We analysed 1,404,808 products across 130+ stores: 92% never sold once in 13 months
A Google Shopping study: 130+ stores, 1.4M products, 17M clicks, 13 months. How much budget goes nowhere, how long a bestseller lives, and when to switch a product off.
There’s a question almost no store owner can answer straight away: what share of your advertising budget over the past year went to products that brought in no sales at all?
Not “what’s your ROAS” — everyone knows that number. Specifically the share of budget spent for nothing, product by product.
We decided to calculate it not on a single store but on a large sample: 130+ stores, 1,404,808 unique products, around 17 million clicks and $3.2M of advertising budget over 13 months.
A warning up front, so as not to waste your time. This is long tables and percentages, not “10 life hacks”. If you run product advertising — Shopping or Performance Max — read on regardless of catalogue size: the same patterns hold at 300 products and at 200,000. Below you’ll almost certainly find your store in one of the rows.
🔍 Compare your catalogue against these numbers
Every number below can be viewed for your own store. Connecting a Google Ads account takes a couple of minutes, and access is read-only: we change nothing in the account. You then see your own distribution — how many products sell, how many eat budget for nothing, how many get no impressions at all. Free.
What the data is
| Parameter | Value |
|---|---|
| Stores | 130+ |
| Unique products | 1,404,808 |
| Products with advertising budget | 632,826 (45.0%) |
| Products with conversions | 111,529 (17.6% of those advertised) |
| Advertising budget for the period | ~$3.2M |
| Clicks | ~17M |
| Campaigns per store | 28 (median) |
| Period | February 2025 — February 2026 (13 months) |
| Countries and currencies | CZ, UA, PL, EU — CZK, UAH, EUR, USD, PLN |
Why 13 months exactly. It’s a full year plus a month. Such a period covers the entire seasonal cycle — both the November–December peak and the February trough — while still allowing a like-for-like month-on-month comparison year over year. A shorter window would show a seasonal spike and pass it off as a trend. And one more thing: if a product hasn’t sold once in over a year of advertising, that isn’t “not enough data” — that’s an answer.
Benchmarks for the sample itself, so you know who you’re comparing yourself with:
| Store metric | Bottom quartile | Median | Top quartile | Mean |
|---|---|---|---|---|
| ROAS | 244% | 420% | 666% | 568% |
| CTR | — | 1.64% | — | 1.77% |
| CVR | — | 2.29% | — | 3.65% |
How to read this table
Median is the middle of the sample: half the stores have ROAS above 420%, half below. Bottom quartile is the line below which the weakest 25% of stores sit: if your ROAS is under 244%, you’re in the bottom quartile. Top quartile is the mirror image: ROAS above 666% means you’re in the best 25%.
Note that the mean (568%) is noticeably higher than the median (420%). A few stores with very high ROAS pull it upwards. So compare yourself against the median — the mean is misleading here.
These are ordinary stores: median ROAS 420%, median CTR 1.64%. Not a showcase of success, and not a collection of failures.
First — two mistakes we made in the calculations
Let me start not with the conclusions but with where we got it wrong. Because if you calculate the same things for yourself, you’ll step on exactly these rakes.
Mistake one: we looked for a hard switch-off threshold. There isn’t one
The hypothesis was simple and very convenient: there must be a spend figure beyond which a product with no conversions can safely be switched off. Spent $50 and zero sales — scrap it.
We tested this on 514,602 products with spend. Here’s what came out:
| Spent on product | Products | Didn’t convert in 13 months | Accuracy of the “switch off” decision |
|---|---|---|---|
| $5+ | 294,669 | 259,582 | 88.1% |
| $20+ | 140,512 | 117,494 | 83.6% |
| $50+ | 63,171 | 50,031 | 79.2% |
| $100+ | 28,382 | 21,413 | 75.4% |
| $200+ | 10,444 | 7,463 | 71.5% |
| $500+ | 2,088 | 1,383 | 66.2% |
This is where it gets interesting. Accuracy doesn’t grow with the amount — it falls and plateaus at 66–69%. A product that has burned $500 without a single sale can be switched off by mistake in a third of cases.
