[{"data":1,"prerenderedAt":1167},["ShallowReactive",2],{"site-chrome-ph":3,"site-chrome-en":-1,"site-chrome-th":-1,"blog-page-ph":287,"blog-categories-ph":345,"blog-posts-ph":365,"article-blog-ph-\u002Fblog\u002Fml-demand-forecasting-for-ecommerce":722,"og-copy-ph":1124,"og-posts-ph-none":1166},{"id":4,"extension":5,"footer":6,"header":116,"indexFacets":202,"meta":243,"notFound":244,"stem":250,"ui":251,"__hash__":286},"chrome_en\u002Fen\u002Fchrome.yml","yml",{"columns":7,"legal":83,"quote":84,"askAi":87,"socials":103},[8,29,47,65],{"label":9,"children":10},"Product",[11,14,17,20,23,26],{"label":12,"to":13},"Ads optimization","\u002Fph\u002Fblog\u002Fads-optimization",{"label":15,"to":16},"Unit economics","\u002Fph\u002Fblog\u002Fprofit-margin",{"label":18,"to":19},"Agentic KAM","https:\u002F\u002Flearn.dataglasslabs.com\u002Fdocs\u002Fai-and-api\u002Fassistants",{"label":21,"to":22},"One-click deploy","https:\u002F\u002Flearn.dataglasslabs.com\u002Fdocs\u002Fgrowth\u002Factions\u002Faction-flow",{"label":24,"to":25},"Integrations","https:\u002F\u002Flearn.dataglasslabs.com\u002Fdocs\u002Foverview\u002Fget-started\u002Fsupported-platforms",{"label":27,"to":28},"Pricing","\u002Fph\u002Fpricing",{"label":30,"children":31},"Resources",[32,35,38,41,44],{"label":33,"to":34},"Blog","\u002Fph\u002Fblog",{"label":36,"to":37},"Customer stories","\u002Fph\u002Fcustomers",{"label":39,"to":40},"Research","\u002Fph\u002Fresearch",{"label":42,"to":43},"Glossary","\u002Fph\u002Fglossary",{"label":45,"to":46},"Docs","https:\u002F\u002Flearn.dataglasslabs.com\u002Fdocs",{"label":48,"children":49},"Company",[50,53,56,59,62],{"label":51,"to":52},"About","\u002Fph\u002Fcompany",{"label":54,"to":55},"Contact","\u002Fph\u002Fcontact",{"label":57,"to":58},"Book a demo","\u002Fph\u002Fbook-a-demo",{"label":60,"to":61},"Privacy","\u002Fph\u002Flegal\u002Fprivacy",{"label":63,"to":64},"Terms","\u002Fph\u002Flegal\u002Fterms",{"label":66,"children":67},"DataGlass for",[68,71,74,77,80],{"label":69,"to":70},"Agencies","\u002Fph\u002Ffor\u002Fagencies",{"label":72,"to":73},"Brands","\u002Fph\u002Ffor\u002Fbrands",{"label":75,"to":76},"Planners and optimizers","\u002Fph\u002Ffor\u002Fplanners",{"label":78,"to":79},"Managers and analysts","\u002Fph\u002Ffor\u002Fmanagers",{"label":81,"to":82},"Sellers and shop owners","\u002Fph\u002Ffor\u002Fsellers","© 2026 DataGlass Labs. All rights reserved.",{"text":85,"caption":86},"Measure what you keep, and then grow what you measure.","The DataGlass team",{"label":88,"prompt":89,"assistants":90},"Ask AI about DataGlass","I want to understand DataGlass, a decision engine for Shopee, Lazada and TikTok Shop sellers that optimizes ads, pricing and promotions for true profit. Use web search to read their website at https:\u002F\u002Fwww.dataglasslabs.com (with an agent-readable summary at https:\u002F\u002Fwww.dataglasslabs.com\u002Fllms.txt), then explain what DataGlass does, who it is for and how it differs from an analytics dashboard.",[91,95,99],{"label":92,"icon":93,"url":94},"ChatGPT","chatgpt","https:\u002F\u002Fchatgpt.com\u002F?prompt={prompt}",{"label":96,"icon":97,"url":98},"Claude","claude","https:\u002F\u002Fclaude.ai\u002Fnew?q={prompt}",{"label":100,"icon":101,"url":102},"Perplexity","perplexity","https:\u002F\u002Fwww.perplexity.ai\u002Fsearch?q={prompt}",[104,108,112],{"label":105,"icon":106,"to":107},"LinkedIn","i-lucide-linkedin","https:\u002F\u002Fwww.linkedin.com\u002Fcompany\u002Fdataglass-labs",{"label":109,"icon":110,"to":111},"Facebook","i-lucide-facebook","https:\u002F\u002Fwww.facebook.com\u002Fdataglasslabsth",{"label":113,"icon":114,"to":115},"Instagram","i-lucide-instagram","https:\u002F\u002Fwww.instagram.com\u002Fdataglasslabs\u002F",{"links":117,"cta":196,"secondaryCta":199},[118,152,191,193,194,195],{"label":9,"to":119,"menu":120},"\u002Fph",{"layout":121,"heading":122,"cards":123,"groups":135,"footer":142},"cards","What DataGlass does",[124,128,132],{"label":125,"description":126,"to":13,"ui":127},"True profit ads optimization","Budgets, pacing and Target ROAS tuned for profit.","budget-tomorrow",{"label":129,"description":130,"to":16,"ui":131},"Unit economics breakdown","What you actually earn from every sale.","unit-economics",{"label":18,"description":133,"to":19,"motif":134},"An AI agent that knows your shops.",true,[136],{"label":137,"size":138,"links":139},"More","secondary",[140,141],{"label":21,"to":22},{"label":24,"to":25},[143,148],{"lead":144,"label":145,"icon":146,"to":147},"DataGlass v.2.6.","See what's new","i-lucide-box","https:\u002F\u002Flearn.dataglasslabs.com\u002Freleases",{"lead":149,"label":150,"icon":151,"to":37},"Read our","customer stories","i-lucide-book-open",{"label":30,"to":34,"menu":153},{"layout":154,"groups":155},"directory",[156,164,170,182],{"label":66,"size":157,"links":158},"primary",[159,160,161,162,163],{"label":69,"to":70},{"label":72,"to":73},{"label":75,"to":76},{"label":78,"to":79},{"label":81,"to":82},{"label":165,"size":157,"links":166},"Insights",[167,168,169],{"label":33,"to":34},{"label":36,"to":37},{"label":39,"to":40},{"label":171,"size":157,"links":172},"Learn",[173,174,177,180],{"label":45,"to":46},{"label":175,"to":176},"Use cases","https:\u002F\u002Flearn.dataglasslabs.com\u002Fuse-cases",{"label":178,"to":179},"Training","https:\u002F\u002Flearn.dataglasslabs.com\u002Ftraining",{"label":181,"to":147},"Changelog",{"label":183,"size":138,"links":184},"Support",[185,186,189],{"label":42,"to":43},{"label":187,"to":188},"Brand kit","\u002Fph\u002Fbrand",{"label":190,"to":55},"Contact us",{"label":192,"to":37},"Customers",{"label":27,"to":28},{"label":48,"to":52},{"label":54,"to":55},{"label":197,"to":198},"Start for free","https:\u002F\u002Fapp.dataglasslabs.com\u002Fsign-up",{"label":200,"to":201},"Sign in","https:\u002F\u002Fapp.dataglasslabs.com\u002Fsign-in",{"clear":203,"groups":204},"Clear filters",[205,218],{"key":206,"label":207,"options":208},"platforms","Platform",[209,212,215],{"value":210,"label":211},"shopee","Shopee",{"value":213,"label":214},"lazada","Lazada",{"value":216,"label":217},"tiktok","TikTok Shop",{"key":219,"label":220,"options":221},"format","Format",[222,225,228,231,234,237,240],{"value":223,"label":224},"guide","Guide",{"value":226,"label":227},"article","Article",{"value":229,"label":230},"comparison","Comparison",{"value":232,"label":233},"calculator","Calculator",{"value":235,"label":236},"paper","Paper",{"value":238,"label":239},"survey","Survey",{"value":241,"label":242},"dataset","Dataset",{},{"title":245,"description":246,"cta":247,"errorTitle":249},"Page not found","The page may have moved, or it never existed.",{"label":248,"to":119},"Back to home","Something went wrong","en\u002Fchrome",{"previous":252,"next":253,"pause":254,"play":255,"goTo":256,"expand":257,"close":258,"readMore":259,"learnMore":260,"seeAll":261,"menu":262,"back":263,"share":264,"shareOn":265,"copyLink":266,"copied":267,"onThisPage":268,"minRead":269,"sources":270,"viewMore":271,"searchPosts":272,"allPosts":273,"all":274,"filters":275,"search":276,"searchClose":277,"searchPlaceholder":278,"searchSuggested":279,"searchResults":280,"searchNoResults":281,"searchLoading":282,"relatedTitle":283,"mainNav":284,"footerNav":285},"Previous","Next","Pause","Play","Go to item","Expand","Close","Read more","Learn more","See all","Menu","Back","Share","Share on","Copy link","Link copied","On this page","{n} min read","Sources","View more","Search posts","All posts","All","Filters","Search","Close search","Search DataGlass","Suggested","Results","No results for","Loading","More to discover","Main","Footer","HZQSEulcpFGY0nWnUTNYhG2n06nO97XnUBc_pEVOFxo",{"id":288,"title":33,"blog":289,"body":327,"customers":334,"description":335,"draft":336,"draftReason":334,"extension":337,"glossary":334,"industry":334,"meta":338,"navigation":336,"ogImage":339,"path":340,"research":334,"robots":334,"schemaOrg":334,"seo":341,"showcase":336,"sitemap":342,"stem":343,"__hash__":344},"pages_en\u002Fblog.md",{"emptyLabel":290,"labels":291,"indexLabels":296,"listenLabels":307,"indexCta":320,"ctaButtons":323},"No posts match yet. Try another category or a shorter search.",{"tldr":292,"published":293,"updated":294,"related":295},"In short","Published","Updated","Keep