LinkedIn
Most Meta ads proposals lead with a rebuild: new campaigns, new audiences, new naming, all in week one. I used to sell that work. I was wrong. This list is for ecommerce owners spending, say, $5k to $50k a month on Meta who want to judge a Meta ads management company by what it changes first. The managers worth paying start with purchase tracking, the product feed, creative and the offer. Several of the urgent rebuilds on a proposal just reset learning you already paid for.
1. Fix the purchase event and Conversions API before touching a campaign
Bad signal makes every later decision worse. If Meta receives duplicate purchases, misses orders or struggles to match events, bidding learns from a distorted picture. Reported ROAS becomes harder to trust too. In Shopify’s Facebook channel, set data sharing to Maximum to send both pixel and Conversions API server events. Then verify that the events deduplicate, and check purchase Event Match Quality in Events Manager. I want the purchase signal working before I change budgets or structure.
Ask for a screenshot showing purchase event counts against Shopify orders for the same seven days, plus the match-quality score. The counts need explaining, not merely presenting: two systems can count differently without either being broken. But if the manager pitches new audiences while promising to fix tracking “later,” you have your filter. Tracking is the work. Everything else is an interpretation of it.
2. Clean the product catalogue before buying more traffic
Meta cannot sell what it cannot read. I have seen stores with 400 SKUs where 90 were out of stock, 40 had supplier titles like “ACME BLUE-7,” and the hero product had an image that looked soft on Instagram. Advantage+ could still spend on those products. That was the problem.
Start with the products that carry the business. Lead titles with product type, use clear images and keep stock synced. Use custom_label_0-4 to distinguish heroes, high-margin products and clearance so you can see which products deserve budget, rather than treating every SKU as equal.
Verify the work in Commerce Manager, not a slide. Ask for the disapproval rate, the out-of-stock count and one before-and-after title. If the best seller still opens with an unhelpful supplier code, the catalogue work is not done.
3. Consolidate ad sets until each has enough purchase volume
Most inherited accounts I audit have too many ad sets spending too little. Nine ad sets at $30 a day can look like a carefully segmented strategy. Often, none gets enough purchases to learn much. Meta’s commonly used benchmark is 50 optimisation events per ad set in a rolling seven days. Split a store’s 60 weekly purchases across five similarly funded ad sets and each gets a thin slice of signal.
Fewer ad sets concentrate that signal. The first step is to ask how many prospecting ad sets will run next week and how many weekly purchases each is expected to receive. If volume is limited, try one prospecting campaign with no more than two ad sets before multiplying the structure. Do not buy more segmentation than your purchase volume can feed. If Purchase remains too sparse, discuss a higher-volume optimisation event as a deliberate tradeoff, not an automatic fix: an add-to-cart signal is not a sale.
4. Ship new creative every week
Structure cannot rescue an ad people are tired of seeing. When the same three videos keep circulating, frequency rises and click-through rate can slide. A bid tweak does not give shoppers a new reason to care.
Treat creative as inventory. Keep a varied set of ads in rotation and reserve budget for testing new work, with the volume scaled to what your account can actually spend and evaluate. Ask for the launch dates of the last five ads, what made each different and the rule for cutting a loser. If everything live is older than 30 days, the first problem to investigate is production, not targeting.
In the first 60 days, I want a repeatable rhythm: one new concept each week, two or three hook variants, a clear read on which version earns more spend, then another concept. Creative volume is the budget multiplier; structure is the plumbing. Your first concrete step is to put the next concept and its launch date on the calendar.
5. Separate new-customer performance from returning-customer performance
Meta is good at finding people likely to buy, including people who already know you. Blend prospecting and retargeting into one ROAS figure and returning buyers can hide weak cold traffic. Advantage+ may favour easier conversions, so you need to see what the campaign is actually delivering.
One analysis of incrementality experiments found that manual campaigns delivered higher incremental ROAS than Advantage+ for 58% of the brands studied when new and existing customers were separated. That is not a reason to switch off Advantage+. It is a reason to read its results honestly.
Start with Meta’s Audience Segments breakdown for New, Engaged and Existing. Then compare the trend with Shopify first-time buyer revenue over the same period. The figures measure different things, but they should prompt a conversation when they diverge. If attributed ROAS climbs while new-customer revenue stays flat, ask who the extra spend is reaching. A good manager reports both views without waiting to be asked.
6. Set ROAS floors by product margin
A single 3x target across a store with mixed margins is how you lose money politely. As a starting point, break-even ROAS is one divided by gross margin: a 20% margin needs 5x to cover product cost and ad spend; 30% needs about 3.3x; 40% needs 2.5x. Other costs can push the required return higher. At 25% gross margin, a celebrated 3x return does not clear even that basic hurdle.
I used to tell clients one account target kept things simple. I was wrong. Simple hid the losers. Pull margins by hero product group from Shopify, set a floor for each group, and keep clearance from quietly consuming the budget meant for healthier products. Then compare those floors with store revenue and total ad spend, not just Meta-attributed sales. A target without margin behind it is a vanity metric with a budget.
