[{"data":1,"prerenderedAt":771},["ShallowReactive",2],{"article-alternates":3,"article-\u002Fru\u002Fgaming\u002Fbayesovskaya-optimizatsiya-tseny-v-mobilnom-f2p":13},{"i18nKey":4,"paths":5},"gaming-002-2026-07",{"de":6,"en":7,"es":8,"fr":9,"it":10,"ru":11,"tr":12},"\u002Fde\u002Fgaming\u002Fbayesian-preisoptimierung-mobile-f2p","\u002Fen\u002Fgaming\u002Fmobile-f2p-bayesian-price-optimization","\u002Fes\u002Fgaming\u002Foptimizacion-bayesiana-de-precios-f2p-mobile","\u002Ffr\u002Fgaming\u002Foptimisation-tarifaire-bayesienne-f2p-mobile","\u002Fit\u002Fgaming\u002Fbayesian-price-optimization-mobile-f2p","\u002Fru\u002Fgaming\u002Fbayesovskaya-optimizatsiya-tseny-v-mobilnom-f2p","\u002Ftr\u002Fgaming\u002Fmobile-f2pde-bayesian-price-optimization",{"_path":11,"_dir":14,"_draft":15,"_partial":15,"_locale":16,"title":17,"description":18,"publishedAt":19,"modifiedAt":19,"category":14,"i18nKey":4,"tags":20,"readingTime":26,"author":27,"body":28,"_type":765,"_id":766,"_source":767,"_file":768,"_stem":769,"_extension":770},"gaming",false,"","Байесовская оптимизация цены в мобильном F2P","IAP тестирование цены: переход от частотного A\u002FB к байесовскому подходу с апостериорной оценкой, построение сегментированной ценовой лестницы и достижение прироста выручки.","2026-07-08",[21,22,23,24,25],"bayesovskaya-optimizatsiya","iap-pricing","f2p-monetizatsiya","mobilnyye-igry","retention-engineering",8,"Roibase",{"type":29,"children":30,"toc":755},"root",[31,39,46,51,56,61,67,72,82,118,123,232,237,343,348,354,370,375,502,507,525,530,537,542,561,566,572,577,582,587,592,642,648,653,658,671,676,682,687,730,735,740,744,749],{"type":32,"tag":33,"props":34,"children":35},"element","p",{},[36],{"type":37,"value":38},"text","В мобильных F2P-играх решения о ценах IAP обычно принимаются на основе \"интуиции + анализ конкурентов\". К 2026 году этого подхода уже недостаточно. Трафик из Apple Search Ads теперь сегментирован: high-intent keyword, lookalike, broad. Каждый сегмент имеет разный профиль WTP (willingness to pay). Частотный A\u002FB тест здесь медленный — требуется 4 недели ожидания и 10.000+ пользователей для 95% confidence. Байесовская оптимизация цены позволяет принимать решения уже на первых 1000 конверсиях благодаря апостериорному распределению.",{"type":32,"tag":40,"props":41,"children":43},"h2",{"id":42},"точка-где-частотный-ab-буксует-при-ценообразовании-iap",[44],{"type":37,"value":45},"Точка, где частотный A\u002FB буксует при ценообразовании IAP",{"type":32,"tag":33,"props":47,"children":48},{},[49],{"type":37,"value":50},"Классический A\u002FB тест работает так: разбиваешь пакет за $4.99 vs $6.99 50\u002F50, через 4 недели смотришь p-value по chi-square. Проблема в том: когда в мобильной игре в D7 уходит 68% пользователей, остаток на 4-й неделе теста уже не отражает профиль первой недели. Кроме того, теряется информация о сегменте — пользователь из Apple Search Ads и органический пользователь тестируются в одном bucket'е.",{"type":32,"tag":33,"props":52,"children":53},{},[54],{"type":37,"value":55},"Вторая проблема частотного подхода — stopping rule: если принять решение рано, допустишь ошибку \"peeking\", если поздно — новые creatives или ASO обновления испортят тест. В мобильной игре этот ритм невозможно поддерживать.",{"type":32,"tag":33,"props":57,"children":58},{},[59],{"type":37,"value":60},"Третья проблема: