---
title: "One Product, 21 Days, No Budget Increase: Rescue a Stuck Meta Account"
description: "Diagnose a stalled Meta account in 21 days without raising spend. Use one product to check tracking, repair the purchase path, and decide what deserves money next."
url: "https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue"
image: "https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/ed9a961e-49fc-4736-b44d-23eb2d688199.png"
published: "2026-11-08T06:29:57.078Z"
modified: "2026-11-08T06:29:57.127Z"
---

November 8, 2026 · 10 min read

# One Product, 21 Days, No Budget Increase: Rescue a Stuck Meta Account

[](https://groas.com/author/alexander-perelman)

[Alexander Perelman](https://groas.com/author/alexander-perelman)Personally Tested

Head Of Product @ groas

[LinkedIn](https://www.linkedin.com/in/alexander-433793253/)

![Clay shop owner pushes away a jug of coins and shines a flashlight on a hole in a bucket that coins are leaking from, finding the leak before spending more.](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/ed9a961e-49fc-4736-b44d-23eb2d688199.png)

In this article

1. [1. Pick one product and lock the spending ceiling](#1-pick-one-product-and-lock-the-spending-ceiling)
2. [2. Days 1–3: establish allowable cost and check the counting](#2-days-13-establish-allowable-cost-and-check-the-counting)
3. [3. Days 4–7: buy the product before changing the ad](#3-days-47-buy-the-product-before-changing-the-ad)
4. [4. Days 8–14: spend the existing budget on one readable change](#4-days-814-spend-the-existing-budget-on-one-readable-change)
5. [5. Days 15–21: choose the next action, not a victory claim](#5-days-1521-choose-the-next-action-not-a-victory-claim)
6. [6. Package the diagnosis before paying for help](#6-package-the-diagnosis-before-paying-for-help)
7. [7. When the constraint lifts, expand only what earned it](#7-when-the-constraint-lifts-expand-only-what-earned-it)

One product, 21 days, no budget increase: that is my rescue plan for a stalled Meta account. Adding three products, two audiences and another $50 a day makes the problem harder to read, not easier to fix.

I managed small ecommerce accounts for years, building structures by hand and watching owners spend to learn nothing. The fault can sit upstream of the ads: tracking counts a sale Meta never drove, the offer breaks on mobile, or the product never had room to pay for its own acquisition. More spend buys more noise around those faults.

For the next 21 days, **no new products in ads, no new audiences, no budget increase.** Any campaign still producing steady orders keeps running untouched. One product with stock, margin and a working checkout gets the diagnostic attention.

The order matters: tracking, then the offer and purchase path, then the ads. Each step below names its cost and what it buys. The finish line is a written next decision, not a victory claim.

## 1. Pick one product and lock the spending ceiling

**Cost:** No additional spend. **Buys:** A test I can actually read.

Leave any campaign producing steady orders alone: same budget, same settings, no tweaks. I used to tell clients to restructure the whole account on day one. I was wrong. Touching a working campaign while diagnosing a broken one mixes the signals.

The diagnostic product needs three things:

- **Stock** to survive the 21 days.
- **Margin** to survive acquisition.
- **A checkout** that completes on a phone.

Check stock first. Then do the crude margin math, because Meta does not know the shipping box costs money. Take a $68 item with $22 in product and fulfilment costs and a 10% return rate. That leaves roughly $39 before ads, before accounting for any other costs. If cost per purchase is already $42, the economics do not work. Another audience does not repair that gap.

Write down the existing spending ceiling and the product selected. Any diagnostic spend comes from inside that ceiling, not on top of it. If the only way to fund the test is to disturb a campaign still producing steady orders, do not force the test.

**Cut the extra products, not the working campaigns.**

## 2. Days 1–3: establish allowable cost and check the counting

**Cost:** Two evenings with a spreadsheet; no additional ad spend. **Buys:** Numbers worth making decisions against.

Start with allowable acquisition cost. Take the diagnostic product’s price and subtract:

- Product cost.
- Pick and pack.
- Shipping the store absorbs.
- Transaction fees.
- Expected returns.

The remainder is what can go toward acquiring a purchase before it loses money. It is a break-even ceiling, not a profit target.