The reason shows up in another table: 35.4% of products in “loser” status did convert in subsequent months, with a median of 2 months. Among “zombies”, 10.6% recovered.
What this means in practice
A rule like “no sales after N spent — switch it off” isn’t supported by 13 months of data. What works isn’t switching off but reallocation: reduce the share of budget going to products with no return, but don’t zero them out completely — otherwise you’re guaranteed to lose a third of your future conversions.
Mistake two: we thought it was about price
A logical hypothesis: products don’t sell because they’re more expensive than the market. We matched Merchant Center price competitiveness data against performance — 213,913 products across 51 accounts matched.
| Price positioning | Products | Share with conversions | Average ROAS |
|---|---|---|---|
| Much cheaper than market (>20% below) | 22,948 | 20.8% | 552% |
| Cheaper (5–20% below) | 40,619 | 21.5% | 863% |
| At market (±5%) | 67,377 | 20.4% | 855% |
| More expensive (5–20% above) | 48,435 | 14.7% | 801% |
| Much more expensive (>20% above) | 34,534 | 11.7% | 674% |
There is a difference: products priced 20%+ above market convert almost half as often. But the correlation between price deviation and product status came out at r = −0.075 — statistically almost nothing. What’s more: among the best-performing products, 38% are cheaper than market and 37.4% are exactly at market. So “being cheaper” guarantees nothing in itself.
Price explains a small part of the picture. The main explanation turned out to be elsewhere.
The five-layer method: how we broke the catalogue down
Instead of looking at an account’s average ROAS, we broke each catalogue into five layers. This is our working classification, which we call the five-layer method — by the first letters of the statuses, also known as WPLZN.
| Layer | Criterion | What it means |
|---|---|---|
| Star ⭐ | has conversions, ROAS ≥ account benchmark | Carries the store |
| Potential | has conversions, ROAS < benchmark | Sells, but inefficiently |
| Loser | spend ≥ one cost per conversion, 0 sales | Spent enough to buy a conversion, but didn’t |
| Zombie 🧟 | spend < cost per conversion, 0 sales | Nibbles away; too early to conclude |
| Invisible 👻 | spend = 0 | Advertising never showed it at all |
The key difference from an ordinary report: Invisible products aren’t “bad products”. They’re products you have no data on at all, because the algorithm never gave them a chance. They can’t be judged on performance — they were never tested.
Here’s what the catalogue looks like across the whole sample:
- 7.9% sold (111,529 products) — they took 78.1% of spend
- 37.1% ate budget without sales (521,372) — they took 21.9% of spend
- 55.0% received no budget at all (771,907)
More than half of a typical store’s catalogue simply isn’t covered by advertising. Another third receives money and returns nothing. And 8% of products carry all the rest.
Five findings the whole exercise was for
1. Budget concentration is sharper than the Pareto principle
The “80/20” rule doesn’t apply here — the distribution is considerably harsher.
| Product group | Share of spend (globally) | Share of spend (median across stores) |
|---|---|---|
| Top 1% of products | 48.3% | 27.7% |
| Top 5% | 72.3% | 54.0% |
| Top 10% | 82.1% | 68.1% |
| Top 20% | 90.5% | — |
In a typical store 1% of products takes a quarter of the budget, and 10% of products two thirds. The remaining 90% of the catalogue share a third between them.
With revenue the picture mirrors this and is no less alarming: a single top product brings in an average of 9.3% of all revenue, the top 5 products 22.7%, the top 10 31.3%. A third of turnover rests on ten items.
