{"id":8088,"date":"2026-10-26T09:00:00","date_gmt":"2026-10-26T13:00:00","guid":{"rendered":"https:\/\/aventurecloud.net\/?p=8088"},"modified":"2026-11-06T13:36:47","modified_gmt":"2026-11-06T18:36:47","slug":"your-rag-tweak-might-be-noise-what-a-one-change-at-a-time-test-showed-us","status":"publish","type":"post","link":"https:\/\/aventurecloud.net\/?p=8088","title":{"rendered":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us"},"content":{"rendered":"<p><!-- ac-update-1793939900 --><\/p>\n<p>Ask ten people how to improve a retrieval-augmented generation (RAG) system and you&#8217;ll get ten lists: better chunking, hybrid search, reranking, query rewriting, HyDE, compression. The lists are fine. What&#8217;s usually missing is proof that any single item helped <em>your<\/em> system.<\/p>\n<p>So we built a reference build to practise the boring, important part: the measurement.<\/p>\n<h2>What the project set out to test<\/h2>\n<p>The RAG optimization reference build is a tuning harness. It runs a RAG pipeline (chunker, retriever, optional reranker, optional compressor), then changes exactly one component at a time from a fixed baseline. Each change is scored on a gold set for recall, precision, keyword hits and token cost. Every change also gets a 95% bootstrap interval, so you can see whether a gain is bigger than the noise.<\/p>\n<p>Important context: everything is synthetic. The corpus is 30 fictional documents, the gold set is 15 questions, the &#8220;dense&#8221; embeddings are hashed bags of words, and the reranker is simulated. The numbers show what the harness measures, not how a real embedding model would score.<\/p>\n<h2>What we observed<\/h2>\n<p><strong>1. Hybrid search plus reranking was a clear win.<\/strong> Against the baseline (fixed 200-token chunks with dense retrieval), hybrid retrieval plus a reranker lifted recall@5 from 0.633 to 0.933. Keyword hit rate jumped from 0.233 to 0.833. The cost? Average context grew from about 190 to about 211 tokens, a small price for that gain.<\/p>\n<p><strong>2. Several popular tricks couldn&#8217;t be told apart from zero.<\/strong> Paragraph chunking and sliding-window chunking each showed +0.100 recall, but their intervals ran from -0.133 to +0.333, which includes zero. HyDE showed +0.033 with an interval that also included zero. Synonym-based query rewriting changed nothing at all. And the chunking changes nearly doubled context size, to roughly 330 tokens. On a 15-question gold set, a +0.10 bump can simply be luck.<\/p>\n<p><strong>3. The simplest method won, and that&#8217;s a warning, not a victory.<\/strong> Plain BM25 keyword search hit a perfect 1.000 recall@5. That&#8217;s not because keyword search beats everything. The synthetic embeddings are lexical, so they favour word-matching methods. With a real embedding model the ranking could change. The honest move is to say so rather than declare a universal winner.<\/p>\n<p><strong>4. The harness itself had a bug.<\/strong> Early on, the A\/B test scored retrieval <em>before<\/em> reranking, so the reranker was never actually measured and always looked useless. Once metrics were taken on the final ranking, reranking on dense retrieval alone showed +0.300 recall. If your evaluation measures the wrong stage, you&#8217;ll throw away good ideas.<\/p>\n<h2>Why this matters for a business<\/h2>\n<p>Every RAG change costs something: engineering time, extra tokens, extra latency, another moving part to maintain. If you can&#8217;t show a change helped beyond noise, you&#8217;re paying for a guess. A small, honest gold set built with your domain experts, plus one-change-at-a-time testing, turns RAG tuning from opinion into evidence. It also shows token cost next to quality, so a &#8220;win&#8221; that doubles your context bill doesn&#8217;t slip through.<\/p>\n<h2>Takeaways<\/h2>\n<ul>\n<li>Change one thing at a time from a fixed baseline.<\/li>\n<li>Report an interval, not just a number; small gold sets are noisy.<\/li>\n<li>Put token cost beside every quality metric.<\/li>\n<li>Check that you&#8217;re measuring the final output, not an earlier stage.<\/li>\n<li>Treat surprising winners with suspicion until real data confirms them.<\/li>\n<li>Grow your gold set from real questions your experts care about.<\/li>\n<\/ul>\n<h2>Reference build<\/h2>\n<p>Explore the harness, its limits and its test suite: AventureCloud RAG optimization. It&#8217;s a learning reference build on synthetic data, not client work.<\/p>\n<h2>Learn it or get it built<\/h2>\n<p>Want your team to evaluate RAG like this? Email <strong>trainings@aventurecloud.net<\/strong> about our Agentic AI trainings. Need trained AI engineers to tune retrieval on your own documents? Reach us at <strong>partnerships@aventurecloud.net<\/strong>.<\/p>\n<hr \/>\n<p><strong>Hashtags:<\/strong> #RAG #GenAI #AIEngineering<\/p>\n<p><strong>LinkedIn teaser:<\/strong><br \/>\nWe tested popular RAG tweaks one at a time. Hybrid search plus reranking lifted recall@5 from 0.633 to 0.933, three tricks couldn&#8217;t be told apart from noise, and our own test harness had a bug.