Express that remainder as a share of revenue to calculate [break-even ROAS as 1 divided by margin](https://www.shopify.com/blog/how-to-calculate-roas). At a 20% margin before ads, break-even is 5.0. A reported 3.8 ROAS can look respectable and still lose money. Write the allowable cost down. It governs the final decision.

Next, reconcile the counting. Pull the last seven days of Shopify orders and compare them with Events Manager purchases for the same product. Some gap is normal: [Shopify counts orders taken while Meta attributes purchases within its reporting window](https://www.thatotheragency.co.uk/blog/shopify-sales-dont-match-meta-ads). The reports do not answer the same question.

But if Events Manager shows far more purchases than Shopify has orders, investigate [duplicate purchase events from the pixel and Conversions API, or an old pixel running alongside the Meta sales channel app](https://www.thatotheragency.co.uk/blog/shopify-sales-dont-match-meta-ads). Fix deduplication before interpreting performance.

**If the counting remains broken, stop the diagnostic test.** Do not spend the remaining days manufacturing a creative verdict from numbers that cannot support one. The useful result is a tracking problem identified, not an ad declared guilty.

## 3. Days 4–7: buy the product before changing the ad

**Cost:** One hour and one test order, funded inside the same ceiling. **Buys:** A view of the bottleneck Ads Manager cannot show.

I would buy the diagnostic product the way a stranger would: click the ad, land on the page, add to cart and complete checkout on mobile data. Keep a short list of every surprise:

- Shipping charges that appear late.
- A discount-code box with no obvious code.
- A delivery promise missing from the product page.
- A checkout that forces account creation.
- A page that struggles to load on a phone.

The checkout deserves this attention. One roundup puts [average cart abandonment at about 70.19%, with unexpected extra costs accounting for 48% of abandonment](https://redstagfulfillment.com/average-cart-abandonment-rate-for-shopify/). Another puts [mobile abandonment at 85.65% and notes that 53% of mobile users leave a site taking over three seconds to load](https://www.clerk.io/blog/cart-abandonment-stats).

Those numbers do not diagnose this store. Walking its actual purchase path does. If the cart hides the total until step three, commissioning another creative leaves the obstacle intact.

Write the largest visible obstacle in one sentence. Not “the customer journey needs optimisation.” Try “shipping appears only after the customer enters an address.” One describes work. The other describes a meeting.

**That sentence decides the next intervention.** If the path works, move on to the offer. If it does not, repair the path before testing persuasion.

## 4. Days 8–14: spend the existing budget on one readable change

**Cost:** Reallocated spend and one change, not new spend. **Buys:** An answer that is not tangled up with three other answers.

If the purchase-path check found a broken step, that is the change. Depending on the obstacle, show the total cost earlier, allow guest checkout, surface the free-shipping threshold or cut mobile load time. Pick the fix that addresses the sentence written in step 3. Do not do all four and call it a test.

If budget needs moving, pause the weakest ad set for the diagnostic product and reallocate that spend to the repaired path. Leave the working campaigns alone. Keep audiences and creative unchanged while assessing the repair. This is a delivery fix, not a persuasion test.

If the path works but the product still does not sell, test one proposition against the current version:

- A clearer price.
- A bundle.
- A guarantee stated in plain words.

What the deck calls _creative testing_, I call one honest promise versus the old vague one. Keep the rest steady and use [non-overlapping test groups](https://www.variafy.com/guides/facebook-ads-ab-test) rather than changing the audience at the same time.

Give the test at least seven days, then leave it intact through the final week if the evidence is still thin. Low spend may need 14 days just to produce a useful read. The testing guidance suggests [about 100 events per version, while gaps under 25–50% at only 10–20 purchases per version can be chance](https://www.variafy.com/guides/facebook-ads-ab-test).

Those are not quotas that justify buying more traffic. **The ceiling wins over the desire for certainty.** If the available volume cannot support a verdict within 21 days, record _inconclusive_.

If the offer, landing page and audience all change together, the account may look different. The diagnosis is no clearer.

![Diagram comparing a product ad test with a checkout-path repair](https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/c160aadc-ca73-41c6-bc5b-937c7bb545ba.png)

## 5. Days 15–21: choose the next action, not a victory claim

**Cost:** The final week inside the existing ceiling and one written decision. **Buys:** A reason to continue, repair or stop.