reading",{"categories":297,"sort":298,"newest":299,"oldest":300,"grid":301,"list":302,"loadMore":303,"results":304,"result":305,"clearSearch":306},"Categories","Sort","Newest first","Oldest first","Grid view","List view","Load more","{n} results for “{q}”","1 result for “{q}”","Clear search",{"listen":308,"loading":309,"play":310,"pause":311,"back":312,"forward":313,"speed":314,"seek":315,"player":316,"close":317,"follow":318,"of":319},"Listen to article","Loading…","Play audio of this article","Pause audio","Back 15 seconds","Forward 15 seconds","Playback speed","Seek","Audio player","Pause audio and close player","Follow the text","{a} of {b}",{"headline":321,"subline":322},"Turn every post into a decision","See the numbers behind these notes for your own shops, with a demo on your data.",{"primaryCta":324,"secondaryCta":325},{"label":57,"to":58},{"label":326,"to":28},"See pricing",{"type":328,"value":329,"toc":330},"minimark",[],{"title":331,"searchDepth":332,"depth":332,"links":333},"",3,[],null,"Notes on profit, pricing and ads for sellers on Shopee, Lazada and TikTok Shop, with the numbers behind them.",false,"md",{},{"component":33},"\u002Fblog",{"title":33,"description":335},{"loc":340},"blog","VuN4_Knpp3IpkJqH-kRoXag4vgoPV2p71ZFiR-VpgzQ",[346,349,351,354,357,360,363],{"slug":347,"label":348},"profit","Profit",{"slug":350,"label":27},"pricing",{"slug":352,"label":353},"ads","Ads",{"slug":355,"label":356},"operations","Operations",{"slug":358,"label":359},"market","Market",{"slug":361,"label":362},"data-science","Data Science",{"slug":364,"label":48},"company",[366,373,379,383,388,393,397,402,407,412,418,423,428,433,438,444,450,456,462,469,475,480,486,491,496,501,506,511,517,523,528,534,540,546,551,557,563,569,574,578,583,588,594,601,607,613,619,625,631,637,643,649,655,661,667,673,679,685,690,695,700,705,710,716],{"title":367,"description":368,"category":352,"date":369,"readMinutes":370,"featured":336,"cover":334,"format":223,"platforms":371,"draft":336,"slug":372},"E-commerce ads optimization for Shopee, Lazada, and TikTok Shop, on profit","E-commerce ads optimization on profit: True ROAS, margin-based Target ROAS and nightly budget allocation for Shopee, Lazada and TikTok Shop ads, applied for you.","2026-10-23",4,[210,213,216],"ads-optimization",{"title":374,"description":375,"category":352,"date":369,"readMinutes":376,"featured":336,"cover":334,"format":232,"platforms":377,"draft":336,"slug":378},"What ROAS is break-even? Calculate break-even ROAS and your maximum ad spend per order","Break-even ROAS is 1 ÷ your contribution-margin ratio. It does not depend on how much you spend on ads. This calculator derives it from a dated Shopee Thailand fee rate — or one you type in — then tells you the number sellers actually need: the most baht one order can pay for ads before it stops making money.",9,[210],"break-even-roas-calculator",{"title":380,"description":381,"category":364,"date":369,"readMinutes":332,"featured":336,"cover":334,"format":229,"platforms":334,"draft":336,"slug":382},"DataGlass vs a marketing agency for marketplace profit","When an agency is the right hire, where the agency model has structural blind spots for marketplace profit, and how DataGlass complements an agency engagement.","dataglass-vs-agency",{"title":384,"description":385,"category":364,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":229,"platforms":334,"draft":336,"slug":387},"DataGlass vs an ERP for marketplace decision-making","When an ERP is the right tool, where it falls short for marketplace ad\u002Fpricing decisions, and how DataGlass complements (not replaces) the ERP for SEA sellers.",2,"dataglass-vs-erp",{"title":389,"description":390,"category":364,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":229,"platforms":391,"draft":336,"slug":392},"DataGlass vs Shopee Seller Center for profit-aware operations","Shopee Seller Center is essential for running a Shopee shop, but it does not show profit-aware metrics or cross-platform views. How DataGlass complements it.",[210],"dataglass-vs-shopee-seller-center",{"title":394,"description":395,"category":364,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":229,"platforms":334,"draft":336,"slug":396},"DataGlass vs spreadsheets for marketplace sellers","When spreadsheets are enough, where they break for marketplace sellers, and how DataGlass connects ads, COGS, fees, and inventory into one decision-ready view.","dataglass-vs-spreadsheet",{"title":398,"description":399,"category":355,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":223,"platforms":400,"draft":336,"slug":401},"Prevent stockouts and overstock across Shopee, Lazada, and TikTok Shop","Prevent Shopee, Lazada, and TikTok Shop stockouts with demand-aware forecasting, campaign-day planning, and cross-platform inventory unification.",[210,213,216],"inventory-stockout",{"title":403,"description":404,"category":355,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":226,"platforms":405,"draft":336,"slug":406},"Lazada integration for marketplace profit optimization","DataGlass integrates with Lazada to model true profit per SKU, detect ad waste, evaluate voucher impact, and forecast stockout risk.",[213],"lazada-integration",{"title":408,"description":409,"category":347,"date":369,"readMinutes":376,"featured":336,"cover":334,"format":232,"platforms":410,"draft":336,"slug":411},"Shopee raised its fees on 4 August 2026 — how much more is your shop paying?","The new rate table is everywhere. Your number is not. Enter last month’s GMV by category and this page computes the baht difference between the schedule effective 1 June 2026 and the one effective 4 August 2026 — including the extra hit for shops that were on the discounted-commission programme.",[210],"marketplace-fee-impact-calculator",{"title":413,"description":414,"category":347,"date":369,"readMinutes":415,"featured":336,"cover":334,"format":223,"platforms":416,"draft":336,"slug":417},"Optimize e-commerce profit across Shopee, Lazada, and TikTok Shop","How to optimize e-commerce profit across Shopee, Lazada and TikTok Shop: one objective, five decision variables, and the constraint that binds each.",7,[210,213,216],"optimize-ecommerce-profit",{"title":419,"description":420,"category":350,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":223,"platforms":421,"draft":336,"slug":422},"Price and promote without destroying margin","Optimize Shopee, Lazada, and TikTok Shop discounts, bundle deals, and campaign-day pricing on true margin. Project promotion impact before publishing.",[210,213,216],"pricing-promotions",{"title":424,"description":425,"category":347,"date":369,"readMinutes":370,"featured":336,"cover":334,"format":232,"platforms":426,"draft":336,"slug":427},"E-commerce profit calculator for Shopee sellers","Calculate true Shopee profit per SKU with a per-order breakdown covering commission fees, ads, COGS, and fulfillment.",[210],"profit-margin",{"title":429,"description":430,"category":355,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":226,"platforms":431,"draft":336,"slug":432},"Shopee integration for profit, ads, stock, and promotion decisions","DataGlass connects to Shopee to model true profit per SKU, detect wasted ad spend, and surface daily decisions. Live integration, no manual exports.",[210],"shopee-integration",{"title":434,"description":435,"category":355,"date":369,"readMinutes":386,"featured":336,"cover":334,"format":226,"platforms":436,"draft":336,"slug":437},"TikTok Shop integration for ads, inventory, and profit decisions","DataGlass integrates with TikTok Shop for True ROAS on video and live ads, margin-aware discount analysis, and stock prioritization before live events.",[216],"tiktok-shop-integration",{"title":439,"description":440,"category":352,"date":441,"readMinutes":415,"featured":336,"cover":334,"format":223,"platforms":442,"draft":336,"slug":443},"Shopee ROAS protection and TikTok ROI protection: when the credit actually protects profit","Shopee ROAS protection credits campaigns back to 80% of target, TikTok to 90%. Why that protects profit only above 1.25× break-even, with a worked example.","2026-10-08",[210,216],"roas-protection",{"title":445,"description":446,"category":352,"date":441,"readMinutes":447,"featured":336,"cover":334,"format":223,"platforms":448,"draft":336,"slug":449},"Shopee Ads optimization: set Target ROAS, budget and timing from profit","How to optimize