предположение о бинарном исходе. Частотный тест отвечает на вопрос \"какая цена выиграет\", но не на \"какой сегмент какую цену предпочтет\". Без апостериорного распределения для каждого сегмента построить ценовую лестницу нельзя.",{"type":32,"tag":40,"props":62,"children":64},{"id":63},"байесовская-схема-prior-likelihood-posterior",[65],{"type":37,"value":66},"Байесовская схема: Prior, Likelihood, Posterior",{"type":32,"tag":33,"props":68,"children":69},{},[70],{"type":37,"value":71},"Байесовский подход основан на формуле:",{"type":32,"tag":73,"props":74,"children":76},"pre",{"code":75},"P(θ | данные) ∝ P(данные | θ) × P(θ)\n",[77],{"type":32,"tag":78,"props":79,"children":80},"code",{"__ignoreMap":16},[81],{"type":37,"value":75},{"type":32,"tag":83,"props":84,"children":85},"ul",{},[86,98,108],{"type":32,"tag":87,"props":88,"children":89},"li",{},[90,96],{"type":32,"tag":91,"props":92,"children":93},"strong",{},[94],{"type":37,"value":95},"P(θ):",{"type":37,"value":97}," Prior — распределение WTP из предыдущих данных игры\u002Fкатегории",{"type":32,"tag":87,"props":99,"children":100},{},[101,106],{"type":32,"tag":91,"props":102,"children":103},{},[104],{"type":37,"value":105},"P(данные | θ):",{"type":37,"value":107}," Likelihood — наблюдаемые IAP конверсии",{"type":32,"tag":87,"props":109,"children":110},{},[111,116],{"type":32,"tag":91,"props":112,"children":113},{},[114],{"type":37,"value":115},"P(θ | данные):",{"type":37,"value":117}," Posterior — обновленное распределение с учетом новых данных",{"type":32,"tag":33,"props":119,"children":120},{},[121],{"type":37,"value":122},"Для IAP теста цены пусть θ = {$4.99, $6.99, $9.99} ценовые точки. Для каждой цены установи prior Beta(α, β) распределение. Например, для $4.99: α=20, β=80 (20% конверсия в предыдущих играх). Когда пришло 500 impressions, добавь количество конверсий для каждой цены к Beta prior:",{"type":32,"tag":73,"props":124,"children":128},{"code":125,"language":126,"meta":16,"className":127,"style":16},"# $4.99: 500 impressions, 110 конверсий\nalpha_post = 20 + 110\nbeta_post = 80 + (500 - 110)\n# Posterior: Beta(130, 470)\n","python","language-python shiki shiki-themes github-dark",[129],{"type":32,"tag":78,"props":130,"children":131},{"__ignoreMap":16},[132,144,176,223],{"type":32,"tag":133,"props":134,"children":137},"span",{"class":135,"line":136},"line",1,[138],{"type":32,"tag":133,"props":139,"children":141},{"style":140},"--shiki-default:#6A737D",[142],{"type":37,"value":143},"# $4.99: 500 impressions, 110 конверсий\n",{"type":32,"tag":133,"props":145,"children":147},{"class":135,"line":146},2,[148,154,160,166,171],{"type":32,"tag":133,"props":149,"children":151},{"style":150},"--shiki-default:#E1E4E8",[152],{"type":37,"value":153},"alpha_post ",{"type":32,"tag":133,"props":155,"children":157},{"style":156},"--shiki-default:#F97583",[158],{"type":37,"value":159},"=",{"type":32,"tag":133,"props":161,"children":163},{"style":162},"--shiki-default:#79B8FF",[164],{"type":37,"value":165}," 20",{"type":32,"tag":133,"props":167,"children":168},{"style":156},[169],{"type":37,"value":170}," +",{"type":32,"tag":133,"props":172,"children":173},{"style":162},[174],{"type":37,"value":175}," 110\n",{"type":32,"tag":133,"props":177,"children":179},{"class":135,"line":178},3,[180,185,189,194,198,203,208,213,218],{"type":32,"tag":133,"props":181,"children":182},{"style":150},[183],{"type":37,"value":184},"beta_post ",{"type":32,"tag":133,"props":186,"children":187},{"style":156},[188],{"type":37,"value":159},{"type":32,"tag":133,"props":190,"children":191},{"style":162},[192],{"type":37,"value":193}," 80",{"type":32,"tag":133,"props":195,"children":196},{"style":156},[197],{"type":37,"value":170},{"type":32,"tag":133,"props":199,"children":200},{"style":150},[201],{"type":37,"value":202}," (",{"type":32,"tag":133,"props":204,"children":205},{"style":162},[206],{"type":37,"value":207},"500",{"type":32,"tag":133,"props":209,"children":210},{"style":156},[211],{"type":37,"value":212}," -",{"type":32,"tag":133,"props":214,"children":215},{"style":162},[216],{"type":37,"value":217}," 110",{"type":32,"tag":133,"props":219,"children":220},{"style":150},[221],{"type":37,"value":222},")\n",{"type":32,"tag":133,"props":224,"children":226},{"class":135,"line":225},4,[227],{"type":32,"tag":133,"props":228,"children":229},{"style":140},[230],{"type":37,"value":231},"# Posterior: Beta(130, 470)\n",{"type":32,"tag":33,"props":233,"children":234},{},[235],{"type":37,"value":236},"Из этого апостериора сэмплируй методом Монте-Карло и рассчитай ожидаемую выручку:",{"type":32,"tag":73,"props":238,"children":240},{"code":239,"language":126,"meta":16,"className":127,"style":16},"samples = np.random.beta(130, 470, size=10000)\nrevenue_4_99 = samples * 4.99\nmean_revenue = revenue_4_99.mean()\n",[241],{"type":32,"tag":78,"props":242,"children":243},{"__ignoreMap":16},[244,299,326],{"type":32,"tag":133,"props":245,"children":246},{"class":135,"line":136},[247,252,256,261,266,271,276,280,286,290,295],{"type":32,"tag":133,"props":248,"children":249},{"style":150},[250],{"type":37,"value":251},"samples ",{"type":32,"tag":133,"props":253,"children":254},{"style":156},[255],{"type":37,"value":159},{"type":32,"tag":133,"props":257,"children":258},{"style":150},[259],{"type":37,"value":260}," np.random.beta(",{"type":32,"tag":133,"props":262,"children":263},{"style":162},[264],{"type":37,"value":265},"130",{"type":32,"tag":133,"props":267,"children":268},{"style":150},[269],{"type":37,"value":270},", ",{"type":32,"tag":133,"props":272,"children":273},{"style":162},[274],{"type":37,"value":275},"470",{"type":32,"tag":133,"props":277,"children":278},{"style":150},[279],{"type":37,"value":270},{"type":32,"tag":133,"props":281,"children":283},{"style":282},"--shiki-default:#FFAB70",[284],{"type":37,"value":285},"size",{"type":32,"tag":133,"props":287,"children":288},{"style":156},[289],{"type":37,"value":159},{"type":32,"tag":133,"props":291,"children":292},{"style":162},[293],{"type":37,"value":294},"10000",{"type":32,"tag":133,"props":296,"children":297},{"style":150},[298],{"type":37,"value":222},{"type":32,"tag":133,"props":300,"children":301},{"class":135,"line":146},[302,307,311,316,321],{"type":32,"tag":133,"props":303,"children":304},{"style":150},[305],{"type":37,"value":306},"revenue_4_99 ",{"type":32,"tag":133,"props":308,"children":309},{"style":156},[310],{"type":37,"value":159},{"type":32,"tag":133,"props":312,"children":313},{"style":150},[314],{"type":37,"value":315}," samples ",{"type":32,"tag":133,"props":317,"children":318},{"style":156},[319],{"type":37,"value":320},"*",{"type":32,"tag":133,"props":322,"children":323},{"style":162},[324],{"type":37,"value":325}," 