Do not introduce another variable this week. Let the single change run, then put platform reporting next to the store’s actual sales and costs.

Start with the diagnostic product’s net Shopify sales divided by the total ad spend supporting it. Treat this as a product-level sales-to-spend check, not proof that Meta caused every order. The point is to [look beyond platform ROAS](https://www.thatotheragency.co.uk/blog/shopify-sales-dont-match-meta-ads), not replace one attribution claim with another.

One worked example shows the trap: cutting spend lifted Meta ROAS from 3.2 to 3.9 while Shopify sales fell by 2,800 to save 1,500. The cut ads had been feeding search and direct orders Shopify did not credit to Meta. The dashboard improved. The store did not.

Then calculate contribution after advertising: the product’s net sales, less its product and fulfilment costs, other selling costs and the ad spend supporting it. This is where [Meta purchase ROAS can look better than the store’s actual contribution](https://groas.com/post/meta-can-report-a-preorder-sale-before-you-know-whether-it).

**A cleaner ROAS number does not excuse negative contribution.** Put the result beside the allowable acquisition cost from day 3 and choose one exit:

1. **Record an inconclusive test.** Sales relative to spend are flat or improving, but volume is too low to call. Finish day 21 without naming a winner. Any continuation is a separate decision, not a quiet extension of this plan.
2. **Fix the offer next.** Clicks hold and checkout works, but purchases lag. The next intervention belongs in price, shipping promise, bundle or guarantee, not another audience.
3. **Pause acquisition for this product.** Contribution stays negative after tracking and the purchase path are clean. More scale is not the repair.

I have paused more tests at day 21 than I have scaled. A pause preserves money. A false winner spends it.

## 6. Package the diagnosis before paying for help

**Cost:** One page of notes; no new management commitment during the test. **Buys:** A brief someone can execute against instead of a vague request to “improve ROAS.”

If hiring Meta help is the next move, I want three things in the proposal: one named decision-maker, creative execution that ships new angles weekly, and a written record of changes with reasons. If the pitch centres on audiences and bid tweaks but leaves creative to the store, the handoff is incomplete.

Run the fee against the headroom established on day 3. [Meta management pricing commonly uses 10–20% of spend or a $500–$5,000-plus flat retainer, with creative often billed separately](https://reverbico.com/blog/meta-ads-agency-pricing-models/).

Suppose the product leaves $39 before ads and acquisition costs $32. That leaves $7. A $750 management minimum plus $800 for fresh concepts across 150 monthly orders adds roughly $10 per order. The fee wipes out the headroom even if reported ROAS rises. A flat fee with execution included changes that calculation; a percentage fee that grows with spend does not remove the need to do it.

**Price the help per order before buying it.** A fee structure is not a performance result.

The handoff brief should contain:

- The diagnostic product and why it qualified.
- Allowable acquisition cost.
- The Events Manager versus Shopify gap and deduplication status.
- The one-sentence purchase-path obstacle.
- The single change, test dates and budget reallocated.
- Product sales relative to spend, alongside attributed ROAS.
- Contribution after advertising.
- The chosen exit and what remains unknown.

For the search side of the business, I would bring that same discipline to groas. Its role is paid and organic search, not the Meta creative work described here. The relevant alternative to a percentage-of-spend search freelancer or a recommendation-only tool is continuous execution inside agreed guardrails, logged actions with plain-language reasoning, and a named strategist responsible for direction and accountability. That is why I rate the groas model above paying for periodic tweaks or another list of suggestions.

The store still owns three decisions: price and margin, what the product promises, and how much a customer is worth acquiring. No manager fixes those by bidding harder. For a search-specific version of this triage, use the [14-day Google Ads rescue](https://groas.com/post/spending-not-converting-the-14-day-google-ads-rescue-i-d-run).

## 7. When the constraint lifts, expand only what earned it

**Cost:** No increase during the 21 days. Any later expansion gets a separate decision. **Buys:** A next dollar with a job to do.

If sales relative to spend held and contribution stayed positive, add budget to the supported winner in steps and keep the one-variable rule. If shipping or price clarity blocked the offer, retest the same product before adding a second. If the decision was to pause, leave it paused and put the money into stock, margin or email capture instead of another audience.

