Why the ad money isn't coming back

Roughly forty percent of this budget was aimed at engagement rather than sales. That was a deliberate choice by the agency running the account, to test a market that does not yet have a name, and it is a reasonable thing to have done. This report treats it as context rather than as an error.

What it does mean is that a lot of the spend cannot be judged on sales. So this page starts by separating what the numbers can settle from what they cannot, and then spends most of its length on the part that is not in dispute: the advertising itself.

Every ad set on Meta is given one instruction: what outcome to go and get. That instruction is what decides who sees the ad. Ask for post engagement and Meta hunts for people likely to tap a heart. Ask for video views and it finds people who watch videos. Ask for a purchase and it goes looking for people it believes are about to spend money.

Your budget was split across four of those instructions.

The lower three, $1,900.66 or 39 percent of the budget, were never asked to find a buyer. We understand this was intentional: a top-of-funnel test to find out whether there is an audience for a product in a category that barely exists yet. Your search data supports the reasoning. Almost nobody types "guitar string protection case", so there is no existing demand to capture, and paying to find out who reacts is a defensible first move.

Being straight about the limits of this matters more than the findings themselves, so here they are.

Those last two are worth a moment, because they are the difference between a test and an ongoing cost. A test runs, concludes, and changes what happens next. This one has been running for six months.

March and April read exactly like a test phase: everything aimed at engagement, nothing aimed at sales. From May the sales campaigns started, and the engagement spend carried on beside them rather than tapering. Post engagement and video spend specifically has trended up rather than down over that period, from $98 in March to $297 in August, its highest month on record. Of the $1,922 total, $625 falls in the March to April test window and $1,298 was spent after the sales campaigns were already live.

We also checked the account for structured tests, the kind Meta sets up and measures for you, and for lift studies. There are none, and there never have been. That does not mean nothing was learned. It means the learning is not written down anywhere in the account, so the only place it exists is with the people who ran it.

There is one more thing to settle here, because it bears on whether that metric can be trusted at all.

First, the geography question, closed. Every one of the fifteen ad sets is set to the United States and nothing else, with no language restriction. The delivery matches: 99.4 percent of the money, $4,844.88, was served to people in the United States, across 119 US regions led by California, Texas, Florida and New York. Brazil received nothing. The only non-US delivery in the account's history is $28.32 in Vietnam, on a campaign paused last December. Every ad's copy is in English. The ads are aimed where you want them aimed.

Your Instagram following is a different story. The profile has 885 followers, and they break down like this.

That is not a targeting failure, and it is worth being precise about why. Of everyone the account reached in the last few days, roughly nine in ten were American and fewer than four in a hundred were Brazilian, which is the paid delivery doing exactly what it was told. But of the people who then engaged, a quarter were Brazilian. A Brazilian who sees your content is roughly ten times more likely to react to it than an American who sees it. Instagram shows content to whoever engages, so a small, highly active pocket compounds: nine months of that produces a following that is majority Brazilian while the budget buys American attention.

The follower split is a lifetime figure across 885 people and is solid. The ten-times gap is measured over a short recent window and should be read as a direction, not a precise multiple.

One thing that has not gone wrong and should stay that way: all four of your saved remarketing audiences are built from website visitors, not from social engagement. That keeps them clean. An audience built from people who like the posts, or a lookalike modelled on them, would be seeded largely on Brazil.

One number that survives the test framing intact, because it belongs to the other half of the account: the ad sets that were aimed at a sale spent $2,920.39 and returned 10 sales. That is $292 per sale on a $59.99 product, and section 07 covers why.

There are 63 ads in the account's history, carrying 44 distinct images between the 60 whose artwork Meta will still hand back. Three of the 63 were built as advertisements, with a written headline, body copy and a destination. The other 60 are existing social posts that were promoted.

The three purpose-built ads returned four of the eleven sales on a quarter of the money, at $310 per sale against the account's $438 blended average. They are the best-performing creative you have ever run.

All three are switched off. They sat inside the campaign that was paused in January, and every one of the 30 ads delivering today, carrying $1,822 of spend between them, is a promoted post.

A promoted post is not automatically a bad ad, and some of these did sell. But a post is written to be read by people who already follow you. An ad has to work on someone who has never heard of you, in about a second and a half, and it needs a headline and a destination to do that.

Video is where most of the original creative effort went. Eleven distinct concepts were produced and run. Here is every one of them, in the order of how well the idea is aimed, not how much it was spent on.

The first thing to say is that the ideas are not the problem. The first three carry the strongest angles in the entire account. One opens with "Your guitar strings are dying faster than you think", which is the problem your customers actually have and the language they actually search. One is a piece to camera about sending six guitars to a corrosion testing lab, which is real proof most competitors could not produce. One breaks the product down into its three layers, which is the closest thing here to explaining what $59.99 buys.

Then look at where the money went.

The two strongest ideas, the dying-strings hook and the corrosion lab, received $141 between them. Twelve percent of the video budget. The single largest share, $463.91, went to the unboxing collage carrying a message announcing a launch that had already happened.

That inversion is the finding. It is not that nobody in this account can write a good hook. Somebody clearly could. It is that the good hooks were never funded enough to prove themselves, and the money followed something else.

