Six Apps, One Week of Data: Why a 13× Impression Gap?
同一個開發者、同一週、零行銷，最高與最低曝光差 13 倍。用 App Store Connect 的每週數據，拆解 ASO 的兩個獨立問題：搜不搜得到，跟想不想安裝。
Same developer, same week, zero marketing — a 13× gap between the top and bottom app. Using one App Store Connect weekly report to separate the two independent ASO problems: getting found, and being wanted.
Every Thursday Apple emails an App Store Connect: App Analytics Weekly Summary — impressions, downloads, and conversion for every app in your account. This week I laid all six on one page, and one thing jumped out: same developer, same week, zero marketing across the board — yet the top and bottom apps differ 13× in impressions. That's not luck. It's something you can take apart.
📊 本週數據總表📊 This week's numbers
2026-07 第一週 · 六個 App · 全 organic（無投放）
First week of July 2026 · six apps · fully organic (no ads)
The 30-day daily-units trend tells the same story: after July 1, daily downloads jumped from 0–4 to 8–20, almost entirely from LetPiano (55 units over 30 days). That's a new app's organic honeymoon — Apple gives freshly-listed apps a brief ranking boost. The window is usually only 2–4 weeks, so optimize metadata now, not after it closes.
App Store Connect · 30 天 Units per Day（Jun 9 – Jul 8）· 7 月初的蜜月期尖峰App Store Connect · 30-day units per day (Jun 9 – Jul 8) · the early-July honeymoon spike
🧾 圖中名詞怎麼看？🧾 What do these terms mean?
這張是 App Store Connect 的 Weekly Summary 摘要，不是每個欄位都代表「下載」。前兩個主要看被看見與被取得，Proceeds 看收入，後兩個則是使用狀況，而且受使用者隱私選擇影響。
This is an App Store Connect Weekly Summary. Not every column means “downloads”: the first two describe discovery and acquisition, Proceeds describes earnings, and the last two describe usage subject to user privacy choices.
Impressions（曝光）：App 在 App Store 的 Today、Games、Apps 或 Search 分頁被看到超過 1 秒的次數，包含產品頁瀏覽。它回答「有多少次機會被看見」，不是「有多少人」；同一台裝置重複看到可能重複計算。
ASO 用途：衡量搜尋字詞、標題與 metadata 是否把 App 帶進使用者視野。ASO use: tells you whether keywords, the title, and metadata are bringing the app into view.
App Units（App 單位／取得次數）：使用者第一次點擊「取得」或「購買」而產生的 App／Bundle 取得數。更新、同一 Apple Account 下載到其他裝置，以及同一裝置重新下載，不會再算新的 App Unit；免費 App 的家人共享下載則會計入。
文章表格的「下載」是用這個概念解讀；它不等於唯一使用者人數，也不等於安裝完成數。The article reads “downloads” using this concept; it is not the number of unique people or completed installations.
Proceeds（開發者實得／預估收益）：Apple 估算你可收到的金額，大致是客戶支付價格扣除適用稅款與 Apple 佣金。它不是 App 的總銷售額；Sales and Trends 的數字是估算值，最終入帳應以 Payments and Financial Reports 的結算報告為準。
免費 App 沒有付費交易時，這欄為 0 很正常；它與 ASO 曝光／轉換分析是另一條收入軸。For a free app with no paid transactions, zero is expected; this is a revenue axis separate from ASO discovery and conversion.
Sessions（工作階段）— Opt-in Only：App 被使用至少 2 秒算一次 session；離開背景後再次使用，也可能算另一個 session。Opt-in Only 表示只有同意與開發者分享分析資料的使用者才會被納入，因此是估計值，不是全體使用者的完整總數。
適合看回訪與使用頻率，不適合拿來反推所有下載者。Useful for return usage and frequency, but not for inferring every downloader.