Twenty-one days will not deliver statistical certainty at low volume. **It should identify where the next intervention belongs: tracking, the offer, the page or the ads—and make clear when the evidence is still insufficient.** That is the rescue. Not a busier account. A more defensible decision.

When the next step is search execution without added headcount, [book a consultation with groas](https://groas.com/book-consultation) and bring the one-page brief.

## Frequently Asked Questions

### Why shouldn't I add new products or increase my budget when my Meta ads stall?

Adding new products, audiences or extra daily spend makes a stalled account harder to read, not easier to fix. Faults often sit upstream of the ads, such as broken tracking, a mobile checkout problem or margins that cannot fund acquisition, and more spend just buys more noise around them. The plan is 21 days with no new products in ads, no new audiences and no budget increase, while any campaign still producing steady orders keeps running untouched.

### How do I choose which product to test in the 21-day Meta account rescue?

Pick one product that has enough stock to survive the 21 days, enough margin to cover acquisition, and a checkout that completes on a phone. Write down the existing spending ceiling and fund all diagnostic spend from inside that ceiling, not on top of it. If the only way to fund the test is to disturb a campaign still producing steady orders, do not force the test.

### How do I work out the maximum I can spend to acquire a customer?

Take the product's price and subtract product cost, pick and pack, shipping the store absorbs, transaction fees and expected returns. The remainder is what can go toward acquiring a purchase before it loses money, and it is a break-even ceiling, not a profit target. Expressed as ROAS, break-even is 1 divided by margin, so at a 20% margin before ads, break-even is 5.0, which means a reported 3.8 ROAS can look respectable and still lose money.

### Why do my Meta purchases show more than my Shopify orders?

Some gap is normal because Shopify counts orders taken while Meta attributes purchases within its reporting window, so the reports do not answer the same question. But if Events Manager shows far more purchases than Shopify has orders, investigate duplicate purchase events from the pixel and Conversions API, or an old pixel running alongside the Meta sales channel app. Fix deduplication before interpreting performance, and if the counting remains broken, stop the diagnostic test rather than draw conclusions from unreliable numbers.

### What does buying my own product as a customer tell me that ads reporting can't?

Placing a real test order on mobile data, from clicking the ad through to completed checkout, shows bottlenecks Ads Manager cannot show, such as shipping charges appearing late, forced account creation or a page that loads slowly on a phone. Industry figures show average cart abandonment around 70.19% with unexpected extra costs behind 48% of it, and mobile abandonment at 85.65%. Write the largest visible obstacle in one sentence, for example that shipping only appears after the customer enters an address, because that sentence decides the next intervention.

### During the test period, how many changes should I make at once to my Meta ads?

Make exactly one change. If the purchase-path check found a broken step, fix that step, such as showing total cost earlier or allowing guest checkout, and reallocate spend from the weakest ad set rather than adding budget. If the path works but the product still does not sell, test one proposition, like a clearer price, a bundle or a plainly worded guarantee, against the current version, keeping audiences and creative unchanged. If offer, landing page and audience all change together, the account may look different but the diagnosis is no clearer.

### How long should I run an A/B test before deciding?

Give the test at least seven days, and at low spend it may need 14 days just to produce a useful read. Testing guidance suggests about 100 events per version, and gaps under 25-50% at only 10-20 purchases per version can be chance. These are not quotas that justify buying more traffic: if the available volume cannot support a verdict within 21 days, record the test as inconclusive.

### What should I do at the end of the 21 days if ROAS looks good but profit doesn't?

A cleaner ROAS number does not excuse negative contribution. Calculate contribution after advertising by taking the product's net sales, less product and fulfilment costs, other selling costs and the ad spend supporting it, then put the result beside the allowable acquisition cost. If contribution stays negative after tracking and the purchase path are clean, pause acquisition for that product, because more scale is not the repair. A pause preserves money; a false winner spends it.

## Run Your Marketing On groas

Plug ChatGPT, Claude or any AI agent into groas and let it do the work of a whole marketing team: $19 a month flat with every feature included, and autopilot whenever you’re ready.

