You get $8k a month, 45 days, and one strategist to revive a stalled ecommerce Meta account. I would spend the first two weeks resisting the most tempting move: rebuilding it.

Most takeovers start with new audiences and a tidier campaign structure. I have watched that kind of cleanup reset learning while the inputs Meta actually reads stayed broken: a double-counted purchase event, out-of-stock catalog variants, and the same three tired ads doing all the spending. A neat account can still buy expensive orders.

Here is the sequence. Days 1 to 14 are for purchase-signal accuracy and feed hygiene, with the structure almost untouched. Days 15 to 30 put a limited testing budget behind new creative. Days 31 to 45 consolidate only what the evidence says to consolidate, then make a scale-or-hold call. No extra budget, no rebuild from scratch, no team to throw at it.

1. Days 1–3: Establish a baseline before making changes

Cost against the constraint: about one full strategist day and $0 diverted to new tests. Existing campaigns keep spending within the $8k monthly cap. What you buy is a baseline you can use to judge every later decision.

I pull the last 60 days by campaign and placement and check where the money actually went, not where the account structure suggests it should have gone. In a stalled store, I often find one Advantage+ sales campaign taking 70% of spend while prospecting starves. I used to rebuild that on sight. I was wrong. A rebuild can reset learning and bury the measurement fault underneath. The audit-first approach that holds structure while checking CAPI, event match, and deduplication looks slower than a new campaign map. It gives you a better reason for the next edit.

The first check is Purchase. If the pixel and Conversions API both fire it, Meta counts the order once when the event name and event ID match exactly, including case, within 48 hours. Miss that match and one order can look like two. ROAS looks 30 to 40% better than Shopify, and every bid decision built on that figure inherits the mistake.

This is not optional plumbing for ecommerce after iOS14. One ecommerce breakdown puts CAPI plus pixel at recovering 70 to 85% of lost signal versus pixel alone, with Purchase ranked first and Event Match Quality at 6 or higher. I verify deduplication in Events Manager, check that email and phone are passed for matching, and compare reported purchases with store orders. If those numbers disagree, I investigate before touching audiences.

What gets cut: the immediate restructure, the fresh interest stacks, and the satisfying but premature account cleanup. No new audiences until one purchase equals one purchase.

2. Days 4–14: Clean the catalog and keep the freeze

Cost against the constraint: about four strategist hours on days 4 to 7, with no media dollars diverted to a new test. The existing account continues to run. The work buys cleaner product inputs and fewer clicks sent to products a shopper cannot buy.

Catalog cleanup is unglamorous, which may be why it gets skipped. For Advantage+ catalog and dynamic ads, the feed does much of the creative work: descriptive titles with brand, type, and attribute; clean 1:1 or 4:5 images at a minimum of 1024×1024 pixels; prices that match the landing page; and availability that suppresses out-of-stock products. I check those fields against the product pages rather than assuming an approved feed is a useful feed.

I have seen stalled accounts send 15% of spend to variants that had been out of stock for weeks because no one set the availability rule. On an $8k month, that is $1,200 buying clicks to a dead product page. Fixing that does not require a new audience. It requires the catalog to stop making promises the store cannot keep.

The freeze stays in place through day 14: one scaling sales campaign, one testing control, and catalog sales where it already exists. I note obvious structural waste, but I do not rebuild around it yet. Split an approximately $267 daily budget across six fresh ad sets and each has less room to collect purchases. The new structure may look diversified; the learning is fragmented.

What gets cut: an Advantage+ versus manual test, a broad copy rewrite, and any audience rebuild whose main benefit is making the account easier to explain in a deck. If the feed lies, every audience looks stupid.

3. Days 15–30: Fund concepts, not a pile of ad files

Cost against the constraint: up to 30% of the existing budget for testing, not 30% on top of $8k. That is roughly $550 a week from an approximately $1,850 weekly budget. It buys enough room to judge a small number of genuinely different ideas.

Here is my planning math: new concepts per week equal weekly testing budget divided by target CPA times ten results. At a $35 target CPA, $550 divided by $350 gives me about 1.5 concepts. I round that to two new concepts a week, each with two or three hook variations. The variations are ways to test the idea, not a license to call six similar files six different concepts. Meta can group similar-looking ads rather than treat every file as a fresh idea.

That sounds thin beside advice to make five to ten creatives a week. Good. $8k cannot feed eight ideas properly at once. I would rather learn from two than starve ten. I keep them in one test ad set, separate from the proven ads so a new concept does not have to fight an established winner for every impression. The dedicated-testing approach caps that work at 30% of budget. For this takeover, the concepts come from existing product-page images and one creator clip.

What gets cut: a fourth angle, an influencer-whitelisting sprint, and a full batch of twelve user-generated videos. None helps if the account cannot spend enough to judge it. I also leave the sales campaign running while the tests collect evidence; the goal is not to replace known revenue with a creative workshop.

I write kill rules before launch because day 16 is when a weak ad starts looking like it might turn around. My sheet says:

  1. Graduate after 10 purchases at or under target CPA.
  2. Kill after spend reaches 1.5 times target CPA with zero purchases, or after 2,000 impressions with no add-to-cart signal.
  3. Scale a winner in 15 to 30% budget steps every two to three days, with no midday edits during learning. That stepwise approach gives each change time to show its effect.

