--- title: "The Learning Phase That Never Ended: How Four Sensible Edits Kept a Campaign Unstable" description: "A postmortem of a typical Google Ads campaign stuck in Learning: the target change, page swap, ad edits, and budget jump that kept moving the question Smart Bidding was trying to answer." image: "https://cdn.prod.website-files.com/6823bbd57170ea42b357cf81/6abc9d2b41d552ebd3029410_ebf96c8d-4c20-418a-a738-3aea67f41f40.png" --- September 30, 2026 • min read # The Learning Phase That Never Ended: How Four Sensible Edits Kept a Campaign Unstable ![Young man with curly hair wearing a black shirt outdoors against green foliage background.](https://cdn.prod.website-files.com/6821efca072e48f6f495a47e/68562d390107b3921a6e3d68_1743932904108.jpg) **Alexander Perleman**, Head Of Product @ groas Ex-Goldman Sachs and Stanford Computer Science alex@groas.ai [LinkedIn](https://groas.com/post/the-learning-phase-that-never-ended-anat#) ![Cover image for: The Learning Phase That Never Ended: How Four Sensible Edits Kept a Campaign Unstable](https://cdn.prod.website-files.com/6823bbd57170ea42b357cf81/6abc9d2b41d552ebd3029410_ebf96c8d-4c20-418a-a738-3aea67f41f40.png) Most campaigns stuck in Learning are not waiting for Google to wake up. The person checking the account every morning keeps changing what the campaign is supposed to learn. Take the version of this every team eventually runs into. This is a typical composite, not a report from one account. A new Search campaign launches on Target CPA at $80 a day. It brings in roughly five or six conversions a week, but by Day 20 the status still says Learning and daily CPA is swinging around. The owner blames the algorithm. I would start with the change history. Each decision in that history makes sense on its own. The target looks too ambitious, so the owner adjusts it. The page looks weak, so someone improves it. An ad strength prompt asks for new headlines. Volume looks thin, so the budget doubles. None of this feels like sabotage. Together, the edits keep moving the question Smart Bidding is trying to answer. #### Day 1: a clean launch with a problem already built in Say this is a home services account. The owner sets Target CPA at $65 because that is the number the business needs to make the math work. The structure is deliberately simple: one campaign, two ad groups split by intent, three exact-ish keywords in each, and one landing page. Tracking checks out. The first clicks come in Tuesday afternoon, and the first conversion lands Wednesday morning. The status column says **Learning**. That is not an alarm. Google uses that status while a new or reactivated Smart Bidding strategy calibrates; [setting and structural changes can also put it into Learning](https://support.google.com/google-ads/answer/13020501). At launch, the label tells the owner very little about whether the $65 target will work. The campaign has barely begun collecting the conversions it needs to inform bids. The trouble is visible elsewhere. The target is tight, volume is low, and the page is one the owner already wants to replace. A campaign can launch clean and still be set up for a nervous first week. **A valid Learning label is not a request to intervene.** #### Day 4: the first sign becomes the wrong call By Friday, the owner sees four clicks at $9 each and no second conversion. He lowers Target CPA from $65 to $50 to “control costs.” I used to make this exact move for clients. I was wrong. A lower target is not a brake pedal that instantly makes the same clicks cheaper. It changes the instruction Smart Bidding is working from while the campaign has almost no conversion history. Now the system has to calibrate against a stricter goal, and the owner has even less basis for judging whether the original one was attainable. Four clicks are not a verdict on CPA. They are four clicks. If those clicks are irrelevant, that is worth investigating. If they are relevant but have not converted by Friday, tightening the target does not tell the owner what they were worth. It just changes the brief before the first brief has had a fair run. This is the first point where the postmortem turns. The owner believes he has responded to evidence. **He has responded to a sample too small to support the decision.