Shopee Ads for profit: break-even ROAS per SKU, which lever binds, and CPC by hour from 9.11M Thai clicks. Worked THB example.",10,[210],"shopee-ads-optimization",{"title":451,"description":452,"category":347,"date":441,"readMinutes":453,"featured":336,"cover":334,"format":223,"platforms":454,"draft":336,"slug":455},"Shopee seller fees in the Philippines 2026: every fee, what it is charged on, and your break-even ROAS","Shopee Philippines seller fees 2026: commission, 2.24% transaction fee, 5.6% platform shipping, seller growth support and ₱5 order fee, worked on one order.",5,[210],"shopee-seller-fees",{"title":457,"description":458,"category":352,"date":441,"readMinutes":459,"featured":336,"cover":334,"format":223,"platforms":460,"draft":336,"slug":461},"TikTok Shop ads optimization: set GMV Max ROI from margin and affiliate mix","How to optimize TikTok Shop GMV Max for profit: break-even ROI after TikTok fees, the affiliate commission trap, and the 3% commission rule. Worked THB example.",8,[216],"tiktok-shop-ads-optimization",{"title":463,"description":464,"category":352,"date":465,"readMinutes":466,"featured":336,"cover":334,"format":226,"platforms":467,"draft":336,"slug":468},"How Shopee ads bidding works: what your Target ROAS and budget actually do to a bid","How Shopee ads bidding works, derived: GMV Max gives you two boxes — a Target ROAS and a daily budget — and every guide treats them as two levers. They are not. Both collapse into a single threshold on value-to-cost before any bid is placed, only one is binding at a time, and the other is doing nothing at all. Here is the derivation, the three regimes a campaign can be in, and the one-line test that tells you which of your two levers is switched off.","2026-09-01",13,[210],"how-shopee-ads-bidding-works",{"title":470,"description":471,"category":352,"date":465,"readMinutes":472,"featured":336,"cover":334,"format":223,"platforms":473,"draft":336,"slug":474},"Break-even ROAS is the zero line: the Target ROAS that actually maximizes Shopee profit","Setting your Target ROAS to your break-even ROAS does not protect profit — it targets exactly zero, for every possible volume. The target that maximizes profit is break-even marked up three times: by the demand your ads did not create, by the price your own expansion pays, and by the gap between your average click and your last one. On mid-market parameters those compound to 3.3×, moving the correct target on a 20%-margin product from 5.00 to 16.67.",12,[210],"profit-maximizing-target-roas",{"title":476,"description":477,"category":361,"date":465,"readMinutes":466,"featured":336,"cover":334,"format":241,"platforms":478,"draft":336,"slug":479},"Shopee ad benchmarks 2026: what a click and a thousand impressions actually cost","Across 333.9 million Shopee impressions and 9.11 million clicks on 80+ Thai shops, the price of a click varies 2.01× by hour of day while the rate at which people click varies 1.14×. CPC ranges THB 3.36 to THB 6.75, CPM ranges THB 89.16 to THB 200.15, and every behavioural input is nearly flat. The full table, the methodology, and the two ways a benchmark like this gets misused.",[210],"shopee-ads-benchmarks-cpc-cpm",{"title":481,"description":482,"category":364,"date":483,"readMinutes":484,"featured":336,"cover":334,"format":229,"platforms":334,"draft":336,"slug":485},"DataGlass is not an OMS — what the difference is, and when to run both","An OMS makes the order happen. DataGlass decides what to do next. The daily-question boundary, an explicit list of what DataGlass does not do, and the one seam: price.","2026-08-25",6,"dataglass-vs-oms",{"title":487,"description":488,"category":352,"date":483,"readMinutes":415,"featured":336,"cover":334,"format":223,"platforms":489,"draft":336,"slug":490},"How much should you spend on Shopee Ads? Derive it from margin, not from feel","Break-even ROAS is 1 ÷ contribution margin ratio. Worked THB example: a fashion SKU allows THB 71 per attributed order and roughly THB 994 a day. Budget is an output.",[210],"shopee-ads-budget",{"title":492,"description":493,"category":347,"date":483,"readMinutes":415,"featured":336,"cover":334,"format":223,"platforms":494,"draft":336,"slug":495},"Losing money on Shopee — the leaks a fee calculator cannot see","Shopee deducts 7.49–17.12% plus a 3.21% payment fee — the half a calculator sees. Six month-level leaks turn a 28.6% \"profit\" into 1.6%. Sourced THB waterfall.",[210],"shopee-losing-money",{"title":497,"description":498,"category":358,"date":483,"readMinutes":415,"featured":336,"cover":334,"format":229,"platforms":499,"draft":336,"slug":500},"Shopee vs TikTok Shop vs Lazada — what survives after fees and the ads it takes to sell","TikTok Shop's 11.77% fashion commission becomes a 23.2% take rate once the Commerce Growth and Transaction Fees stack. The crossover with Shopee sits near THB 3,720.",[210,213,216],"where-to-sell-shopee-tiktok-lazada",{"title":502,"description":503,"category":352,"date":504,"readMinutes":376,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":505},"CTR and Target ROAS under CPM vs CPC: what changes when marketplaces bill for impressions","Under cost-per-click billing, a better creative is worth nothing to your unit economics — it buys the same clicks from fewer impressions, and your cost per click does not move. Under cost-per-impression billing the same improvement cuts the price of every click by g\u002F(1+g). Five lemmas on the transition, and why click-through rate is the input most marketplace sellers are least prepared to have priced.","2026-08-18","ctr-target-roas-cpm-vs-cpc",{"title":507,"description":508,"category":352,"date":509,"readMinutes":447,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":510},"What is ROAS — and why a high ROAS doesn't always mean more sales or more profit","ROAS (Return on Ad Spend) is sales ÷ ad spend: spend THB 1, get THB 20 of sales back, and your ROAS is 20. But a high ROAS only means a high return per ad baht — not the highest total sales, and not real profit once you subtract COGS, platform fees, payment fees, vouchers, shipping, and returns. Here is what ROAS is, what it hides, and the number sellers should actually watch.","2026-06-01","what-is-roas",{"title":512,"description":513,"category":352,"date":514,"readMinutes":459,"featured":336,"cover":334,"format":223,"platforms":515,"draft":336,"slug":516},"What is Target ROAS? How to set it in Shopee GMV Max without burning profit","What is Target ROAS? It is the revenue-to-ad-spend ratio Shopee GMV Max optimises toward — and on the Shopee fee schedule effective 4 August 2026, a THB 1,000 product with THB 600 of COGS is left with 7.7% margin before ads, so a Target ROAS of 5 recovers THB 0.39 of contribution per THB 1.00 of ad spend while the campaign reports that it hit its target. What Target ROAS means, how to set it in GMV Max, and the two formulas — break-even ROAS and target-margin ROAS — that protect your margin.","2026-05-30",[210],"how-to-set-target-roas-gmv-max",{"title":518,"description":519,"category":352,"date":520,"readMinutes":484,"featured":336,"cover":334,"format":226,"platforms":521,"draft":336,"slug":522},"Platform ROAS vs true ROAS: why your ads make sales but no profit","A ROAS of 5 feels like a win — until you notice the platform computed it on gross attributed sales, before fees, vouchers, shipping, and COGS, and while still counting orders that were later returned or cancelled. The dashboard number is a revenue ratio wearing a profit costume. A primer on what platform ROAS hides and how true ROAS is calculated across Shopee, Lazada, and TikTok Shop.","2026-05-29",[210,213,216],"platform-roas-vs-true-roas",{"title":524,"description":525,"category":361,"date":520,"readMinutes":484,"featured":336,"cover":334,"format":226,"platforms":526,"draft":336,"slug":527},"What is e-commerce optimization?","Optimization is not a marketing word — it is an operations-research discipline: use data and math to compute the best decision against an explicit objective. Point it at a Shopee, Lazada, or TikTok Shop store and the objective is true profit, not sales. A primer on why those two numbers stopped moving together, and what optimization actually answers.",[210,213,216],"what-is-ecommerce-optimization",{"title":529,"description":530,"category":347,"date":531,"readMinutes":376,"featured":336,"cover":334,"format":223,"platforms":532,"draft":336,"slug":533},"How