Two concepts a week for two weeks gives me four judged bets by day 30. I am looking for one that can beat the control at a cost the store can afford, not a slide full of promising thumbnails. The weekly cycle can continue through day 45 if the budget still supports it. Test less, judge faster, and write the kill price down first.

4. Days 31–37: Consolidate only where the math demands it

Cost against the constraint: no additional media budget and no new campaign build. I spend strategist time reviewing purchases per ad set per week. What it buys is a decision about which splits are preventing the account from learning.

The common benchmark is about 50 optimization events per ad set per week to exit learning. Meta has also said ad sets under about 50 weekly conversions ran roughly 28% more expensive per purchase. That is a reason to take fragmentation seriously, not a promise that merging ad sets makes purchases appear.

At a $35 CPA, 50 purchases require about $1,750 of weekly spend. The whole account has roughly $1,850 a week. The 50-conversion ceiling math therefore leaves room for at most one ad set at that volume if nearly all spend goes to it. While the test campaign takes roughly $550 a week, even the main sales campaign may not reach that threshold. I do not pretend it will.

I merge prospecting splinters into the winning sales campaign and keep the testing control separate at its cap. The catalog campaign remains where it exists, but it does not get a fresh budget claim merely because its name sounds useful. If three interest ad sets each have 11 purchases and no distinct job, I cut that version of “diversification.” Fewer ad sets give the available purchases a better chance to inform a decision; they do not change the size of the budget.

What gets cut: structural changes made for appearances, and ad sets that split the same small pool of learning. One ad set with a chance to learn beats four that never do.

5. Days 38–45: Make the call in store numbers

Cost against the constraint: the same $8k monthly cap, with any increase to a winner funded by a decrease elsewhere. What it buys is a controlled decision, not a relaunch disguised as scaling.

On day 38, I look for seven straight days at target CPA in the consolidated sales campaign and one clear graduate from testing. If both are there, I move 70% of the test allocation toward the winner. I aim for a 20% increase to the sales campaign, funded from that transfer, and leave any amount that will not fit within the cap where it is. Then I wait three days. No midday edits. If CPA runs more than 20% above target for three days, I hold instead and fund another two-concept test cycle from the existing allocation.

The distinction matters: moving money is not adding money. Nor does a 20% raise guarantee that an ad set exits learning. At this budget, the test and the sales campaign compete for limited purchase volume. I would rather see that tradeoff plainly than claim I can fully fund every campaign at once.

On day 45, I open Shopify before Ads Manager. Then I check the ad account and the creative sheet. The scorecard has three parts:

  1. Blended return and acquisition cost. I compare all spend with GA4 or warehouse orders, and check new-customer ROAS through a post-purchase survey or holdout. Ads Manager ROAS can rise while the business stands still if retargeting claims demand it did not create.
  2. Contribution margin after ads. I calculate revenue minus cost of goods, shipping, the $8k in media, and management cost, then divide by revenue. If blended ROAS rises 15% but margin stays flat because discounts did the work, the plan did not work.
  3. Creative win rate. I count how many tested concepts beat the control at target CPA. If the two-concept weekly cycle has produced eight judged concepts by day 45, one winner suggests I am still guessing; two or three give me an angle worth feeding when the constraint lifts. If fewer concepts have enough purchases to judge, I report fewer. I do not turn an unfinished test into a win.

Those numbers also tell me whether to keep the manager. If blended performance, margin, and creative win rate move together, keep going. If only Ads Manager ROAS improves, change the manager, not the budget.

Who should skip this takeover

This plan will not work for everyone. I would skip it if the average order value is under $25 and a $35 acquisition cost is not supported by the store’s economics. A cleaner feed cannot repair that gap. I would also skip it if Meta drives fewer than 30 purchases a month. That volume cannot feed an ad set to 50 purchase events a week; the honest options are to use a higher-volume event such as add to cart where it helps, or hold spend until purchase volume builds.

A catalog maintained in spreadsheets, or product pages whose prices never match the feed, calls for a data build rather than this media takeover. And if the business needs a studio shooting twelve new videos a week, $8k is not funding that operation alongside two judged concepts a week. Change the budget or change the constraint. The same-account switch playbook keeps history intact while ownership changes, but history alone cannot supply missing data or creative.

6. When the constraint lifts, buy more learning—not another rebuild

The sequence changes only when the budget does. I keep the consolidated sales campaign that has learned, work the test share back toward four concepts a week, and add a second prospecting angle only when weekly budget, target CPA, and the 50-event benchmark can support it. The 45-day inherited-account guide makes the same case for recovery without an immediate budget increase. I agree with its order of operations.

This is also where I stop pretending one strategist checking in twice a week is enough. Feed checks, deduplication monitoring, creative judging, and budget steps need daily attention. groas runs Meta ads for a flat $1,499 a month, with audiences, budgets, and ads worked around the clock, changes logged, and one human owning the result. I would rather pay for the inputs Meta reads than for the rebuild that resets them.