The second thing is harder to say plainly, so here it is plainly. Laid side by side, these do not look like eleven ads from one company. There are at least seven different display typefaces across them. Several carry heavy outer glows and drop shadows, which is the visual signature of a template rather than a design. The colour treatment swings from flat pink illustration to black-and-white lightning to purple smoke to warm domestic wood. There is no recurring logo position, no consistent type treatment and no palette. A guitarist who saw four of these in a week would have no reason to connect them to each other.

This is the same finding the store audit made about your homepage, appearing again in a different place. The absence there was a design system. The absence here is the same one, and it costs more on Meta than it does on the site, because on Meta repetition is the only thing that builds recognition and nothing here repeats.

Four of the images carrying real spend are shown here at the size they were served. Meta only makes full-resolution files available for a minority of the account's creative, so this is not the whole set; it is the part that can be inspected honestly.

$644.40 · highest-spending adThe product appears as seven thin vertical strips on a synthetic stadium backdrop. At the size this runs in a feed, it is not clear what the object is or what it is for. There is no price, no offer and no text in the image.

$422.74 · second-highestA phone photograph of a domestic living room. The product is competing with a Christmas tree, a curtain, a gift box and eight guitars. Shot in portrait, so the feed crops it further.

$172.56 · thirdThe cleanest composition of the three, and the one where the product reads as a plain dark strap. Nothing in the frame says what it does or why it costs $59.99.

The one that is closest to workingClean background, product legible, a headline in the frame. It asks “What's String Protech?”, which requires the viewer to already care about the brand name. The same image asking “Why do your strings die so fast?” would be doing the same job against a problem people have.

The three purpose-built ads all carry the same headline. It contains a grammatical error, and it sells the wrong thing.

Two things are happening there. The first is a typo in the highest-spending headline in the account. The second matters more: the ad leads with decoration and treats protection as a bonus. Your own site analytics point the other way. The page that holds a stranger longest is the one about strings going dead, and the search terms guitarists actually type are about rust and dead strings, not about customising a guitar's appearance. The ad is selling the feature the market is not asking for and burying the one it is.

A second piece of copy ran twice, in April and again in June, for $463.91 combined.

By June the product had been on sale for months. Money was still going into building anticipation for a launch that had already happened, pointed at a profile rather than a shop.

Set the engagement spend aside. The $2,920.39 that was aimed at a sale did the right things: 81 percent of its clicks landed on the site, 66 carts, 23 checkouts, 10 sales. It also cost $292 per sale on a $59.99 product.

The reason is volume, and it is structural rather than a matter of patience. Meta's system improves by watching conversions happen; its own published guidance is that an ad set needs somewhere around 50 purchases a week before it has enough signal to settle into efficient delivery. Across the last 90 days, your best-performing sales ad set recorded two purchases in total. Not two a week. Two.

Four remarketing ad sets ran at the same time against overlapping slices of the same pool of site visitors: everyone from the last 30 days, the last 14, the last 7, and the 7-to-14-day window. A 30-day list already contains the 14-day and 7-day lists inside it. They were competing with each other for the same people.

$455.71 was spent showing the same message to 1,005 people an average of eighteen times each. It produced one sale. A frequency that high on an audience that small has one meaning: there was nobody new left to reach, so the system kept going back to the same handful of people. Meta's own recommendation tool flags a related problem elsewhere in the account, on a separate pair of ad sets it says are splitting the same audience with the same creative.

1. $823.97 pointed American attention at a Brazilian audience. Six ads carried a button that opens your Instagram profile rather than a page on your site. They drew 6,791 clicks, more than the other fifty-seven ads combined, and only 306 of those people reached your website. Under a profile-growth objective that is coherent. It is less coherent given section 02: the profile those American guitarists arrived at has a following that is 56 percent Brazilian, so the largest single block of attention this account ever bought was spent sending buyers to the weakest piece of social proof you own.
2. One live ad cannot run, and probably should not have been stopped. An ad in your sales campaign sits disapproved, and Meta's stated reason is its Drugs and Pharmaceuticals policy. On an ad for a guitar accessory that reads as a misclassification rather than a real violation, which means it is appealable. Its status in the ordinary list still reads active, so it looks like part of your working inventory. Nobody has appealed it.
3. Spend is concentrated without a reason. The single top ad absorbed 13 percent of everything spent on Meta; the top five took 47 percent. At the other end, 17 ads never cleared $5 in their lifetime, which reads as variants switched off within hours rather than tests allowed to run.

Nothing here says paid social cannot work for this product, and nothing here says the decision to test the market with engagement campaigns was wrong. It says that on the $4,821 spent so far, nobody can tell you whether it works, because almost none of that money was spent in a way that would produce an answer.

The encouraging half of that list is not a consolation prize. The dying-strings hook, the corrosion lab footage and the three-layer explainer are the hard part of advertising, and they already exist, sitting at the bottom of the spend table. What is missing around them is everything that turns a good idea into a result: a consistent look so that four exposures feel like one brand, a headline and a destination on every ad, and any mechanism at all for telling which idea is working.

And underneath all of it sits the arithmetic from the earlier report. At $59.99 a unit, a cost per sale anywhere near today's has no version that works, whoever is running the account. Better creative moves that number. It does not on its own solve it.

Every figure on this page was read directly from your Meta account between 1 and 3 September 2026, and each headline number was checked by at least two different queries. Nothing in the account was created, changed, paused or deleted. One note on precision: Meta's own spend figures move slightly through the day as delivery is reconciled, so treat these as accurate to the dollar rather than the cent.