to cut Shopee seller costs: the supply-side margin audit","The median gap between COGS refreshes on a top-50 Shopee catalogue is 87 days, and over that window supplier and packaging drift alone compresses contribution margin by 1.5–3.0 percentage points. The eight supply-side cost lines that compound against Shopee margin between audits, ranked by recoverable magnitude, with the review cadence that holds each one.","2026-05-08",[210],"how-to-increase-shopee-profit-without-more-sales",{"title":535,"description":536,"category":352,"date":531,"readMinutes":537,"featured":336,"cover":334,"format":223,"platforms":538,"draft":336,"slug":539},"How to sell more on Shopee without burning margin","The standard Shopee growth advice — discount harder, scale ad budget, opt into every campaign — works against the operator at scale. A research note on the lever order that actually compounds Shopee revenue, and the metric that tells you whether the lift was earned or bought.",11,[210],"how-to-sell-more-on-shopee",{"title":541,"description":542,"category":347,"date":543,"readMinutes":484,"featured":336,"cover":334,"format":223,"platforms":544,"draft":336,"slug":545},"Increase profit on Shopee without selling more","Shopee ROAS hides commission, fees, vouchers and shipping. Optimize Shopee profit with the per-SKU break-even framework that replaces it — 280 Thai accounts.","2026-05-07",[210],"increase-shopee-profit",{"title":547,"description":548,"category":358,"date":543,"readMinutes":484,"featured":336,"cover":334,"format":223,"platforms":549,"draft":336,"slug":550},"Increase sales on Shopee in 2026","Shopee CPC is climbing — buying demand got pricier while repeat demand stayed flat. The three levers that compound revenue, plus a 90-day operator cadence.",[210],"increase-shopee-sales",{"title":552,"description":553,"category":361,"date":543,"readMinutes":554,"featured":336,"cover":334,"format":235,"platforms":555,"draft":336,"slug":556},"Mechanism-Aware Hierarchical Causal Elasticity Modeling for Platform E-Commerce","A methodological working paper proposing MA-HCEM: a hierarchical Bayesian, double-machine-learning, mechanism-aware estimator for SKU-level price elasticity on Shopee, Lazada, and TikTok Shop, with Bundle Deal and voucher mechanisms treated as structured price discrimination.",27,[210,213,216],"mechanism-aware-hierarchical-causal-elasticity-modeling-platform-ecommerce",{"title":558,"description":559,"category":350,"date":543,"readMinutes":560,"featured":336,"cover":334,"format":226,"platforms":561,"draft":336,"slug":562},"Price elasticity modeling for marketplace sellers — why Shopee, Lazada, and TikTok Shop pricing decisions need it","A practical guide from DataGlass Labs Research on the most under-used lever in Southeast Asian e-commerce: knowing how customers actually respond to price changes.",14,[210,213,216],"price-elasticity-modeling-shopee-lazada-tiktok-shop",{"title":564,"description":565,"category":361,"date":566,"readMinutes":567,"featured":336,"cover":334,"format":235,"platforms":334,"draft":336,"slug":568},"From Gut Feel to Posterior Inference: A Research Article on the DataGlass Decision-Intelligence System for E-Commerce Ad Budget Allocation","A rigorous public communication of the DataGlass decision-intelligence system for daily ad-budget allocation on platform-controlled marketplaces — the pre-algorithmic baseline, the analytical reasons it fails (the reported-vs-true ROAS gap and the rolling-mean bias), the Bayesian + bandits-with-knapsacks methodology, and the empirical 18–24% portfolio-profit lift with reduced reallocation frequency.","2026-05-04",20,"bayesian-budget-allocation-for-marketplace-ads",{"title":570,"description":571,"category":361,"date":566,"readMinutes":572,"featured":336,"cover":334,"format":238,"platforms":334,"draft":336,"slug":573},"Decision Intelligence for E-commerce: How Retailers Optimise Pricing, Demand Forecasting, Inventory, Promotions & Personalization","Pricing, forecasting, inventory, promotions, and personalization — a deep technical survey of the techniques large retailers use, the variants that matter, and how to deploy them",42,"decision-intelligence-playbook-for-e-commerce",{"title":575,"description":576,"category":361,"date":566,"readMinutes":332,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":577},"DataGlass Research Methodology","A single canonical reference for the sample frame, time window, attribution rules, and exclusions that calibrate every \"in our data\" claim across the blog and research surfaces.","methodology",{"title":579,"description":580,"category":361,"date":566,"readMinutes":581,"featured":336,"cover":334,"format":238,"platforms":334,"draft":336,"slug":582},"Prediction and Risk Optimization Under Uncertainty: A Cross-Domain Meta-Review","Mature decision systems in finance, operations, insurance, energy, healthcare, causal inference, and e-commerce share the same four primitives — a calibrated probabilistic model, a coherent risk-aware objective, an operational constraint set, and a principled exploration mechanism. Eleven worked cases ground the framework, with the DataGlass marketplace ad-budget system as the connecting tissue.",49,"prediction-and-risk-optimization-meta-review",{"title":584,"description":585,"category":355,"date":566,"readMinutes":560,"featured":336,"cover":334,"format":226,"platforms":586,"draft":336,"slug":587},"Shopee optimization in 2026 is a feedback loop, not a checklist","The popular \"10 ways to optimize your Shopee shop\" advice optimises the wrong objective. The lever that actually moves margin is the tightness of the diagnose → decide → operate loop — and most checklists skip the diagnose stage entirely.",[210],"shopee-optimization-in-2026",{"title":589,"description":590,"category":358,"date":566,"readMinutes":591,"featured":134,"cover":334,"format":241,"platforms":592,"draft":336,"slug":593},"Shopee Sellers in 2026: Southeast Asia E-commerce Market Research, GMV & Seller Economics","Market growth has resumed; seller economics have become more exacting. Southeast Asia's platform e-commerce reached US$157.6B in 2025 (up 22.8% YoY), top-three platforms now control about 98.8% of platform GMV, and content commerce accounts for ~32% of platform GMV. The 2026 question for sellers is no longer \"how big is the market?\" — it is \"who controls the decision loop?\"",22,[210],"southeast-asia-seller-market-research-2026",{"title":595,"description":596,"category":355,"date":597,"readMinutes":598,"featured":336,"cover":334,"format":223,"platforms":599,"draft":336,"slug":600},"Shopee fee structure and commission calculation in 2026: the six layers, the four-term formula","Most 'Shopee fees' explainers stop at commission and transaction fee. The four layers underneath — voucher co-funding, Free Shipping Program subsidy, Mall premium, payment-channel pass-through — are where the gap between dashboard revenue and bank deposit actually lives. With a sourced layer-by-layer breakdown, the four-term commission-per-order formula, two worked THB examples, a sensitivity table, and the reconciliation procedure.","2026-05-03",18,[210],"shopee-fee-structure-2026",{"title":602,"description":603,"category":352,"date":604,"readMinutes":537,"featured":336,"cover":334,"format":223,"platforms":605,"draft":336,"slug":606},"Cross-platform ad budget allocation for SEA marketplace sellers","Most multi-platform sellers split ad budget across Shopee, Lazada, and TikTok Shop by historical revenue share. The math says that's wrong. Optimal allocation equalises marginal ROAS, not historical share — and the gap between the two on a typical account is 4–7 percentage points of net contribution margin per quarter.","2026-04-29",[210,213,216],"cross-platform-ad-budget-allocation",{"title":608,"description":609,"category":352,"date":610,"readMinutes":415,"featured":336,"cover":334,"format":223,"platforms":611,"draft":336,"slug":612},"Why scaling Shopee ad budget grows revenue and shrinks profit","A Shopee account moving its monthly ad budget from THB 40,000 to THB 100,000 commonly grows revenue 60% and loses a third of its net profit — and not one decision in that sequence looks wrong on the dashboard. The arithmetic of the scaling trap, why platform-reported