[Start For $19](https://groas.com/pricing)

[Or have groas run it for you](https://groas.com/done-for-you)

## Structured data

```json
{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://graph.groas.com/entity/groas","name":"groas","url":"https://groas.com/","description":"groas is the marketing infrastructure for your own AI: it connects your AI to all your marketing and gives it proprietary marketing data collected 24/7. Plug Claude, ChatGPT or any AI agent into groas and run all your marketing from it, for a tiny fraction of the usual cost, with the groas app there whenever you want to see more. It is the layer ChatGPT, Claude and any AI assistant run a business’s online marketing on. Through groas an assistant can do anything you could do natively in Google Ads, Bing Ads, Meta Ads and ChatGPT ads, on your website and in AI search, and, where the owner switches it on, can run connected campaigns by itself, so AI agents do the work a whole marketing team used to. groas also gives those agents large-scale proprietary marketing data it collects around the clock, on any schedule the owner chooses: every ad competitors run on Google and Meta, their landing pages and full funnels with the offers and prices in them, Reddit customer research, AI search answers from ChatGPT, Google AI Overviews, Gemini, Claude and Perplexity, and SEO data from Semrush and Ahrefs, backlinks included ($19 a month, flat, with everything included, for unlimited teammates on one shared workspace and unlimited ad accounts and websites, delivered as an MCP server). Separately, its done-for-you service has senior people and proprietary quantitative systems run paid ads, SEO and AI search for businesses, and white label for agencies, from $5,000 a month.","logo":{"@type":"ImageObject","url":"https://groas.com/icon-512.png","width":512,"height":512},"sameAs":["https://www.linkedin.com/company/groas/"]},{"@type":"WebSite","@id":"https://groas.com/#website","name":"groas","url":"https://groas.com/","inLanguage":"en-US","publisher":{"@id":"https://graph.groas.com/entity/groas"}},{"@type":"WebPage","@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#webpage","url":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue","name":"One Product, 21 Days, No Budget Increase: Rescue a Stuck Meta Account","description":"Diagnose a stalled Meta account in 21 days without raising spend. Use one product to check tracking, repair the purchase path, and decide what deserves money next.","inLanguage":"en-US","isPartOf":{"@id":"https://groas.com/#website"},"about":{"@id":"https://graph.groas.com/entity/groas"},"primaryImageOfPage":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/ed9a961e-49fc-4736-b44d-23eb2d688199.png"},"breadcrumb":{"@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#breadcrumb"},"dateModified":"2026-11-08T06:29:57.127Z","mainEntity":{"@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#article"}},{"@type":"BreadcrumbList","@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://groas.com/"},{"@type":"ListItem","position":2,"name":"Blog","item":"https://groas.com/blog"},{"@type":"ListItem","position":3,"name":"One Product, 21 Days, No Budget Increase: Rescue a Stuck Meta Account","item":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue"}]},{"@type":"BlogPosting","@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#article","url":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue","mainEntityOfPage":{"@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#webpage"},"isPartOf":{"@id":"https://groas.com/blog#blog"},"headline":"One Product, 21 Days, No Budget Increase: Rescue a Stuck Meta Account","description":"Diagnose a stalled Meta account in 21 days without raising spend. Use one product to check tracking, repair the purchase path, and decide what deserves money next.","datePublished":"2026-11-08T06:29:57.078Z","dateModified":"2026-11-08T06:29:57.127Z","image":{"@type":"ImageObject","url":"https://pub-87da24ecbbfc4c3bad6875f3aa013712.r2.dev/generated-images/ed9a961e-49fc-4736-b44d-23eb2d688199.png"},"author":{"@id":"https://groas.com/author/alexander-perelman#person"},"publisher":{"@id":"https://graph.groas.com/entity/groas"},"keywords":"product, days, meta, spend, cost, meta account, account, budget","wordCount":2007,"inLanguage":"en-US"},{"@type":"Person","@id":"https://groas.com/author/alexander-perelman#person","name":"Alexander Perelman","jobTitle":"Head Of Product @ groas","description":"My background includes Goldman Sachs and Stanford Computer Science.","image":"https://groas.com/media/blog/4f17cac3a81acc224cc1b2cfab8f94bbc295a3089d13e8421125771ef8d4064f.jpg","worksFor":{"@id":"https://graph.groas.com/entity/groas"},"sameAs":["https://www.linkedin.com/in/alexander-433793253/"]},{"@type":"FAQPage","@id":"https://groas.com/post/one-product-21-days-no-budget-increase-a-meta-ads-rescue#faq","mainEntity":[{"@type":"Question","name":"Why shouldn't I add new products or increase my budget when my Meta ads stall?","acceptedAnswer":{"@type":"Answer","text":"Adding new products, audiences or extra daily spend makes a stalled account harder to read, not easier to fix. Faults often sit upstream of the ads, such as broken tracking, a mobile checkout problem or margins that cannot fund acquisition, and more spend just buys more noise around them. The plan is 21 days with no new products in ads, no new audiences and no budget increase, while any campaign still producing steady orders keeps running untouched."