** My rule at that moment is plain: while a new campaign is learning, leave the target alone unless there is an actual setup problem to correct. #### Day 8: a better-looking page changes the experiment By the following Monday, CPA is above the new $50 target. The owner looks at the landing page, decides the bounce rate looks soft, and asks his web guy to swap in a template with a shorter form and a click-to-call button near the top. Same offer, different page. It feels like basic hygiene. That Friday, he adds two responsive search ad headlines because Google's ad strength meter says Average. Also reasonable in isolation. The page and the ads could improve the campaign. The problem is timing: neither has been given room to show what it does before the next change arrives. I call this the *invisible reset*, with an important distinction. A page swap does not necessarily flip the status column back to Learning that afternoon. Neither does every ad edit. But both can change what happens after a click. The bid model has started observing which queries, devices, and hours lead to a filled form on the old page. A shorter form and a prominent call button can change the mix of outcomes it sees. New headlines can change which clicks the ads attract. The owner is still reading one continuous campaign history. Underneath it, the conditions have changed. When CPA moves, which decision caused it: the $50 target, the new page, the new headlines, or the ordinary noise of a thin account? There is no clean answer. That is the damage these edits do even when the interface never announces a fresh start. Google identifies setting and structural changes as Learning triggers, and [calibration can continue after the label clears](https://support.google.com/google-ads/answer/13020501). Page and ad changes belong in the same operational conversation, not because every one produces the same status change, but because they change the evidence available to judge the bids. **A page swap during Learning makes the old CPA harder to interpret.** If the page is broken, fix it and accept the disruption. If it merely looks less polished than a new template, wait. #### Day 12: the conversion math the owner skipped By Day 12, spend is about $960. CPA is uncomfortable, the status still says Learning, and the owner concludes that Google is taking too long. I would check the conversion pace before calling the platform slow. Smart Bidding needs completed conversion cycles to inform its bids. Google describes calibration as typically taking [one to two conversion cycles](https://support.google.com/google-ads/answer/13020501), and this business can take five to seven days to get from a click through a signed form and confirmed call. That delay matters. A click from late last week may not yet have produced the outcome the owner cares about. Meanwhile, the target, page, and ads have already changed. ##### The math: six weeks of evidence is not six weeks of status Here is the arithmetic I wish had been on the launch sheet. Google recommends judging Smart Bidding over periods with [at least 30 conversions, or 50 for Target ROAS](https://support.google.com/google-ads/answer/7065882), measured over a month or longer. At five conversions a week, reaching 30 takes roughly six weeks. That is a planning horizon for evaluating performance, **not a promise that the Learning label will disappear at conversion 30**. Run the click math, too. At $80 a day and $9 a click, the budget buys about nine clicks a day. At a 7% landing-page conversion rate, that is roughly 0.6 conversions a day, or around four a week. These are useful approximations, not a forecast down to the decimal. Either way, this is a low-volume campaign. Daily CPA will swing because one conversion arriving today instead of tomorrow changes the daily number dramatically. The owner is asking a different question: why has the label not cleared? It is understandable, but the status column is a poor substitute for the change history and the conversion lag. Even if the label cleared tomorrow, the account would still need enough comparable outcomes to judge its target. And if the label stays, checking it twice a day does not create those outcomes. **Calculate the conversion pace before launch.** If the budget buys roughly five conversions a week, do not build a management plan that requires a confident CPA verdict after four days. #### Day 18: more budget makes the moving target bigger On Day 18, the owner tries to buy his way out. The budget goes from $80 to $160 overnight to “get more data faster.” I understand the logic. More spend can bring in more clicks, and more clicks can bring in more conversions. But doubling the budget also asks the strategy to handle substantially more traffic while the owner is still trying to understand performance under the previous conditions. Practitioners who scale cautiously often use increases of [around 20% held for at least a week](https://www.nautilusmarketing.co.uk/how-to-scale-google-ads-without-wasting-budget), or [15 to 20% with a week or two between moves](https://www.aviddemand.com/blog/5-things-to-verify-before-increasing-your-google-ads-budget). Those are operating rules, not a magical threshold below which nothing changes. Their value here is that they leave time to see what the last move did. A 100% jump does the opposite. By Day 20, the owner is staring at volatile CPA and a campaign that still has not settled. He reads that as the algorithm failing. I read four changes in less than three weeks: 1. A $65 Target CPA became $50 after four clicks. 2. The landing page changed before the first setup had much history. 3. New headlines changed the ads while the page change was still fresh. 