ROAS is blind to it by construction, and the per-SKU break-even bar that stops it.","2026-04-22",[210],"how-to-increase-profit-on-shopee",{"title":614,"description":615,"category":347,"date":616,"readMinutes":466,"featured":336,"cover":334,"format":223,"platforms":617,"draft":336,"slug":618},"How to increase profit on Lazada in 2026","The LazMall badge lifts conversion. It also raises commission, mandates free-shipping subsidies, and pulls the price ceiling down through the platform's own competitive-parity rules. Whether the badge pays is a per-SKU question. A research note on the LazMall economics, why Sponsored Discovery leaks more margin than Sponsored Search, and the audit that recovers 4–6 percentage points of margin in 30 days.","2026-04-15",[213],"how-to-increase-profit-on-lazada",{"title":620,"description":621,"category":352,"date":622,"readMinutes":466,"featured":336,"cover":334,"format":223,"platforms":623,"draft":336,"slug":624},"How to reduce Shopee ad waste without killing sales","On a typical Shopee account, 20–30% of ad spend runs at a structural loss the platform dashboard ranks as winning campaigns. Pausing \"underperformers\" misses the leak. A research note on the two structural defaults that cause hidden ad waste — and the audit that surfaces it without losing revenue.","2026-04-08",[210],"how-to-reduce-shopee-ad-waste",{"title":626,"description":627,"category":347,"date":628,"readMinutes":537,"featured":336,"cover":334,"format":223,"platforms":629,"draft":336,"slug":630},"How to calculate Shopee seller margin","The math, the inputs, and the program-specific traps. Why Shopee's in-dashboard income view almost always overstates margin — and how to reconstruct true contribution margin per SKU using the order-line data Shopee already exposes via Open Platform.","2026-04-05",[210],"how-to-calculate-shopee-seller-margin",{"title":632,"description":633,"category":347,"date":634,"readMinutes":537,"featured":336,"cover":334,"format":223,"platforms":635,"draft":336,"slug":636},"How to calculate Lazada seller margin","The math, the inputs, and the program-specific traps. Why Lazada's in-platform P&L summary almost always overstates margin — and how to reconstruct true contribution margin per SKU using the order-line data Lazada exposes via the Open Platform.","2026-04-01",[213],"how-to-calculate-lazada-seller-margin",{"title":638,"description":639,"category":352,"date":640,"readMinutes":560,"featured":336,"cover":334,"format":223,"platforms":641,"draft":336,"slug":642},"How to calculate true Shopee ROAS for profit","A methodology note. Shopee's in-platform ROAS is gross-revenue based and structurally biased toward overspend at scale. True ROAS is the same formula with one input substituted — and that substitution flips winners into losses on roughly half the typical Shopee catalog. With charts, three SKU profiles, sensitivity analysis, and the operating procedure that applies the substitution at production cadence.","2026-03-25",[210],"how-to-calculate-true-shopee-roas",{"title":644,"description":645,"category":352,"date":646,"readMinutes":537,"featured":336,"cover":334,"format":223,"platforms":647,"draft":336,"slug":648},"How to reduce Lazada ad waste","Lazada runs two structurally different ad products — Sponsored Search and Sponsored Discovery — with break-even economics that diverge by ~1.4–1.8× in true-ROAS terms. A single account-wide ROAS target misjudges both. A research note on the audit that surfaces the structural waste, the pruning order that recovers margin without losing revenue, and the per-placement framework that survives Pay Day and 11.11.","2026-03-18",[213],"how-to-reduce-lazada-ad-waste",{"title":650,"description":651,"category":347,"date":652,"readMinutes":466,"featured":336,"cover":334,"format":223,"platforms":653,"draft":336,"slug":654},"How to find low-margin SKUs on Shopee","On a typical Shopee account, the top-10 SKUs by revenue and the top-10 by contribution profit overlap by roughly 50%. Half of every shop's bestsellers are not the most profitable products. A research note on the audit that surfaces the gap, the patterns hiding inside it, and the per-SKU operating decisions that recover margin.","2026-03-12",[210],"how-to-find-low-margin-skus-on-shopee",{"title":656,"description":657,"category":350,"date":658,"readMinutes":560,"featured":336,"cover":334,"format":223,"platforms":659,"draft":336,"slug":660},"How to plan Lazada campaigns around profit","Pay Day, Mega Sale, 11.11, and the long tail of category windows make participation feel mandatory. The platform documents the eligibility tiers; the seller absorbs them. A research note on what each campaign actually costs, when participation pays, and the margin-first rubric that survives all three.","2026-03-04",[213],"how-to-plan-lazada-campaigns-around-profit",{"title":662,"description":663,"category":347,"date":664,"readMinutes":472,"featured":336,"cover":334,"format":223,"platforms":665,"draft":336,"slug":666},"How to increase profit on TikTok Shop in 2026","TikTok Shop is the only SEA marketplace with a stacked second commission — affiliate commission (10–25% via the Open Affiliate Plan) layered on top of platform commission. A 6.0 platform ROAS routinely becomes ~1.4 true ROAS once the full four-line cost stack is subtracted. A research note on the affiliate-stack arithmetic, live-stream pricing discipline, and the per-SKU framework that recovers margin without retreating from the platform.","2026-02-26",[216],"how-to-increase-profit-on-tiktok-shop",{"title":668,"description":669,"category":358,"date":670,"readMinutes":671,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":672},"The SEA marketplace in 2026: a fragmented arena with one survival rule","Eighteen distinct platform-market cells, six countries, three platforms, and roughly a million sellers competing on a saturated product surface. The aggregate is growing; the per-cell margin distribution is compressing. A research note on the structural shifts and the operating rule that survives them.","2026-02-22",16,"sea-marketplace-2026",{"title":674,"description":675,"category":361,"date":676,"readMinutes":537,"featured":336,"cover":334,"format":226,"platforms":677,"draft":336,"slug":678},"Data ingestion for Shopee sellers: why zero-setup analytics matters","Most Shopee sellers don't have a strategy problem first. They have a data plumbing problem — orders, ads, COGS, fees, vouchers, inventory, pricing, and returns live in seven different surfaces, and by the time the seller has stitched them together the campaign is over. A research note on the data-source matrix, the canonical-entity model, and the zero-setup architecture that recovers ~10 hours per week.","2026-02-18",[210],"data-ingestion-for-shopee-sellers",{"title":680,"description":681,"category":358,"date":682,"readMinutes":415,"featured":336,"cover":334,"format":226,"platforms":683,"draft":336,"slug":684},"The race to zero margin — and how Shopee, Lazada, and TikTok Shop got there","Read the platform documentation end-to-end and the conclusion is uncomfortable: the discounts buyers see, the free shipping that drives conversion, and the affiliate spend that drives reach are all paid by the seller. The race-to-zero is the equilibrium output of that design.","2026-02-08",[210,213,216],"race-to-zero-margin",{"title":686,"description":687,"category":361,"date":688,"readMinutes":560,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":689},"ML demand forecasting for e-commerce sellers","Machine learning in e-commerce gets discussed in vague terms; for marketplace sellers the operating question is concrete — how many units of this SKU will sell in the next N days, with what confidence, and what decision flows from the answer? A research note on the practical model architecture, the stockout-distortion problem, sensitivity analysis, and the operating decisions forecasts feed.","2026-02-04","ml-demand-forecasting-for-ecommerce",{"title":691,"description":692,"category":358,"date":693,"readMinutes":484,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":694},"Complexity is the new tax on small sellers","A multi-shop seller in 2026 logs into seven platforms, reconciles four fee schedules, exports six CSVs, and re-enters COGS by hand. Big brands absorb the cost with a data team. Small sellers absorb it