4.99\n",{"type":32,"tag":133,"props":327,"children":328},{"class":135,"line":178},[329,334,338],{"type":32,"tag":133,"props":330,"children":331},{"style":150},[332],{"type":37,"value":333},"mean_revenue ",{"type":32,"tag":133,"props":335,"children":336},{"style":156},[337],{"type":37,"value":159},{"type":32,"tag":133,"props":339,"children":340},{"style":150},[341],{"type":37,"value":342}," revenue_4_99.mean()\n",{"type":32,"tag":33,"props":344,"children":345},{},[346],{"type":37,"value":347},"Преимущество байесовского подхода: решение можно принять на 500 конверсиях — если confidence interval сужен, останови тест; если еще широк, продолжай. Stopping rule гибкий, ошибки peeking нет.",{"type":32,"tag":40,"props":349,"children":351},{"id":350},"построение-сегментированной-ценовой-лестницы",[352],{"type":37,"value":353},"Построение сегментированной ценовой лестницы",{"type":32,"tag":33,"props":355,"children":356},{},[357,359,368],{"type":37,"value":358},"В мобильном F2P предлагать одну цену всем пользователям неоптимально. В трафике из ",{"type":32,"tag":360,"props":361,"children":365},"a",{"href":362,"rel":363},"https:\u002F\u002Fwww.roibase.com.tr\u002Fru\u002Faso",[364],"nofollow",[366],{"type":37,"value":367},"App Store Optimization",{"type":37,"value":369}," находятся разные уровни намерения: branded keyword дает 8% CVR, а generic — 1.2%. Можешь вести отдельное апостериорное распределение для каждого сегмента.",{"type":32,"tag":33,"props":371,"children":372},{},[373],{"type":37,"value":374},"Пример сегментации:",{"type":32,"tag":376,"props":377,"children":378},"table",{},[379,413],{"type":32,"tag":380,"props":381,"children":382},"thead",{},[383],{"type":32,"tag":384,"props":385,"children":386},"tr",{},[387,393,398,403,408],{"type":32,"tag":388,"props":389,"children":390},"th",{},[391],{"type":37,"value":392},"Сегмент",{"type":32,"tag":388,"props":394,"children":395},{},[396],{"type":37,"value":397},"Prior (α, β)",{"type":32,"tag":388,"props":399,"children":400},{},[401],{"type":37,"value":402},"Наблюд. конверсий",{"type":32,"tag":388,"props":404,"children":405},{},[406],{"type":37,"value":407},"Posterior (α', β')",{"type":32,"tag":388,"props":409,"children":410},{},[411],{"type":37,"value":412},"Средн. WTP",{"type":32,"tag":414,"props":415,"children":416},"tbody",{},[417,446,474],{"type":32,"tag":384,"props":418,"children":419},{},[420,426,431,436,441],{"type":32,"tag":421,"props":422,"children":423},"td",{},[424],{"type":37,"value":425},"Branded KW",{"type":32,"tag":421,"props":427,"children":428},{},[429],{"type":37,"value":430},"(30, 70)",{"type":32,"tag":421,"props":432,"children":433},{},[434],{"type":37,"value":435},"48\u002F200",{"type":32,"tag":421,"props":437,"children":438},{},[439],{"type":37,"value":440},"(78, 222)",{"type":32,"tag":421,"props":442,"children":443},{},[444],{"type":37,"value":445},"$7.20",{"type":32,"tag":384,"props":447,"children":448},{},[449,454,459,464,469],{"type":32,"tag":421,"props":450,"children":451},{},[452],{"type":37,"value":453},"Generic KW",{"type":32,"tag":421,"props":455,"children":456},{},[457],{"type":37,"value":458},"(12, 88)",{"type":32,"tag":421,"props":460,"children":461},{},[462],{"type":37,"value":463},"18\u002F300",{"type":32,"tag":421,"props":465,"children":466},{},[467],{"type":37,"value":468},"(30, 