<br \/>\nHere&#8217;s what a 15-question gold set taught us about measuring before you tune.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A small RAG tuning experiment found that some popular tricks did nothing measurable, one did a lot, and the test harness itself had a bug.<\/p>\n","protected":false},"author":19,"featured_media":8123,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[20],"tags":[],"class_list":["post-8088","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-ai-research"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v28.8 - https:\/\/yoast.com\/product\/yoast-seo-wordpress\/ -->\n<title>Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/aventurecloud.net\/?p=8088\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities\" \/>\n<meta property=\"og:description\" content=\"A small RAG tuning experiment found that some popular tricks did nothing measurable, one did a lot, and the test harness itself had a bug.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/aventurecloud.net\/?p=8088\" \/>\n<meta property=\"og:site_name\" content=\"Infinite Possibilities\" \/>\n<meta property=\"article:published_time\" content=\"2026-10-26T13:00:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-11-06T18:36:47+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1400\" \/>\n\t<meta property=\"og:image:height\" content=\"1294\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Aventure Blog\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Aventure Blog\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\\\/\\\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#article\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088\"},\"author\":{\"name\":\"Aventure Blog\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#\\\/schema\\\/person\\\/5bb333ca4287c7b8bcd6b4f668d02dba\"},\"headline\":\"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us\",\"datePublished\":\"2026-10-26T13:00:00+00:00\",\"dateModified\":\"2026-11-06T18:36:47+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088\"},\"wordCount\":679,\"commentCount\":0,\"publisher\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#organization\"},\"image\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/8088_rag.jpg\",\"articleSection\":[\"AI Research\"],\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"CommentAction\",\"name\":\"Comment\",\"target\":[\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#respond\"]}]},{\"@type\":\"WebPage\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088\",\"url\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088\",\"name\":\"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities\",\"isPartOf\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#primaryimage\"},\"image\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#primaryimage\"},\"thumbnailUrl\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/8088_rag.jpg\",\"datePublished\":\"2026-10-26T13:00:00+00:00\",\"dateModified\":\"2026-11-06T18:36:47+00:00\",\"breadcrumb\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\\\/\\\/aventurecloud.net\\\/?p=8088\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#primaryimage\",\"url\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/8088_rag.jpg\",\"contentUrl\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2026\\\/11\\\/8088_rag.jpg\",\"width\":1400,\"height\":1294,\"caption\":\"Open books and library shelves representing document retrieval\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/?p=8088#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\\\/\\\/aventurecloud.net\\\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#website\",\"url\":\"https:\\\/\\\/aventurecloud.net\\\/\",\"name\":\"Infinite Possibilities\",\"description\":\"\",\"publisher\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\\\/\\\/aventurecloud.net\\\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#organization\",\"name\":\"Infinite Possibilities\",\"url\":\"https:\\\/\\\/aventurecloud.net\\\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#\\\/schema\\\/logo\\\/image\\\/\",\"url\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/cropped-Logo2-1-3.png\",\"contentUrl\":\"https:\\\/\\\/aventurecloud.net\\\/wp-content\\\/uploads\\\/2024\\\/06\\\/cropped-Logo2-1-3.png\",\"width\":1560,\"height\":758,\"caption\":\"Infinite Possibilities\"},\"image\":{\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#\\\/schema\\\/logo\\\/image\\\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\\\/\\\/aventurecloud.net\\\/#\\\/schema\\\/person\\\/5bb333ca4287c7b8bcd6b4f668d02dba\",\"name\":\"Aventure Blog\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g\",\"url\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g\",\"contentUrl\":\"https:\\\/\\\/secure.gravatar.com\\\/avatar\\\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g\",\"caption\":\"Aventure