with their evenings — and it's the single biggest reason multi-shop operators stall before they reach scale.","2026-01-25","complexity-is-the-tax",{"title":696,"description":697,"category":355,"date":698,"readMinutes":472,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":699},"Stockout math for e-commerce sellers","A stockout is not one cost; it is five compounding costs. Lost contribution profit on the missed unit, plus wasted ad spend during the stockout window, plus algorithmic ranking demotion, plus repeat-buyer trust erosion, plus distorted forecasting that increases the likelihood of the next stockout. A research note on the multi-line stockout cost function, the per-SKU reorder-point math that accounts for it, and the campaign-aware adjustment that survives Pay Day and 11.11.","2026-01-18","inventory-stockout-math-ecommerce",{"title":701,"description":702,"category":350,"date":703,"readMinutes":537,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":704},"Dynamic pricing for marketplace sellers","Discounting is easy. Profitable pricing is hard. A 30% volume lift on a 10% price cut routinely lowers total contribution profit — the math says volume must lift by ~33% just to break even, and most SKUs underperform that bar. A research note on the price-elasticity arithmetic, the inventory × demand four-quadrant framework, and the per-SKU pricing decision that survives campaign-window pressure.","2026-01-04","dynamic-pricing-ecommerce-marketplaces",{"title":706,"description":707,"category":361,"date":708,"readMinutes":472,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":709},"E-commerce Decision Engine: How Marketplace Sellers Turn Data Into Profit Recommendations","A dashboard tells you what happened. A decision engine tells you what to do next, ranks the options by projected profit lift, and surfaces the math behind every recommendation. A research note on the five-layer architecture that separates the two, why marketplace commerce now requires the latter, and where the operating model breaks.","2025-12-12","how-ecommerce-decision-engines-work",{"title":711,"description":712,"category":355,"date":713,"readMinutes":459,"featured":336,"cover":334,"format":226,"platforms":714,"draft":336,"slug":715},"Multi-shop analytics for Shopee, Lazada, and TikTok Shop sellers","Modern marketplace sellers rarely operate on one channel. They have multiple dashboards but not one operating view. They know sales by channel — but not profit by channel.","2025-11-26",[210,213,216],"multi-shop-analytics-lazada-shopee-tiktok",{"title":717,"description":718,"category":358,"date":719,"readMinutes":720,"featured":336,"cover":334,"format":226,"platforms":334,"draft":336,"slug":721},"The new competitive world of commerce","A research note on what changed in marketplace commerce between 2022 and 2026. Platforms became active algorithmic counterparties; sellers who treat them as passive listing markets are competing against optimization systems they cannot see. The fix is to run your own optimization layer on top.","2025-11-12",15,"new-competitive-world-of-commerce",{"entry":723,"document":770,"toc":1113},{"id":724,"title":686,"category":361,"cover":334,"cta":725,"date":688,"description":687,"draft":336,"draftReason":334,"extension":337,"featured":336,"format":226,"listen":334,"math":336,"meta":728,"navigation":334,"ogImage":334,"path":729,"platforms":334,"readMinutes":560,"related":730,"robots":334,"schemaOrg":334,"seo":731,"sitemap":734,"sources":735,"stem":760,"tldr":761,"topic":768,"updatedAt":334,"__hash__":769},"blog_en\u002Fblog\u002Fml-demand-forecasting-for-ecommerce.md",{"title":726,"body":727},"Forecast demand before stockouts cost you sales.","DataGlass connects demand, inventory, ads, and margin so sellers can act before problems become expensive.",{},"\u002Fblog\u002Fml-demand-forecasting-for-ecommerce",[699,709,678],{"title":732,"description":733},"ML Demand Forecasting for E-commerce Sellers | DataGlass","Learn how machine learning demand forecasting helps marketplace sellers reduce stockouts, plan ads, manage inventory, and protect profit.",{"loc":729},[736,740,744,748,752,756],{"title":737,"url":738,"note":739},"ScienceDirect — Machine learning approaches in inventory control: a systematic review (2025)","https:\u002F\u002Fwww.sciencedirect.com\u002Fscience\u002Farticle\u002Fpii\u002FS2214716025000430","122 reviewed articles categorising ML applications across demand forecasting, joint optimisation, and reinforcement-learning inventory policies. The structural literature for the model architecture described in this note.",{"title":741,"url":742,"note":743},"arXiv — Dynamic Retail Pricing via Q-Learning","https:\u002F\u002Farxiv.org\u002Fabs\u002F2411.18261","Research on reinforcement-learning approaches to retail pricing and demand response — relevant context for the broader ML-meets-marketplace literature and the joint-optimisation layer downstream of forecasting.",{"title":745,"url":746,"note":747},"M5 Forecasting Competition — methodology and results","https:\u002F\u002Fwww.kaggle.com\u002Fc\u002Fm5-forecasting-accuracy","M5 hierarchical retail forecasting competition. Reference for the gradient-boosted-tree baseline and the typical accuracy ceiling on hierarchical retail demand series.",{"title":749,"url":750,"note":751},"Google, Temasek & Bain — e-Conomy SEA 2025","https:\u002F\u002Fwww.temasek.com.sg\u002Fen\u002Fnews-and-resources\u002Fnews-room\u002Fnews\u002F2025\u002Fe-conomy-sea-2025-report-aseans-digital-economy-poised-to-surpass-300-billion","Volatility context: SEA e-commerce GMV growth and the video-commerce demand spikes (live-stream sessions, creator-driven volume) that forecasting on a Shopee\u002FLazada\u002FTikTok Shop catalog has to handle.",{"title":753,"url":754,"note":755},"Sea Limited — Investor Relations","https:\u002F\u002Fwww.sea.com\u002Finvestor\u002Fhome","Sea Limited 4Q25 \u002F 1Q26 disclosures documenting the AI investment that drives platform-side demand-shaping (ad auctions, recommendation systems) the forecasting model has to account for.",{"title":757,"url":758,"note":759},"Shopee — Ranking and stockout policy documentation","https:\u002F\u002Fhelp.shopee.co.th\u002Fportal\u002Farticle\u002F77792","Shopee Help Center documentation referenced for the stockout-distortion and ranking-demotion mechanisms — the structural reason stockouts compound forecasting errors.","blog\u002Fml-demand-forecasting-for-ecommerce",[762,763,764,765,766,767],"Thesis: useful ML demand forecasting for marketplace sellers is decision infrastructure, not prediction theatre. The output is not \"expected units sold\"; it is \"reorder now \u002F scale ads \u002F cap ad spend \u002F refuse this campaign window,\" with confidence attached.","A simple additive model — base demand + seasonality + promotion lift + ad effect + price effect + channel effect − stock constraint — captures most of what matters. The hard part is the stock-constraint term, not the ML.","In our data, forecasts that censor stockout-period demand and incorporate ad-pressure and campaign-window features beat naive moving-average baselines by 20–40% on stockout-prevention recall, depending on category.","Stockout-period sales data is censored, not zero. Treating it as zero produces a self-reinforcing under-forecast that the model \"learns\" until the SKU is structurally under-ordered.","Forecasts feed three downstream decisions: reorder timing (with confidence-tiered logic), ad-budget allocation (scale only when forecast supports the demand and inventory supports the supply), and campaign-window participation (decline windows where forecast volume cannot be served from current stock).","Limitation: the framework assumes ≥6 months of clean order-line history per SKU. Below that, simpler heuristics (moving average, supplier-lead-time-based reorder) outperform the ML model on most categories.","models","UjiQM5mlAefmWXXQ_GBQimaTy_ly8OGO17smc0SK6a4",{"frontmatter":771,"meta":783,"nodes":815},{"title":686,"description":687,"seo":772,"category":361,"format":226,"topic":768,"date":688,"readMinutes":560,"tldr":773,"related":774,"sources":775,"cta":782},{"title":732,"description":733},[762,763,764,765,766,767],[699,709,678],[776,777,778,779,780,781],{"title":737,"url":738,"note":739},{"title":741,"url":742,"note":743},{"title":745,"url":746,"note":747},{"title":749,"url":750,"note":751},{"title":753,"url":754,"note":755},{"title":757,"url":758,"note":759},{"title":726,"body":727},{"toc":784},{"title":331,"searchDepth":386,"depth":386,"links":785},[786,789,792,795,798,801,804,807,810,813],{"id":787,"depth":386,"text":788},"why-marketplace-demand-is-hard-to-forecast","Why