Crashes（閃退次數）— Opt-in Only：符合條件的裝置上回報的 App crash 總次數。同樣只涵蓋 opt-in 使用者，所以「0」不代表絕對沒有閃退；要查是哪一段程式、哪個版本或哪種裝置出錯，應到 Xcode 的 crash reports 看詳細資料。
判讀重點：這是品質警訊，不是使用人數；先搭配版本與裝置分布再決定是否發修正版。Treat it as a quality signal, not a user count; check version and device breakdown before deciding on a fix release.
一句話：Impressions 看「被看見」、App Units 看「被取得」、Proceeds 看「估算實收」，Sessions／Crashes 看「使用與穩定度」；後兩者因為 Opt-in Only，不能當成完整母體統計。In short: Impressions = seen, App Units = acquired, Proceeds = estimated earnings, Sessions/Crashes = usage and stability — with the last two limited by opt-in privacy data.
The evidence is in my own data. LetPiano and SunnyWalker are the same developer, same week, same zero marketing, yet their impressions are 1,036 vs 77 — a 13× gap. The one structural difference is the title:
「Piano for Kids」是真實搜尋詞——每天有家長真的在搜。
「SunnyWalker」是零搜尋量的自創字——沒有人會去搜這個字。
"Piano for Kids" is a real search term — parents actually type it.
"SunnyWalker" is an invented word with zero search volume — nobody searches for it.
App Store 的搜尋權重排序是：App Name > Subtitle > Keyword 欄位。如果標題裡沒有任何人會搜的字，在演算法眼中這個 App 幾乎等於隱形。SunnyWalker 這種「純品牌名」標題，等於把最高權重的欄位浪費掉了。
App Store search weighting goes: App Name > Subtitle > Keyword field. If your title contains no word anyone searches for, the algorithm treats the app as nearly invisible. A pure brand-name title like "SunnyWalker" wastes the single highest-weighted field.
💡 修法（App Name 上限 30 字元）：改成「品牌 + 一組真實描述詞」——例如 SunnyWalker: Sunrise Alarm（26 字元）。這是安全格式；SunnyWalker Alarm Clock Wake Up Morning 這種堆疊會被 Apple 以「關鍵字塞字」退件。Subtitle（30 字元）放第二組詞（morning / wake up / alarm clock），Keyword 欄位（100 字元）別重複、只放長尾。💡 Fix (App Name max 30 chars): use "brand + one real descriptor" — e.g. SunnyWalker: Sunrise Alarm (26 chars). That's the safe format; SunnyWalker Alarm Clock Wake Up Morning gets rejected as keyword stuffing. Put a second phrase in the Subtitle (30 chars), and only long-tail, non-repeating terms in the Keyword field (100 chars).
每週摘要信原圖 · LetPiano（1,036→47）對照 Lode iOS（997→6）The raw weekly-summary email · LetPiano (1,036→47) next to Lode iOS (997→6)
🎯 一個容易混淆的重點：這是兩個獨立的問題🎯 The key confusion: these are two separate problems
The most important lesson from this data isn't any single number — it's realizing that impressions and conversion measure two entirely different things with entirely different fixes. Mix them up and you fix neither.
Lode iOS is the textbook case: 997 people saw it, only 6 installed. Its impressions nearly match LetPiano's, so the problem isn't keywords at all — 991 people saw the product page and swiped away. Optimizing keywords here does nothing; the fix is the page. SunnyWalker, by contrast, has both holes: can't be found (77 impressions) and isn't wanted when found (0% conversion) — so renaming and reworking the first screenshot have to happen together, or it stays at zero.
