Meta spend rose 8.3% to $21,480 at 1.14 back per dollar, and one tired ad is where most of the waste sits.

The account spent $21,480, up 8.3% on the week, and Meta credits it with $24,520 of sales, 1.14 per dollar. The winner is clear: the dentist-reacts UGC ad took $4,980 and returned 1.69. The problem is just as clear: the 15-second before-and-after cut spent $2,890 at 0.54, stops only 9.8% of the people it is shown to, and its click rate has fallen 41% from its own peak. Three ads returning under 0.6 hold $3,270 of the week. Move that money to the proven openers and to Reels, where the return is highest, before raising the total again.

$21,480 of spend, up 8.3% from $19,826, against $24,520 of sales Meta credits to its own ads: 1.14 back per dollar at a 1.4% click rate. Treat the return as an upper limit and check it against all revenue over all ad spend.

• Sales here are what Meta credits to its own ads (a click within 7 days or a view within 1 day). It is the most generous view — some buyers would have bought anyway — so treat the return figure as an upper limit, and check it against all revenue divided by all ad spend before believing it.
• Purchase numbers for the newest week keep truing up for about 7 days as Meta finishes attributing. Spend is stable; purchases are provisional.

Twelve ads spent money and the top five took 74.9% of it. The dentist-reacts UGC ad leads on every count: $4,980 spent, $8,420 credited, 168 purchases. The before-and-after cut is the third-largest spender and returns 0.54, which makes it the first ad to cut.

12 ads spent money this week; the top 5 took 74.9% of it.

• Return per ad uses Meta's own credited sales — an upper limit, not incremental truth.

The median video stops 33.0% of the people it is shown to, a good figure. The outlier is the before-and-after cut at 9.8%: it has taken $11,271 over four weeks with an opener that almost nobody stops for. Its first frame needs replacing, not its budget raising.

Median hook rate across 10 videos: 33.0% (good). Ladder: 40%+ excellent · 25–40% good · 15–25% average · under 15% means the first frame needs rethinking (practitioner ladder, CTC/Foxwell).

• Hook rate = of everyone shown the ad, the share who stopped and watched at least 3 seconds (Meta counts a video view at 3 seconds). Image ads are excluded — they have no watch signal.
• The benchmark ladder is a practitioner reference for headroom, not a grade; the account's own median is the fairer bar.

The median video keeps 25.0% of its viewers to the end. 'How it works 30s' is the one to fix: it opens well at 34.0% and then keeps only 11.0%, so the middle is losing the room. Cut it shorter or move the payoff earlier.

Median hold rate across 10 videos: 25.0% (good). Ladder: 35%+ excellent · 25–35% good · 15–25% average · under 15% weak (practitioner ladder, CTC), read alongside the account's own median.

• Hold rate = of the people who stopped to watch 3 seconds, the share who stayed to a ThruPlay — Meta's count of watching to the end, or at least 15 seconds.
• The retention curve reads: of everyone who started the video, the share still watching at a quarter, half, three quarters, and the end.
• A high hook with a low hold means the opener over-promises or the middle drags — fix the body of the ad, not the first frame.

Of 23,060 clicks, 16,670 reached a loaded page and 1,501 added to cart, the step that loses the most people you can win back. At its own four-week rate of 9.4% the cart step would have passed about 66 more people this week. The before-and-after cut has a separate problem: its clicks often do not load the page, so check where that ad sends people.

• Every count after the click is Meta's pixel view of your site — an attribution read, not the order system's truth. The funnel is multiplicative: fixing the single weakest step moves the whole thing most.
• The click step is left out of the biggest-leak pick on purpose. Only a small share of people shown any ad ever click, so that step would win every week and hide the step you can actually fix.
• Per-creative stages use 4 weeks of data and are judged against the account's own median for that stage — a stage is flagged only when it runs at less than 75% of the account's normal.
• The size-of-the-prize figure is a what-if, not a forecast: it re-runs this week's own traffic at that step's 4-week rate. It sizes what the step is costing you now; it does not promise what a fix delivers.

Women 25 to 44 carry the account: $9,690 of spend between the two groups at 1.35 and 1.44 back per dollar. Men return less, 0.83 at 25 to 34 and 0.77 at 35 to 44, so they are the first place to trim if budget must come down.

• Demographics come from the account-level feed — Meta does not provide them per creative at this grain.

Reels earns the most back per dollar: 1.37 on Instagram and 1.21 on Facebook. The Facebook feed takes the most money, $8,410, and returns 0.70, and Facebook stories returns 0.35. Shift budget toward Reels and test whether the feed spend is doing a reach job worth paying for.

• Return per surface is Meta's own credited sales for people who saw the ad there. Coverage is partial: some purchases reach Meta with no surface attached, so the surface rows can add up to fewer purchases than the account total. Read the surfaces against each other, not as a full account tally.

iPhone returns 1.16 per dollar against 0.74 on Android at similar spend. Before moving money, check the site on Android: a gap this size is often the checkout on that phone, and part of it is how much Meta can see on each.

• Device data exists only at the account level — Meta does not provide a per-creative device split, so 'which ad wins on iPhone' cannot be answered from this feed.
• Before moving budget between phone types, check the landing experience on the loser. A device gap this large is often the site on that device, not the audience on it — and iOS and Android also differ in how much of a purchase Meta is allowed to see, so part of any gap is tracking rather than sales.

California, Texas and New York take 61.0% of spend. A click cost between $0.58 and $0.67 in every state, within 8% of the $0.63 account average, so there is no state to cut on cost alone. Meta reports no sales by state, so this section cannot show a return.

• There is no return column here, and that is a data limit, not an oversight. Meta has never returned a single dollar of purchase value on the state-level breakdown for any of your brands: across every state row ever synced, more than 50,000 of them, the purchase value is zero — while the same accounts' age and device breakdowns do carry sales for the same weeks. A per-state return figure would therefore be invented, so this section shows what is real: where the money went, how well it clicked, and what a click cost.
• Cost per click is compared to this account's own average, so a state reading 1.4 costs 40% more per click than your typical state. Cheap clicks are not automatically better clicks — a state can be cheap because the people there are less likely to buy.

$2,890 of the week sits on one worn-out ad: the before-and-after cut, whose click rate fell from 1.9% to 1.1%, down 41.0% from its own peak. Rotate in a fresh angle rather than rebuilding an ad that has already had its run.

$2,890 of this week's spend is on ads whose click rate has slid 30%+ from their own peak.

• Two signals make fatigue: the audience keeps seeing the ad AND responds less each time. One bad week alone is weather; a slide from the ad's own peak while money still flows is the pattern worth acting on.
• The fix for a worn-out ad is rotation — a fresh angle or creator — not a redesign of an ad that already proved it can work.

$3,270, about 15% of spend, sits in three ads returning under 0.6. Take it back, put roughly 40% into what is proven, and let the weekly total ease toward $18,210 to $19,518 while returns are below break-even. The most-shared ad this week is also the best seller, which is the pattern to make more of.

• These checks distill what working practitioners look at (Foxwell, Motion, Common Thread, from the scraped knowledge base) into fixed rules run on your numbers. They are references for headroom, never grades.
• Shares and saves move before purchases do — an ad people send to friends this week is often next month's winner. A leading signal, not a guarantee.