marketplace demand is hard to forecast",{"id":790,"depth":386,"text":791},"a-basic-demand-forecasting-formula","A basic demand forecasting formula",{"id":793,"depth":386,"text":794},"the-stockout-distortion-problem","The stockout distortion problem",{"id":796,"depth":386,"text":797},"how-ml-can-help","How ML can help",{"id":799,"depth":386,"text":800},"forecasting-without-decisions-is-not-enough","Forecasting without decisions is not enough",{"id":802,"depth":386,"text":803},"the-ad-and-inventory-connection","The ad and inventory connection",{"id":805,"depth":386,"text":806},"forecasting-confidence-matters","Forecasting confidence matters",{"id":808,"depth":386,"text":809},"sensitivity-what-changes-the-operating-decision","Sensitivity — what changes the operating decision",{"id":811,"depth":386,"text":812},"limitations-and-where-this-argument-breaks","Limitations and where this argument breaks",{"id":577,"depth":386,"text":814},"Methodology",[816,820,824,827,830,833,836,838,841,851,854,856,859,864,867,869,877,903,906,909,911,914,916,919,923,925,928,933,936,941,943,946,952,1048,1053,1055,1058,1077,1079,1097,1106,1109],[817,818,819],"p",{},"Machine learning in e-commerce often gets described in vague terms. For sellers, the most useful ML question is much more concrete:",[821,822],"blog-pullquote",{"text":823},"How much will this SKU sell soon?",[817,825,826],{},"That question feeds into every other operational decision a marketplace seller makes — inventory, ads, pricing, promotions, cashflow. Underestimate demand and the SKU stocks out at exactly the worst time. Overestimate demand and cash gets trapped in inventory that turns slowly. Misunderstand the shape of demand and ad budget pours into products that cannot support the extra sales. Demand forecasting is not really prediction for its own sake. It is decision infrastructure.",[828,829,788],"h2",{"id":787},[817,831,832],{},"Marketplace demand is volatile in ways that classical retail forecasting was never designed for. A Shopee seller can watch demand swing because of payday campaigns, double-day campaigns, vouchers, competitor pricing, their own ad spend, keyword ranking changes, creator content, reviews, stock availability, seasonality, weather, shipping-speed changes, or even small product-page edits — sometimes several of these at once.",[817,834,835],{},"A simple historical average does not survive contact with that environment. A product that sold 100 units last month will not necessarily sell 100 units this month. Maybe it was out of stock for a week. Maybe ad spend was higher. Maybe a campaign artificially boosted demand. Maybe a competitor dropped price. Maybe a review issue hurt conversion. The seller who treats last month's number as next month's plan is fitting a straight line to a curve.",[828,837,791],{"id":790},[817,839,840],{},"Even a simplified additive model captures most of what matters:",[842,843,846],"pre",{"language":844,"filename":845},"text","Expected demand",[847,848,850],"code",{"class":849},"language-text","Expected demand = base demand\n                + seasonality\n                + promotion lift\n                + ad effect\n                + price effect\n                + channel effect\n                − stock constraint",[817,852,853],{},"Each component does specific work. Base demand captures the normal pattern. Seasonality captures recurring spikes. Promotion lift captures campaign behavior. Ad effect captures paid visibility. Price effect captures elasticity. The stock constraint prevents the model from misreading a stockout as a drop in real demand — which sounds obvious until you see what happens when it is missing.",[828,855,794],{"id":793},[817,857,858],{},"Stockouts corrupt historical data. If a product was out of stock for ten days, sales data will show lower units sold for that period — but that does not mean demand disappeared. It means demand could not be served.",[860,861,863],"blog-callout",{"title":862},"Naive vs smart forecasts","A naive forecast concludes \"sales dropped, so demand is lower.\" A smarter forecast asks: did sales drop because buyers did not want the product, or because the seller had no stock?",[817,865,866],{},"The reason this matters in practice is the feedback loop. Underestimate demand after a stockout, reorder a smaller batch, stock out again sooner — and the model \"learns\" to expect lower and lower demand from a SKU whose real customers are still there.",[828,868,797],{"id":796},[817,870,871,872,876],{},"Machine learning can learn patterns across many signals at once — historical sales, campaign dates, ad spend, price changes, inventory availability, product category, day of week, seasonality, order velocity, and channel behavior — without the seller having to choose which one matters in advance. A 2025 systematic review of machine learning in inventory control (ScienceDirect) analyzed 122 articles and categorised ML applications into demand forecasting before optimisation, ML embedded directly into optimisation, and dynamic approaches such as reinforcement learning for inventory policies ",[873,874],"blog-cite",{"n":875},"1",":blog-cite.",[878,879],"blog-chart",{"kind":880,"title":881,"labels":882,"values":888,"notes":893,"unit":899,"max":900,"highlight":901,"caption":902},"bar","Forecast accuracy by method — typical Shopee SKU, 28-day horizon (sMAPE, lower is better)",[883,884,885,886,887],"Naive moving average (last 28 days)","Seasonal naive (same window last year)","Exponential smoothing + holiday calendar","Gradient-boosted trees + ad \u002F price \u002F stock features","Same model with stockout-period censoring",[889,890,891,892,466],38,31,24,17,[894,895,896,897,898],"Baseline; no censoring of stockout periods","Captures recurring spikes","Adds campaign-window dummies","Feature-engineered; production baseline","The single highest-leverage data fix","%","45",[332,370],"sMAPE = symmetric mean absolute percentage error. The single highest-leverage step — censoring stockout-period sales rather than treating them as observed demand — produces a larger accuracy gain than the model-architecture upgrade. Result: in production, data quality dominates model choice on hierarchical retail series.",[817,904,905],{},"For sellers, the practical takeaway is not \"use ML because ML is trendy.\" The takeaway is more pointed:",[821,907],{"text":908},"Forecasting should connect to decisions.",[828,910,800],{"id":799},[817,912,913],{},"A forecast that says \"you may sell 500 units next month\" is interesting but operationally inert. The seller needs to know what to do — should I reorder now, should I increase ads, should I reduce ads because stock is running low, should I raise price to slow demand, should I avoid a campaign because inventory cannot support it, should I bundle this SKU with something else? A forecast that does not answer at least one of those questions is doing analysis instead of work. This is why DataGlass connects forecasting directly to action.",[828,915,803],{"id":802},[817,917,918],{},"Ads and inventory should not be managed in separate tabs. A SKU with strong margin and enough inventory can absorb more ad budget. A SKU with strong demand but low stock will be hurt by more ad budget — every additional click lands on a page that is about to disappoint a buyer. A SKU with excess inventory but weak margin can sometimes benefit from a discount, but only if the discount is capped by contribution margin instead of by clearance instinct.",[860,920,922],{"title":921},"A useful decision rule","Scale ads only when margin, conversion, and inventory are all healthy.",[828,924,806],{"id":805},[817,926,927],{},"A forecast should not pretend to be perfect. A good system shows confidence alongside the prediction, because the right reorder behavior is different at high, medium, and low confidence:",[842,929,930],{"language":844},[847,931,932],{"class":849},"High confidence: reorder now\nMedium confidence: monitor and prepare supplier\nLow confidence: avoid aggressive inventory bet",[817,934,935],{},"This is especially important for new products and for products that have been moved by viral content, where the historical signal is short, noisy, or both.",[937,938,940],"blog-question",{"question":939},"How accurate is ML demand forecasting for marketplace sellers in practice?","On a typical Shopee SKU with 6+ months of clean order-line history, a feature-engineered gradient-boosted-tree model with stockout-period censoring achieves ~13% sMAPE on the 28-day horizon — vs. ~38% sMAPE for a naive moving-average baseline. The single highest-leverage step is censoring stockout-period sales (treating them as missing rather than observed); the model-architecture upgrade contributes a smaller share of the accuracy gain. New SKUs (under 6 months of history) and SKUs moved by viral content have meaningfully wider confidence intervals, and the operating procedure should reflect that — high-confidence forecasts feed automated reorder, low-confidence forecasts feed human review.",[828,942,809],{"id":808},[817,944,945],{},"A forecasting system in production exists to feed decisions, and the table below stress-tests how three operating decisions — reorder, ad scale, campaign participation — shift under different forecast-confidence regimes.",[817,947,948],{},[949,950,951],"em",{},"Forecast confidence × operating decision",[953,954,955,974],"table",{},[956,957,958],"thead",{},[959,960,961,965,968,971],"tr",{},[962,963,964],"th",{},"Decision",[962,966,967],{},"High confidence",[962,969,970],{},"Medium confidence",[962,972,973],{},"Low confidence",[975,976,977,992,1006,1020,1034],"tbody",{},[959,978,979,983,986,989],{},[980,981,982],"td",{},"Reorder timing",[980,984,985],{},"Auto-reorder at lead-time + safety buffer",[980,987,988],{},"Manual review; supplier on standby",[980,990,991],{},"Avoid aggressive inventory commitment",[959,993,994,997,1000,1003],{},[980,995,996],{},"Ad budget allocation",[980,998,999],{},"Scale to forecast-supported demand level",[980,1001,1002],{},"Hold current spend; monitor weekly",[980,1004,1005],{},"Cap spend to current run-rate",[959,1007,1008,1011,1014,1017],{},[980,1009,1010],{},"Campaign-window participation",[980,1012,1013],{},"Participate at full eligibility tier",[980,1015,1016],{},"Participate at lower voucher tier",[980,1018,1019],{},"Decline; protect inventory for baseline demand",[959,1021,1022,1025,1028,1031],{},[980,1023,1024],{},"Pricing change",[980,1026,1027],{},"Implement price test with monitoring",[980,1029,1030],{},"Defer until forecast stabilises",[980,1032,1033],{},"Hold current pricing",[959,1035,1036,1039,1042,1045],{},[980,1037,1038],{},"Stockout-risk alert",[980,1040,1041],{},"Alert at 14 days of supply",[980,1043,1044],{},"Alert at 21 days of supply",[980,1046,1047],{},"Alert at 28+ days; treat as soft signal",[817,1049,1050],{},[949,1051,1052],{},"The matrix is the integration layer between the forecast and the operating decisions. A forecast without confidence-tiered decision rules is analysis without action; a confidence-tiered system produces auto-handled cases at the high-confidence end and human-review cases at the low-confidence end, concentrating operator attention where it matters.",[828,1054,812],{"id":811},[817,1056,1057],{},"Five explicit limits.",[1059,1060,1061,1065,1068,1071,1074],"ul",{},[1062,1063,1064],"li",{},"History-length lower bound. The framework assumes ≥6 months of clean order-line history per SKU. Below that, simpler heuristics (moving average, supplier-lead-time-based reorder, category-mean inference) outperform the ML model. New-product launches need a different operating procedure: human-set reorder cadence with auto-graduation to the model once history accumulates.",[1062,1066,1067],{},"Viral-content distortion. SKUs moved by creator content or live-stream sessions exhibit demand distributions that classical models handle poorly — bursty, short-window, high-variance. Wider confidence intervals and human-in-the-loop reorder are appropriate; production code should detect and flag these cases rather than auto-decide.",[1062,1069,1070],{},"Cross-platform demand interaction. The framework treats per-platform demand independently. Real demand sometimes shifts between Shopee, Lazada, and TikTok Shop on the same SKU as price-mirror automation closes arbitrage. The cross-platform interaction term is non-trivial to model and is a known underestimation source.",[1062,1072,1073],{},"Censoring quality. The stockout-period censoring step is only as good as the inventory-state input. If inventory-state data is noisy (delayed updates, multi-warehouse SKUs), censoring incorrectly applied creates either over-forecasting (if too aggressive) or under-forecasting (if too conservative).",[1062,1075,1076],{},"Internal-data scope. The accuracy figures (~13% sMAPE production baseline, the chart's comparative numbers) are aggregated across the SEA-6 Thai Shopee accounts we model directly. They are not population claims about all e-commerce demand-forecasting setups; they explicitly exclude the bottom of the size distribution noted above.",[828,1078,814],{"id":577},[817,1080,1081,1082,1084,1085,1088,1089,1092,1093,1096],{},"Public-data citations are taken from the 2025 ScienceDirect systematic review of ML approaches in inventory control ",[873,1083],{"n":875},", the M5 Forecasting Competition methodology and results (Kaggle) ",[873,1086],{"n":1087},"3",", the Bain e-Conomy SEA 2025 commentary on regional demand volatility drivers ",[873,1090],{"n":1091},"4",":blog-cite, and Shopee's Help Center documentation on stockout policy and ranking-algorithm response to inventory state ",[873,1094],{"n":1095},"6",".",[817,1098,1099,1100,1105],{},"Internal-data claims — the sMAPE figures across forecasting methods, the 20–40% stockout-prevention recall lift, the typical operating-decision split under confidence-tiered rules — are aggregated across approximately 400 active marketplace seller accounts across the ",[1101,1102,1104],"a",{"href":1103},"\u002Fph\u002Fblog\u002Fmethodology","DataGlass research methodology"," sample frame (Jan 2024 – Apr 2026, 28-month observation window), with at least 6 months of clean order-line history per SKU included in the forecasting evaluation set. Forecasts are evaluated on rolling out-of-sample 28-day windows; accuracy figures are reported as sample medians across SKUs.",[821,1107],{"text":1108},"The seller does not need to see the model. The seller needs to see the decision.",[860,1110,1112],{"title":1111},"Forecasting is decision infrastructure, not prediction","Useful ML demand forecasting for marketplace sellers is not better prediction; it is decision infrastructure — confidence-tiered reorder, ad-scale, and campaign-participation rules feeding off a censored, feature-engineered model that treats stockout periods as missing data rather than observed demand.",[1114,1115,1116,1117,1118,1119,1120,1121,1122,1123],{"id":787,"depth":386,"text":788},{"id":790,"depth":386,"text":791},{"id":793,"depth":386,"text":794},{"id":796,"depth":386,"text":797},{"id":799,"depth":386,"text":800},{"id":802,"depth":386,"text":803},{"id":805,"depth":386,"text":806},{"id":808,"depth":386,"text":809},{"id":811,"depth":386,"text":812},{"id":577,"depth":386,"text":814},{"id":1125,"categories":1126,"domain":1133,"extension":5,"fallback":1134,"home":1137,"industry":1146,"meta":1151,"pricing":1152,"stem":1164,"__hash__":1165},"og_en\u002Fen\u002Fog.yml",[1127,1128,1129,1130,1131,1132],{"slug":347,"label":348},{"slug":350,"label":27},{"slug":352,"label":353},{"slug":355,"label":356},{"slug":361,"label":362},{"slug":364,"label":48},"dataglasslabs.com",{"title":1135,"description":1136},"DataGlass","The decision engine for Shopee, Lazada and TikTok Shop sellers.",{"chart":1138,"metrics":1139},"Net profit, last 30 days",[1140,1143],{"label":1141,"value":1142},"Net profit","฿1.24M",{"label":1144,"value":1145},"Profit ROAS","3.4x",{"axis":1147,"point":1148},"Ad spend",{"label":1149,"value":1150},"Next baht still pays","+฿0.42",{},{"plans":1153},[1154,1158,1161],{"name":1155,"price":1156,"period":1157},"Free","฿0","\u002Fmonth",{"name":1159,"price":1160,"period":1157},"Standard","฿1,499",{"name":1162,"price":1163,"period":1157},"Pro","฿2,499","en\u002Fog","63L8y4Lv8I2q3NvR26fMXU1sE9Wd7TZaz2l_tV6oZuo",[],1792785020259]