🩺 逐一診斷（按投資報酬排序）🩺 App-by-app triage (by ROI)
1Lode iOS · 轉換率急救Lode iOS · conversion first aid
最大的浪費：997 曝光只換 6 下載。問題不在關鍵字（曝光已接近 LetPiano），在商店頁面賣相。檢查順序：① 第一張截圖有沒有一句話賣點（不是純 UI 圖）② icon 在搜尋列表跟一堆筆記/掃描競品排在一起時顯不顯眼 ③ promotional text 前兩行。目標：0.6% → 2%，同樣曝光下載直接 ×3。
The biggest waste: 997 impressions, 6 downloads. Not a keyword problem (impressions rival LetPiano's) — a product-page problem. Check, in order: ① does the first screenshot carry one clear selling line (not a bare UI shot); ② does the icon stand out among notes/scanner rivals in the results list; ③ the first two lines of promotional text. Target 0.6% → 2% — same impressions, 3× the installs.
Adding alarm is right — follow the "brand + descriptor" formula. But don't forget the second hole: 0 downloads from 77 impressions. Even if a rename lifts impressions to 500, sub-0.6% conversion still yields ~3 installs. So the first screenshot must say "what it is + why it's lovable" in half a second — study how LetPiano won.
3Go Kids · 剛上架，先觀察 indexGo Kids · just launched, watch the index
Go Kids 於 7/7 才正式上架，這週摘要的搜尋曝光是 0——對一個才兩三天大的新 listing 來說是正常的（還沒進搜尋索引池，這週的 9 個下載幾乎都來自直連分享）。要盯的是：接下來 3–7 天曝光有沒有從 0 開始長。若超過一週在 App Store 直接搜全名仍找不到，就該開 ticket。
Go Kids went live only on July 7, so this week's search impressions are 0 — normal for a two-day-old listing (not yet in the search index; this week's 9 units came mostly from direct links). What to watch: whether impressions start climbing from zero over the next 3–7 days. If a full-name search still can't find it after a week, open a ticket.
4LetPiano · 趁蜜月期加碼LetPiano · press the honeymoon
目前唯一有自然流量的引擎。趁排名加權還在：確認 zh-Hant metadata 已填、iPad 截圖齊全（兒童 App 的 iPad 佔比高）、可考慮開 Apple Search Ads 最小預算收割 piano for kids 流量。之後所有 App 的標題都照它的公式：品牌字 + 真實搜尋詞。
The only app with real organic traffic right now. While the boost lasts: confirm zh-Hant metadata is filled, iPad screenshots are complete (kids apps skew heavily to iPad), and consider a minimal Apple Search Ads budget to harvest piano for kids traffic. Every future title follows its formula: brand word + real search term.
179 impressions is too low. Check whether the Subtitle/Keyword field covers what parents actually search: sight words / spelling / flashcards / ESL — "vocabulary" is grown-up phrasing that parents don't necessarily type.
🗂 全家族共同事項：每個 App 的 zh-Hant（台灣）metadata 獨立填好——那是免費的第二組關鍵字空間，台灣家長搜的是「鬧鐘」「起床」「注音」不是英文。兒童 App 之間在 promotional text 互相提到家族其他 App（App Store 允許提自家 App）。每週摘要存檔後只盯兩個數字：曝光（搜不搜得到）和轉換率（想不想裝）——兩個問題解法完全不同，不要混在一起改。🗂 Whole-family to-dos:Fill each app's zh-Hant (Taiwan) metadata separately — it's a free second keyword field; Taiwanese parents search their own terms, not English. Have the kids' apps cross-reference each other in promotional text (Apple allows mentioning your own apps). Each week, archive the summary and track just two numbers: impressions (get found) and conversion (get wanted) — different problems, different fixes, never mixed.
LetPiano already proved the formula with data: put a real search term in the title → 13× impressions. For every other app, first ask which axis it's stuck on — can't be found, fix metadata; found but not wanted, fix the product page. Measuring and fixing the two axes separately beats any "ASO trick."
數據來源：App Store Connect Weekly Summary（2026-07-09）· 分析與建議由 Fable 5 協作整理Source: App Store Connect Weekly Summary (2026-07-09) · analysis co-drafted with Fable 5