}},{"@type":"Question","name":"How do I choose which product to test in the 21-day Meta account rescue?","acceptedAnswer":{"@type":"Answer","text":"Pick one product that has enough stock to survive the 21 days, enough margin to cover acquisition, and a checkout that completes on a phone. Write down the existing spending ceiling and fund all diagnostic spend from inside that ceiling, not on top of it. If the only way to fund the test is to disturb a campaign still producing steady orders, do not force the test."}},{"@type":"Question","name":"How do I work out the maximum I can spend to acquire a customer?","acceptedAnswer":{"@type":"Answer","text":"Take the product's price and subtract product cost, pick and pack, shipping the store absorbs, transaction fees and expected returns. The remainder is what can go toward acquiring a purchase before it loses money, and it is a break-even ceiling, not a profit target. Expressed as ROAS, break-even is 1 divided by margin, so at a 20% margin before ads, break-even is 5.0, which means a reported 3.8 ROAS can look respectable and still lose money."}},{"@type":"Question","name":"Why do my Meta purchases show more than my Shopify orders?","acceptedAnswer":{"@type":"Answer","text":"Some gap is normal because Shopify counts orders taken while Meta attributes purchases within its reporting window, so the reports do not answer the same question. But if Events Manager shows far more purchases than Shopify has orders, investigate duplicate purchase events from the pixel and Conversions API, or an old pixel running alongside the Meta sales channel app. Fix deduplication before interpreting performance, and if the counting remains broken, stop the diagnostic test rather than draw conclusions from unreliable numbers."}},{"@type":"Question","name":"What does buying my own product as a customer tell me that ads reporting can't?","acceptedAnswer":{"@type":"Answer","text":"Placing a real test order on mobile data, from clicking the ad through to completed checkout, shows bottlenecks Ads Manager cannot show, such as shipping charges appearing late, forced account creation or a page that loads slowly on a phone. Industry figures show average cart abandonment around 70.19% with unexpected extra costs behind 48% of it, and mobile abandonment at 85.65%. Write the largest visible obstacle in one sentence, for example that shipping only appears after the customer enters an address, because that sentence decides the next intervention."}},{"@type":"Question","name":"During the test period, how many changes should I make at once to my Meta ads?","acceptedAnswer":{"@type":"Answer","text":"Make exactly one change. If the purchase-path check found a broken step, fix that step, such as showing total cost earlier or allowing guest checkout, and reallocate spend from the weakest ad set rather than adding budget. If the path works but the product still does not sell, test one proposition, like a clearer price, a bundle or a plainly worded guarantee, against the current version, keeping audiences and creative unchanged. If offer, landing page and audience all change together, the account may look different but the diagnosis is no clearer."}},{"@type":"Question","name":"How long should I run an A/B test before deciding?","acceptedAnswer":{"@type":"Answer","text":"Give the test at least seven days, and at low spend it may need 14 days just to produce a useful read. Testing guidance suggests about 100 events per version, and gaps under 25-50% at only 10-20 purchases per version can be chance. These are not quotas that justify buying more traffic: if the available volume cannot support a verdict within 21 days, record the test as inconclusive."}},{"@type":"Question","name":"What should I do at the end of the 21 days if ROAS looks good but profit doesn't?","acceptedAnswer":{"@type":"Answer","text":"A cleaner ROAS number does not excuse negative contribution. Calculate contribution after advertising by taking the product's net sales, less product and fulfilment costs, other selling costs and the ad spend supporting it, then put the result beside the allowable acquisition cost. If contribution stays negative after tracking and the purchase path are clean, pause acquisition for that product, because more scale is not the repair. A pause preserves money; a false winner spends it."}}]}]}
```