4. The daily budget doubled before the owner could separate the effects of the first three moves. The budget increase is not proof that every earlier observation has been erased. The more useful diagnosis is that **there is no stable stretch of comparable performance to examine**. Spend has continued, but each new decision has made the previous days less useful as a guide to the next ones. Calling that a Google problem lets the change history off too easily. ![Cartoon hamster wheel labeled Learning, representing repeated campaign edits](https://cdn.prod.website-files.com/6823bbd57170ea42b357cf81/6abc9d2e41d552ebd3029449_ab29e284-e986-4c00-8a60-8640d29194db.png) #### The decision that could have prevented the loop The mistake was not that the owner cared about CPA, or that the page could be improved. It was launching a setup with thin expected volume and a page he already distrusted, then treating early readings as instructions to tune. On Day 0, he chose a tight $65 target because the business needed that number. On Day 4, the account had not produced enough evidence to show whether the target was workable. He responded by making it tighter. The alternative decision belongs at launch, before the first click. Choose the page the campaign is actually going to use. Keep the conversion action clear. If the goal is to learn which traffic converts before imposing a target, start on Maximize Conversions without a tCPA cap for that first stretch, within the planned budget. Then put the review point in writing. Do not wait until the status column makes everyone nervous to decide what counts as enough evidence. Consolidation matters, but I would not pretend it manufactures conversions. This example already has one campaign. Splitting its modest traffic into more campaigns or rotating page versions would make each comparison thinner, not create more total data. Keep the setup simple so the conversions that do arrive answer the same question. A single page and a single conversion action give the owner a cleaner history to read. The uncomfortable part is that the business may not like what that history eventually says. The actual CPA might remain above $65. A stable test does not guarantee an affordable campaign; it tells the owner whether the campaign is affordable without making him guess which edit caused the latest swing. **Stability is how you get an answer, not how you force a favorable one.** #### The fix: freeze the setup, then read the result When I see this pattern, I stop optimizing for a stretch. I check that the conversion action and page are the ones the business intends to use, then freeze the moving parts: no target changes, new ads, page swaps, or new keywords while the account builds a comparable run of conversions. Pure-junk search terms can still get negative keywords, batched rather than handled as a daily excuse to keep tinkering. A genuinely broken page or broken tracking gets fixed immediately. A soft bounce rate does not get the same privilege. The first week feels like neglect. Clicks cost what they cost. CPA wobbles. The Learning label may sit there, and nobody gets the satisfaction of pointing to a busy change log. That is exactly why the freeze is hard. In a low-volume account, motion looks like management even when it destroys the comparison the owner needs. The review is about outcomes, not the ceremony of watching a status change. At roughly five conversions a week, 30 conversions takes around six weeks from a clean starting point; the actual pace may differ. Once there is a meaningful run of results, the owner can compare actual CPA with the business's $65 requirement. If setting a tCPA cap then makes sense, base it on observed CPA rather than the number everyone wished for on Day 1. If budget needs to rise, make the move in smaller steps and give each one time to show its effect. None of that guarantees the label clears on a particular day. It does prevent the account from answering a different question every four days. The fix is boring, which is why people keep missing it. My rule now fits on a sticky note: **before I change a campaign in Learning, I name the problem the change will fix and the evidence that it exists.** Four clicks and a nervous owner do not qualify. ## Related Posts [![](https://cdn.prod.website-files.com/plugins/Basic/assets/placeholder.60f9b1840c.svg) ##### Agency AEO: Managing Knowledge Graphs and Auto-Publishing to Client CMS Guide for agencies on evaluating AEO software for knowledge-graph and citation management and auto-publishing to client CMS, and how groas documents that workflow. September 30, 2026 • min read Written by](https://groas.com/post/agency-aeo-knowledge-graph-cms-publishing-guide) [![](https://cdn.prod.website-files.com/plugins/Basic/assets/placeholder.60f9b1840c.svg) ##### Google Ads Measurement in 2026: Consent Mode v2, GA4 Updates, and Privacy Sandbox Google Ads measurement in 2026: Consent Mode v2, GA4 updates, enhanced conversions and Chrome third-party cookies status explained. September 30, 2026 • min read