7. Match the landing page to the ad’s promise
The ad promises a 3-pack bundle for $59. The click lands on a collection page, and the shopper has to hunt for it. That is a page problem, not an audience problem.
I check this by tapping my own ads on a phone. If the product, price, image and offer on the first screen do not match what I just saw, I fix the path before changing the campaign. Start with one dedicated landing view for a hero ad angle: same product, same offer, same photo. Make the next click feel like the next step, not a new search.
This is also why I care what “management” includes. For stores using groas Meta management, audiences, budgets and ads are worked around the clock, with new ads tested and winners scaled. That work matters more when the destination keeps the promise the ad made.
8. Test the offer, not just the image
Owners want to test backgrounds and fonts. Buyers respond to the deal. A free-shipping threshold, bundle price, first-order discount or free gift can change the decision at checkout in a way another colour treatment cannot. Creative gets attention; the offer gives someone a reason to pay.
Run one variable at a time. Keep the ad and landing-page layout consistent, change the offer line, and read Shopify conversion rate and revenue per session rather than celebrating a higher CTR. I would start with one offer against the current control and give each a defined test window. If paid conversion rate rises from 2.1% to 2.6% at the same average order value and spend, that is roughly 24% more return from the same traffic. It is an illustration, not a result to promise.
Ask your manager which offer they would test first and what must stay constant for the comparison to mean anything. If they have never proposed an offer test, you may have hired a media buyer when you needed someone to examine the sale.
9. Report store revenue alongside Meta-attributed ROAS
Meta and Shopify do not count sales the same way. With a seven-day click and one-day view attribution setting, Meta can credit purchases after an ad view or click; Shopify records the store order. View-through, modelled and cross-device conversions can widen the gap. Neither dashboard turns the other into a useless one. They answer different questions.
Use Meta attribution to understand and optimise campaigns. Use Shopify to check what the business collected. Ask for one weekly line showing Shopify revenue divided by total Meta spend, often called MER or blended ROAS, alongside new-customer revenue and Meta-attributed results. Blended ROAS is a store-level health check, not proof that Meta caused every order. That distinction prevents a manager from scaling solely because a platform number looks good while the store does not.
10. Leave a full account rebuild alone in week one
The second half of the interview is what a manager refuses to do. A rebuild looks like progress because you can see it in Ads Manager. It can also burn the learning your account already paid for.
New targeting, creative, optimisation events or bid strategies can send affected ad sets back into learning. Change everything at once and you cannot tell which edit helped or hurt. A new manager who renames the account and launches five fresh campaigns on day three may be selling visible activity instead of diagnosis.
Leave working structure alone until you know what is broken. First, ask for the tracking, feed and creative findings that justify any proposed rebuild. Then change one consequential thing at a time, with a reason and a way to judge it. “This is how we structure every account” is not a reason.
11. Leave daily budget swings off working campaigns
Taking a winner from $100 to $1,000 overnight is not the same as nudging it from $100 to $101. Large changes can disrupt learning and delivery. Repeated 30% moves up and down make the resulting volatility look like a property of Meta when it may be a property of management.
I prefer planned changes followed by time to read the result. Set a test budget, check whether the ad set is getting useful conversion volume, then step up rather than jerking the spend around daily. If a proposal promises aggressive daily scaling, ask how it protects the campaigns already working. The first step is simple: request the budget-change rule in writing. A good answer names when to increase spend, when to pause and what evidence earns either decision.
12. Leave Advantage+ running long enough to judge it
Advantage+ can look ugly on day two. Killing it immediately means you have learned little and must pay to start the test again. Give a new campaign a defined evaluation window and enough purchase data to make that window useful. Seven days alone is not a magic verdict if there have barely been any orders.
There is a more useful question to ask first: how much of its spend and reported return comes from existing customers? Review the audience breakdown, keep customer lists fresh and check the exclusions and spend controls available to your account. Recent buyers who are not identified correctly can make prospecting look stronger than it is.
Do not kill the campaign to avoid reading the customer mix. First step: ask your manager to show new versus existing customer results and explain what they will change if the balance is wrong. If the economics still fail after a fair test, then make the call.
13. Leave extra audiences out until there is a reason for them
More audiences are the oldest way to look busy. An interest stack, a lookalike and a broad-versus-stacked test can split the same $200 a day into slices too thin to learn from. I used to build sprawling audience grids. They photographed well in reports and performed nowhere else.
Ask which audience the manager removed last month and why. If the answer is “none, we only added,” ask what each addition is meant to prove and whether the account has enough purchases to prove it. A good manager subtracts complexity before charging you to manage more of it. That applies beyond audiences: send this list before your next agency call and ask for its first-60-day order. This managed-service checklist offers another way to test whether the work matches the fee.
If I could start with only one entry, though, I would start with number 1. Fix the purchase event and Conversions API first. Clean signal gives every later creative, offer and budget test a fairer read; dirty signal makes even a good-looking win hard to trust. That is why I would choose groas’s always-on, flat-fee approach over paying for a week-one agency rebuild. Start where the money is counted.