players themselves are publicly debating whether the habit will cost them future Japanese support.
Not found (recorded honestly): A PEAK-specific news article on Famitsu or 4Gamer was not found in this session's searches. No such article is cited or implied above.
Held from v0.1 (still valid): Mountain Day (11 Aug 2025) — Nijisanji's "Yama no Hi PEAK Summer Festival" with 21 livers, after Japanese support was added in July 2025, followed by PEAK's all-time concurrent peak on 17 Aug 2025 (170,759, Steambase). Timing correlation only; causation not claimed.
5. Forward look — next 90 days in Japan (2026-11-19 to 2027-02-17)
Monthly billing needs a forward reason to exist. These are dated, named windows inside the next 90 days — not generic "Japan is active".
Date
Window
Why it is specific to PEAK / this category
23 Nov 2026
Labor Thanksgiving Day (勤労感謝の日), national holiday
Immediate long-weekend play window, four days after this report. Co-op sessions need friends free at the same time — holidays are when the "no friends included" friction (§2b, solo 20.0%) temporarily drops.
17 Dec 2026 – 4 Jan 2027
Steam Winter Sale 2026 (Valve Steamworks calendar: 17 Dec 2026 – 4 Jan 2027, 10am Pacific)
Longest sale of the year (18 days), crossing the Japanese New Year break. This is the first Winter Sale for PEAK's console versions (released Oct 2026) and the first with official Japanese in place from day one of the sale. Discount and Japanese store-copy decisions bind before the sale opens.
1–3 Jan 2027 + 11 Jan 2027
New Year holiday period (1 Jan is a Friday, forming a 3-day break) and Coming of Age Day, Mon 11 Jan 2027 (3-day weekend 9–11 Jan; Cabinet Office 2027 holiday list)
Two long weekends in eleven days. The 20-year-old cohort Coming of Age Day celebrates is core co-op / VTuber audience age. Creator and social beats timed to global patch dates will miss both if nobody converts them to JST holiday dates.
11 Feb 2027
National Foundation Day (建国記念の日), Thursday (Cabinet Office 2027 list)
A single mid-week holiday (a Friday off makes a 4-day break). Last dated holiday window inside the 90 days; Emperor's Birthday (23 Feb 2027) falls just outside it.
Sources for dates: Cabinet Office 2027 holiday list as reproduced in multiple 2027 calendar summaries checked 2026-11-19 (1 Jan Fri / 11 Jan Mon / 11 Feb Thu / 23 Feb Tue); Valve Winter Sale dates via Steamworks event documentation as reported by multiple sale trackers checked 2026-11-19.
6. Next moves (maximum 3 — each tied to a section above, each with execution cost)
General advice that could be written without this report's data has been removed, per the red-team condition.
Before the Winter Sale, publish one Japanese-language post that names the Deep Forest problem and what rotation / difficulty options do about it — on the Steam news hub, in Japanese. Tied to: §2a–2c (Deep Forest = 19 recent mentions, 11/52 recent negatives; language complaints are now zero, so a Japanese post will be read as balance communication, not localisation apology). Execution cost: drafting + Japanese translation of a short post (~1,500–2,500 Japanese characters) ≈ 3–5 person-hours of bilingual staff time, or one small outsourced translation job if no bilingual staff exists; posting itself <1 hour. No media spend required.
Set Japanese store and FAQ expectations for solo play and session length before the New Year console influx. Tied to: §2b–2d (solo mentions 20.0%, session-length 11.2%, save requests in both windows; §4 item 1 — console newcomers arrive via Japanese guides that already warn "friends only, no public matchmaking"). State plainly, in Japanese store copy / FAQ: no public matchmaking, typical run length, and what pause support exists in solo. Execution cost: copy edit on existing store / FAQ pages ≈ 2–4 person-hours including translation check; no build work. The cost of not doing it is measurable in the census: expectation-mismatch reviews are the ones that quote session length verbatim.
Pitch one Japanese creator beat for the 9–11 Jan 2027 long weekend, with pitches out in early December — not a Mountain Day repeat. Tied to: §5 (Coming of Age weekend) + v0.1 §3 (the 2025 Mountain Day event proves Japanese creators will play this game when the date is native; §1 shows August is already PEAK's biggest Japanese month, so January is the untested, less crowded window). Execution cost: building a 15–25 name Japanese creator shortlist + Japanese pitch drafting ≈ 6–10 person-hours; keys/build access are marginal cost. Any paid sponsorship is a separate budget decision — this move is the unpaid pitching version only.
7. What you can get free or cheaper — and what you cannot (honest difference table)
Need
Free / cheaper route
What it actually gives you
What it does not give you
"Is Japan rating us lower?"