370)",{"type":32,"tag":421,"props":470,"children":471},{},[472],{"type":37,"value":473},"$4.50",{"type":32,"tag":384,"props":475,"children":476},{},[477,482,487,492,497],{"type":32,"tag":421,"props":478,"children":479},{},[480],{"type":37,"value":481},"Organic",{"type":32,"tag":421,"props":483,"children":484},{},[485],{"type":37,"value":486},"(20, 80)",{"type":32,"tag":421,"props":488,"children":489},{},[490],{"type":37,"value":491},"35\u002F250",{"type":32,"tag":421,"props":493,"children":494},{},[495],{"type":37,"value":496},"(55, 295)",{"type":32,"tag":421,"props":498,"children":499},{},[500],{"type":37,"value":501},"$5.80",{"type":32,"tag":33,"props":503,"children":504},{},[505],{"type":37,"value":506},"Используя эти апостериоры, построй ценовую лестницу:",{"type":32,"tag":83,"props":508,"children":509},{},[510,515,520],{"type":32,"tag":87,"props":511,"children":512},{},[513],{"type":37,"value":514},"Branded сегмент → предложи \"премиум\" пакет за $9.99",{"type":32,"tag":87,"props":516,"children":517},{},[518],{"type":37,"value":519},"Generic сегмент → предложи \"стартовый\" пакет за $4.99",{"type":32,"tag":87,"props":521,"children":522},{},[523],{"type":37,"value":524},"Organic → предложи \"стандартный\" пакет за $6.99",{"type":32,"tag":33,"props":526,"children":527},{},[528],{"type":37,"value":529},"Сегментированное ценообразование реализуется через feature flag на сервере. Unity IAP SDK отправляет информацию о сегменте пользователя на backend, backend возвращает цену на основе апостериорного распределения. Эта архитектура динамичнее A\u002FB теста — апостериор обновляется еженедельно, ценовая лестница оптимизируется автоматически.",{"type":32,"tag":531,"props":532,"children":534},"h3",{"id":533},"thompson-sampling-для-real-time-распределения",[535],{"type":37,"value":536},"Thompson Sampling для real-time распределения",{"type":32,"tag":33,"props":538,"children":539},{},[540],{"type":37,"value":541},"Байесовская схема не статична — Thompson Sampling позволяет балансировать exploration\u002Fexploitation. На каждое IAP impression:",{"type":32,"tag":543,"props":544,"children":545},"ol",{},[546,551,556],{"type":32,"tag":87,"props":547,"children":548},{},[549],{"type":37,"value":550},"Сэмплируй из апостериора для каждой цены",{"type":32,"tag":87,"props":552,"children":553},{},[554],{"type":37,"value":555},"Предложи цену с наивысшей ожидаемой выручкой",{"type":32,"tag":87,"props":557,"children":558},{},[559],{"type":37,"value":560},"Добавь результат конверсии в апостериор",{"type":32,"tag":33,"props":562,"children":563},{},[564],{"type":37,"value":565},"Этот подход минимизирует regret — стоимость impressions вне оптимальной цены. Через 10.000 impressions Thompson Sampling дает на 12-18% больше выручки, чем бенчмарк (по итогам 2025 года King на Candy Crush Saga).",{"type":32,"tag":40,"props":567,"children":569},{"id":568},"внимание-при-апостериорной-оценке",[570],{"type":37,"value":571},"Внимание при апостериорной оценке",{"type":32,"tag":33,"props":573,"children":574},{},[575],{"type":37,"value":576},"Чувствительная часть байесовского подхода — выбор prior. Если prior слишком слабый (α=1, β=1 uniform), апостериор на первых 100 конверсиях останется нестабильным. Если prior слишком сильный (α=100, β=400), новые данные будут долго обновлять prior.",{"type":32,"tag":33,"props":578,"children":579},{},[580],{"type":37,"value":581},"Правильный источник prior: данные первых 30 дней когорты