Blog\"},\"url\":\"https:\\\/\\\/aventurecloud.net\\\/?author=19\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/aventurecloud.net\/?p=8088","og_locale":"en_US","og_type":"article","og_title":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities","og_description":"A small RAG tuning experiment found that some popular tricks did nothing measurable, one did a lot, and the test harness itself had a bug.","og_url":"https:\/\/aventurecloud.net\/?p=8088","og_site_name":"Infinite Possibilities","article_published_time":"2026-10-26T13:00:00+00:00","article_modified_time":"2026-11-06T18:36:47+00:00","og_image":[{"width":1400,"height":1294,"url":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg","type":"image\/jpeg"}],"author":"Aventure Blog","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Aventure Blog","Est. reading time":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/aventurecloud.net\/?p=8088#article","isPartOf":{"@id":"https:\/\/aventurecloud.net\/?p=8088"},"author":{"name":"Aventure Blog","@id":"https:\/\/aventurecloud.net\/#\/schema\/person\/5bb333ca4287c7b8bcd6b4f668d02dba"},"headline":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us","datePublished":"2026-10-26T13:00:00+00:00","dateModified":"2026-11-06T18:36:47+00:00","mainEntityOfPage":{"@id":"https:\/\/aventurecloud.net\/?p=8088"},"wordCount":679,"commentCount":0,"publisher":{"@id":"https:\/\/aventurecloud.net\/#organization"},"image":{"@id":"https:\/\/aventurecloud.net\/?p=8088#primaryimage"},"thumbnailUrl":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg","articleSection":["AI Research"],"inLanguage":"en-US","potentialAction":[{"@type":"CommentAction","name":"Comment","target":["https:\/\/aventurecloud.net\/?p=8088#respond"]}]},{"@type":"WebPage","@id":"https:\/\/aventurecloud.net\/?p=8088","url":"https:\/\/aventurecloud.net\/?p=8088","name":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us - Infinite Possibilities","isPartOf":{"@id":"https:\/\/aventurecloud.net\/#website"},"primaryImageOfPage":{"@id":"https:\/\/aventurecloud.net\/?p=8088#primaryimage"},"image":{"@id":"https:\/\/aventurecloud.net\/?p=8088#primaryimage"},"thumbnailUrl":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg","datePublished":"2026-10-26T13:00:00+00:00","dateModified":"2026-11-06T18:36:47+00:00","breadcrumb":{"@id":"https:\/\/aventurecloud.net\/?p=8088#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/aventurecloud.net\/?p=8088"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/aventurecloud.net\/?p=8088#primaryimage","url":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg","contentUrl":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2026\/11\/8088_rag.jpg","width":1400,"height":1294,"caption":"Open books and library shelves representing document retrieval"},{"@type":"BreadcrumbList","@id":"https:\/\/aventurecloud.net\/?p=8088#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/aventurecloud.net\/"},{"@type":"ListItem","position":2,"name":"Your RAG Tweak Might Be Noise: What a One-Change-at-a-Time Test Showed Us"}]},{"@type":"WebSite","@id":"https:\/\/aventurecloud.net\/#website","url":"https:\/\/aventurecloud.net\/","name":"Infinite Possibilities","description":"","publisher":{"@id":"https:\/\/aventurecloud.net\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/aventurecloud.net\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/aventurecloud.net\/#organization","name":"Infinite Possibilities","url":"https:\/\/aventurecloud.net\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/aventurecloud.net\/#\/schema\/logo\/image\/","url":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2024\/06\/cropped-Logo2-1-3.png","contentUrl":"https:\/\/aventurecloud.net\/wp-content\/uploads\/2024\/06\/cropped-Logo2-1-3.png","width":1560,"height":758,"caption":"Infinite Possibilities"},"image":{"@id":"https:\/\/aventurecloud.net\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/aventurecloud.net\/#\/schema\/person\/5bb333ca4287c7b8bcd6b4f668d02dba","name":"Aventure Blog","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/secure.gravatar.com\/avatar\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/825128403eec8be7b8ab517bdc5bcf08e15a5294c816412cf445094e2ad734a9?s=96&d=mm&r=g","caption":"Aventure Blog"},"url":"https:\/\/aventurecloud.net\/?author=19"}]}},"_links":{"self":[{"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/posts\/8088","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/users\/19"}],"replies":[{"embeddable":true,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=8088"}],"version-history":[{"count":0,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/posts\/8088\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=\/wp\/v2\/media\/8123"}],"wp:attachment":[{"href":"https:\/\/aventurecloud.net\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=8088"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=8088"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/aventurecloud.net\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=8088"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}