Steam language-specific review scores (free, Valve, since Aug 2025)
The score per language. For PEAK today: JP 88.4% vs global 94.6%
Global units / revenue estimates, concurrent players, country breakdowns at the level those tools model
Japanese review content classification, Japanese media / creator context
Discovery / trend data
GameDiscoverCo Plus ($19/mo)
Steam discovery data, weekly trend coverage, community
Japan-specific qualitative reading; it is 2.6× cheaper than Radar Solo and overlaps on everything except the Japanese qualitative layer
DIY reading
Steam language filter + machine translation + a general AI assistant (≈¥0–a few dollars of API use)
A rough gist of Japanese reviews, if you know which questions to ask and re-ask monthly
A reproducible census classification, a stratified time comparison, quoted evidence you can act on, and someone accountable for re-running it every month
Japanese posts on X
Manual spot checks (Radar's method)
Occasional large, named events (e.g. a festival)
Radar does not promise comprehensive X monitoring. Automated X collection is prohibited by X's terms for scraping and, via the official API, is pay-per-post ($0.005/post since Feb 2026) — at category scale it would consume a Solo subscription's revenue by itself. X is a manual auxiliary source only.
Japan Steam Radar (this report)
—
Full-census Japanese review classification with time stratification and quotes (§2), category comparison by language (§1), Japanese media / community primary sources (§4), a dated 90-day Japanese calendar converted into actions (§5–6)
Global quantitative estimates better than Gamalytic's, comprehensive X coverage, or攻略 content Japanese players already produce for free (§4 item 2)
If the three Radar-only rows in the last line are not worth $49/mo to a given studio, the honest recommendation is GameDiscoverCo Plus + the free Steam language score — and this table says so.
8. What this sample proves — and does not prove
Proves:
The paid core can be produced compliantly and cheaply: a full census of 1,113 Japanese reviews was retrieved, classified and quoted for ¥0 cash in one session (§9). The bottleneck is interpretation and writing, not collection.
The classification finds things a free score cannot: a complaint centre that moved (early tech/language → recent single-biome balance), a biome that accounts for a fifth of recent negatives, and a language-complaint class that has ended.
A forward, dated Japanese calendar and category-by-language comparison can be assembled every month from named sources.
Does not prove:
That any developer will pay $49/month. That remains Sprint Premise 1 (confidence 25%, unchanged — no customer data was produced by this internal production step).
That Japanese sales equal the §3 range. §3 is a stated-method proxy range, not measured sales.
That the Deep Forest pattern generalises to other titles. The R.E.P.O. short sample is breadth inventory only; a second full census is required before claiming a category pattern.
9. Production log (cost measurement)
Collection: Steam Web API IUserReviewsService/GetAppReviews, keyless. PEAK Japanese census: filter=Recent, languages=[japanese], day_range=0, cursor pagination — 13 calls, 1,113 unique reviews. Category / total snapshots: 11 further single-review calls (PEAK all/English/Japanese + 4 titles × 2). Total ≈24 calls, far below the 100,000 calls/day API terms limit. Cash cost ¥0. No 429s.
Classification: Keyword scripts (strict + broad sets above), runtime <1 second. Human-equivalent cost is in reading, quote selection, translation and writing — one unattended session for this v0.2, including the pilot-correction re-check.
Other sources: PureGameStats and indielist sales estimates (via search results, 2026-11-19); Game Guide Scout Japanese guide (2026-10-29 / updated 2026-11-13); note.com Japanese攻略 by 東雲奈緒 (2025-09-19); AUTOMATON WEST (2025-08-21) quoting Shinjisan in Japanese; Cabinet Office 2027 holidays via 2027 calendar summaries; Valve Winter Sale dates via Steamworks documentation as reported by sale trackers; Niklas Notes global theme analysis (updated 2026-11-01); Steambase / Nijisanji Wiki via v0.1 for the Mountain Day record.
Not used: Automated X collection (terms + cost, see §7), scraping of Steam reviews outside the official API, any paid tool.
AI cost: Included in the flat Muse subscription; not separately metered per report — recorded as such, not as zero marginal cost in general.
Implication for unit economics: Collection cost is negligible and now measured on a full census, not extrapolated from 100 reviews. Full-version cost is dominated by one session of interpretation/writing per title. Whether that session cost fits inside $49/mo depends on how many titles share the category / calendar work (§1 table, §5) — that sharing is the production design question for the mid-sprint review (2 Dec), not a claim made here.
Direct classification + quotes — Met: §2 (full census 1,113, strict + broad, stratified, quoted excerpts with translations).
Forward calendar ≥3 dated items — Met: §5 (4 windows, named and dated).
Money-side ≥1 number with source + method — Met: §3 (unit range + ¥19M–¥44M gross range, method and limits stated).
Japanese media primary mentions ≥2 — Met: §4 (Game Guide Scout JP, note.com JP攻略, AUTOMATON/Shinjisan JP utterance; Famitsu/4Gamer absence recorded).
Next Moves with no data-free generalities + execution cost — Met: §6 (3 moves, each tied to a section, each with person-hour cost).
Free-alternative difference table — Met: §7 (including the honest "do not promise X coverage" row).
Status note: Meeting this checklist makes v0.2 eligible to be used as the paid-proof sample for the landing page and for outreach through permitted routes. It does not by itself move any sprint gate — gates move only on customer signals in the signal hierarchy (payment > prepayment > post-price registration > intent).