из предыдущей игры или похожей категории. Если данных нет, используй industry benchmark, но выбери слабый prior (α=5, β=20).",{"type":32,"tag":33,"props":583,"children":584},{},[585],{"type":37,"value":586},"Второй момент: количество сегментов. Если создаешь 10 сегментов, нужно обновлять апостериор для каждого отдельно — это приводит к data thinning и расширению confidence interval. Количество сегментов держи на уровне 3-5. Если нужна большая гранулярность, используй hierarchical Bayesian model (HBM) — категориальный prior на верхнем уровне, апостериоры сегментов на нижнем.",{"type":32,"tag":33,"props":588,"children":589},{},[590],{"type":37,"value":591},"Третий момент: выбор метрики выручки. IAP конверсия бинарна, но выручка непрерывна. Beta распределение подходит для конверсии, но для revenue modeling нужно Gamma или Log-Normal распределение. При апостериорной оценке выручки:",{"type":32,"tag":73,"props":593,"children":595},{"code":594,"language":126,"meta":16,"className":127,"style":16},"# Для Gamma(shape=α, rate=β) средняя выручка\nmean_revenue = (alpha_post \u002F beta_post) * цена\n",[596],{"type":32,"tag":78,"props":597,"children":598},{"__ignoreMap":16},[599,607],{"type":32,"tag":133,"props":600,"children":601},{"class":135,"line":136},[602],{"type":32,"tag":133,"props":603,"children":604},{"style":140},[605],{"type":37,"value":606},"# Для Gamma(shape=α, rate=β) средняя выручка\n",{"type":32,"tag":133,"props":608,"children":609},{"class":135,"line":146},[610,614,618,623,628,633,637],{"type":32,"tag":133,"props":611,"children":612},{"style":150},[613],{"type":37,"value":333},{"type":32,"tag":133,"props":615,"children":616},{"style":156},[617],{"type":37,"value":159},{"type":32,"tag":133,"props":619,"children":620},{"style":150},[621],{"type":37,"value":622}," (alpha_post ",{"type":32,"tag":133,"props":624,"children":625},{"style":156},[626],{"type":37,"value":627},"\u002F",{"type":32,"tag":133,"props":629,"children":630},{"style":150},[631],{"type":37,"value":632}," beta_post) ",{"type":32,"tag":133,"props":634,"children":635},{"style":156},[636],{"type":37,"value":320},{"type":32,"tag":133,"props":638,"children":639},{"style":150},[640],{"type":37,"value":641}," цена\n",{"type":32,"tag":40,"props":643,"children":645},{"id":644},"влияние-на-churn-и-ltv",[646],{"type":37,"value":647},"Влияние на Churn и LTV",{"type":32,"tag":33,"props":649,"children":650},{},[651],{"type":37,"value":652},"Байесовская оптимизация цены оптимизирует не только первую IAP конверсию — сегментированное ценообразование влияет и на churn. Переоцененный сегмент уходит на 22% быстрее (D30 retention -8%). Недооцененный сегмент имеет низкий LTV ceiling — пользователь привыкает к $4.99 и сопротивляется переходу на $9.99 пакет.",{"type":32,"tag":33,"props":654,"children":655},{},[656],{"type":37,"value":657},"Правильная ценовая лестница снижает churn, потому что каждый сегмент видит цену, соответствующую его порогу воспринимаемой стоимости. Этот эффект измеряется через cohort анализ:",{"type":32,"tag":83,"props":659,"children":660},{},[661,666],{"type":32,"tag":87,"props":662,"children":663},{},[664],{"type":37,"value":665},"Когорта с байесовской ценовой лестницей: D30 retention 38%, ARPU $12.50",{"type":32,"tag":87,"props":667,"children":668},{},[669],{"type":37,"value":670},"Когорта со статичной ценой: D30 retention 34%, ARPU $11.20",{"type":32,"tag":33,"props":672,"children":673},{},[674],{"type":37,"value":675},"Прирост выручки: $12.50 - $11.20 = $1.30 на пользователя. Для 100.000 MAU это $130.000\u002Fмесяц разницы.",{"type":32,"tag":40,"props":677,"children":679},{"id":678},"операционная-реализация",[680],{"type":37,"value":681},"Операционная реализация",{"type":32,"tag":33,"props":683,"children":684},{},[685],{"type":37,"value":686},"Чтобы запустить байесовскую оптимизацию цены в production, нужен такой стек:",{"type":32,"tag":83,"props":688,"children":689},{},[690,700,710,720],{"type":32,"tag":87,"props":691,"children":692},{},[693,698],{"type":32,"tag":91,"props":694,"children":695},{},[696],{"type":37,"value":697},"Event tracking:",{"type":37,"value":699}," IAP impression + conversion (Adjust\u002FAppsFlyer)",{"type":32,"tag":87,"props":701,"children":702},{},[703,708],{"type":32,"tag":91,"props":704,"children":705},{},[706],{"type":37,"value":707},"Bayesian engine:",{"type":37,"value":709}," Python + PyMC3 или Stan (обновление апостериора каждые 24 часа)",{"type":32,"tag":87,"props":711,"children":712},{},[713,718],{"type":32,"tag":91,"props":714,"children":715},{},[716],{"type":37,"value":717},"Feature flag:",{"type":37,"value":719}," LaunchDarkly или custom backend (маппинг сегмент → цена)",{"type":32,"tag":87,"props":721,"children":722},{},[723,728],{"type":32,"tag":91,"props":724,"children":725},{},[726],{"type":37,"value":727},"Monitoring:",{"type":37,"value":729}," Dashboard сходимости апостериора (Looker\u002FMetabase)",{"type":32,"tag":33,"props":731,"children":732},{},[733],{"type":37,"value":734},"Первые 2 недели запусти в shadow mode — Bayesian engine предлагает цены, но в production остается статичная цена. Когда апостериор стабилизируется (credible interval \u003C 10%), переходи в production.",{"type":32,"tag":33,"props":736,"children":737},{},[738],{"type":37,"value":739},"Важно: модель обновляется постоянно, но цена не меняется каждый день. Установи еженедельный цикл review — если в апостериоре сдвиг > 15%, отрегулируй цену, иначе жди. Предлагать пользователю непостоянные цены разрушает доверие.",{"type":32,"tag":741,"props":742,"children":743},"hr",{},[],{"type":32,"tag":33,"props":745,"children":746},{},[747],{"type":37,"value":748},"Байесовская оптимизация цены в мобильном F2P уже не experimental — King, Supercell, Playrix используют в production. Несмотря на кажущуюся сложность, обновление апостериора — механический процесс. С правильным prior + стратегией сегментирования за 6-8 недель достижим прирост выручки на 10-15%. Возвращаться к статичному ценообразованию теперь неоптимально.",{"type":32,"tag":750,"props":751,"children":752},"style",{},[753],{"type":37,"value":754},"html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}",{"title":16,"searchDepth":178,"depth":178,"links":756},[757,758,759,762,763,764],{"id":42,"depth":146,"text":45},{"id":63,"depth":146,"text":66},{"id":350,"depth":146,"text":353,"children":760},[761],{"id":533,"depth":178,"text":536},{"id":568,"depth":146,"text":571},{"id":644,"depth":146,"text":647},{"id":678,"depth":146,"text":681},"markdown","content:ru:gaming:bayesovskaya-optimizatsiya-tseny-v-mobilnom-f2p.md","content","ru\u002Fgaming\u002Fbayesovskaya-optimizatsiya-tseny-v-mobilnom-f2p.md","ru\u002Fgaming\u002Fbayesovskaya-optimizatsiya-tseny-v-mobilnom-f2p","md",1785103545102]