Written by](https://groas.com/post/google-ads-measurement-consent-mode-ga4-privacy-sandbox) [![](https://cdn.prod.website-files.com/plugins/Basic/assets/placeholder.60f9b1840c.svg) ##### ROAS Agency vs Profitability Optimization: What to Choose ROAS agency vs profit optimization: when to chase ROAS, when to optimize for profit, and how to fix stagnant Google Ads. September 30, 2026 • min read Written by](https://groas.com/post/roas-agency-profitability-optimization) ![](https://cdn.prod.website-files.com/6821efca072e48f6f495a47e/6823c03839b2b618430d5ae6_Light-p-500.png) [![White stylized owl eyes with green background icon.](https://cdn.prod.website-files.com/6821efca072e48f6f495a47e/6a68a036a5dd5627409d3579_Groas%20Icon.png)](https://groas.com/post/the-learning-phase-that-never-ended-anat#) [contact](mailto:scale@groas.ai) Explore [Home](https://groas.com/)[Philosophy](https://groas.com/our-philosophy)[Blog](https://groas.com/blog) What We Do [Paid Search](https://groas.com/paid-search)[Earned Search](https://groas.com/earned-search) Who We Serve [Businesses](https://groas.com/for-businesses)[Agencies](https://groas.com/for-agencies) MCP [ChatGPT](https://chatgpt.com/plugins/plugin_asdk_app_6a9c2d7108f08191beb5b3d736be8655?q=groas)[Claude](https://claude.ai/directory/mcp-groas-ai) Legal [Privacy Policy](https://groas.com/legal/privacy-policy)[Terms of Service](https://groas.com/legal/terms) Get Started [Apply](https://groas.typeform.com/to/xC1bQNUT) © 2026 groas 🇺🇸 [![](https://cdn.prod.website-files.com/62434fa732124a0fb112aab4/62434fa732124a389912aad8_linkedin%20small.svg)](https://www.linkedin.com/company/groas/) ```json { "@context": "https://schema.org", "@graph": [ { "@type": "", "@id": "https://www.groas.com/post/the-learning-phase-that-never-ended-anat#article", "headline": "The Learning Phase That Never Ended: How Four Sensible Edits Kept a Campaign Unstable", "description": "A postmortem of a typical Google Ads campaign stuck in Learning: the target change, page swap, ad edits, and budget jump that kept moving the question Smart Bidding was trying to answer.", "url": "https://www.groas.com/post/the-learning-phase-that-never-ended-anat", "image": { "@type": "ImageObject", "url": "https://cdn.prod.website-files.com/6823bbd57170ea42b357cf81/6abc9d2b41d552ebd3029410_ebf96c8d-4c20-418a-a738-3aea67f41f40.png" }, "thumbnailUrl": "https://cdn.prod.website-files.com/6823bbd57170ea42b357cf81/6abc9d2b41d552ebd3029410_ebf96c8d-4c20-418a-a738-3aea67f41f40.png", "datePublished": "2026-09-30T05:25:02.757Z", "dateModified": "2026-09-30T05:25:02.757Z", "inLanguage": "en-US", "isAccessibleForFree": true, "articleSection": "", "keywords": "Google Ads, ", "mainEntityOfPage": { "@type": "WebPage", "@id": "https://www.groas.com/post/the-learning-phase-that-never-ended-anat" }, "author": { "@id": "https://www.groas.com/author/#person" }, "publisher": { "@id": "https://www.groas.com#organization" }, "about": [ { "@type": "Thing", "name": "Google Ads" }, { "@type": "Thing", "name": "Pay-per-click advertising" }, { "@type": "Thing", "name": "Performance marketing" } ], "mentions": [ { "@type": "Organization", "name": "groas", "url": "https://www.groas.com" }, { "@type": "Thing", "name": "Google Ads" } ], "speakable": { "@type": "SpeakableSpecification", "cssSelector": ["h1"] } }, { "@type": "Person", "@id": "https://www.groas.com/author/#person", "name": "", "description": "", "jobTitle": "", "email": "", "url": "https://www.groas.com/author/", "image": { "@type": "ImageObject", "url": "" }, "sameAs": [ "" ], "knowsAbout": [ "Google Ads", "Performance Max", "AI Max", "Pay-per-click advertising", "Conversion tracking", "Bid management", "Search advertising" ], "worksFor": { "@id": "https://www.groas.com#organization" } }, { "@type": "Organization", "@id": "https://www.groas.com#organization", "name": "groas", "alternateName": "groas.com", "url": "https://www.groas.com", "logo": { "@type": "ImageObject", "url": "https://cdn.prod.website-files.com/6821efca072e48f6f495a47e/6834a970e03a7016560e3515_Logo%20Design%20256x256.png" }, "description": "Dedicated strategists run your account and a proprietary engine trained on $500B in profitable ad spend optimises the execution underneath them", "sameAs": [ "https://www.linkedin.com/company/groas" ], "knowsAbout": [ "Google Ads", "Performance Max", "AI Max for Search", "Pay-per-click advertising", "Autonomous campaign management", "AI advertising agents" ], "areaServed": "Worldwide" }, { "@type": "BreadcrumbList", "itemListElement": [ { "@type": "ListItem", "position": 1, "name": "Home", "item": "https://www.groas.com" }, { "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://www.groas.com/blog" }, { "@type": "ListItem", "position": 3, "name": "", "item": "https://www.groas.com/category/" }, { "@type": "ListItem", "position": 4, "name": "The Learning Phase That Never Ended: How Four Sensible Edits Kept a Campaign Unstable", "item": "https://www.groas.com/post/the-learning-phase-that-never-ended-anat" } ] } ] } ```