2. A bestseller lives three months
This is probably the most underrated number in the whole study. We traced how many months a product holds winner status:
- 64.7% of winners are “one-offs”: one conversion in one month and that’s it
- 3 months — median lifespan of a stable winner (≥2 months of sales)
- 71% of bestsellers were in the top for no more than 3 months out of 13
- 38% — the share of active months in which a bestseller was actually a bestseller
Here’s what follows from this. If you picked “our best products” once, built a campaign structure around them and left it for six months — by the third month you’re advertising yesterday’s winners. The data says the composition of the top changes faster than most stores review their structure.
And one more counterintuitive observation: we found no pronounced seasonality in the composition of winners. First half ≈ second half; the November–December peaks add only +12–25% to the average. So “let’s wait for the season, it’ll sort itself out” isn’t a strategy.
3. How many clicks a product needs before its first sale
Here we built a survival curve: if a product has received N clicks and hasn’t sold yet, what’s the probability it sells later?
| Clicks received | Cost of the test | Probability it sells later | Share of future buyers already “caught” |
|---|---|---|---|
| 1 | $0.20 | 14.4% | 6.2% |
| 3 | $0.60 | 25.6% | 16.9% |
| 5 | $1.00 | 33.2% | 25.2% |
| 10 | $2.00 | 46.4% | 39.8% |
| 20 | $4.00 | 61.4% | 56.9% |
| 50 | $10.00 | 79.7% | 78.0% |
Average CPC across the whole sample is $0.20 (converted into dollars across all currencies).
Here’s what the table means. A product that has received 1–4 clicks — you know nothing about it, there simply are no statistics. A product with 20 clicks and no sales is already a signal, but 39% of those will still sell. And even at 50 clicks, one in five converts later.
The practical conclusion: the minimum honest test of a product is 15–20 clicks, i.e. roughly $3–4. Any less and you’re drawing conclusions from noise.
By category the threshold differs noticeably:
| Category | Min. test, clicks | Min. budget | Share of products with sales |
|---|---|---|---|
| Bags and luggage | 15 | $3 | 27.9% |
| Apparel and accessories | 20 | $4 | 23.0% |
| Home and garden | 20 | $4 | 18.1% |
| Electronics | 20 | $4 | 14.6% |
| Health and beauty | 30 | $6 | 29.6% |
| Furniture | 30 | $6 | 9.4% |
| Car parts | 30 | $6 | 8.7% |
| Hobbies and crafts | 50 | $10 | 18.8% |
If you sell furniture and switch a product off after 10 clicks, you’re switching it off blind. If you sell bags and wait for 50 clicks, you’re overpaying for an answer that was already obvious.
4. Half of “invisible” products stay invisible
We took 1,196,357 products that started the period with zero spend and looked at what happened to them over 13 months:
| What happened | Products | Share |
|---|---|---|
| Never received a single impression in 13 months | 886,245 | 74.1% |
| Received budget, but no sales | 279,106 | 23.3% |
| Received at least one conversion | 31,006 | 2.59% |
Three quarters of the “sleeping” catalogue never got a chance in a year. Meanwhile the products that did wake up took a median of 4 months to their first sale — and more than half of them passed through a “getting clicks but not selling” stage first.
So a product’s route to a sale usually looks like this: invisible → gets clicks without conversions → sells. If you switch products off at the middle stage, you’re cutting that route off systematically.
5. The main difference between strong and weak stores
We split the sample into the top and bottom quartile by ROAS and compared not bids and not creatives but the structure of budget distribution.
| Metric (median) | Top 25% | Bottom 25% | Difference |
|---|---|---|---|
| Share of budget on products with no sales | 39.8% | 71.8% | −32 pp |
| Share of budget on selling products | 60.3% | 28.2% | +32 pp |
| Click efficiency | 55.1% | 26.8% | +28 pp |
| Catalogue size | 19,375 | 2,774 | ×7 |
| Top 5 products = share of revenue | 14.3% | 24.4% | −10 pp |
There it is, the answer to the question at the start of the article. The difference between a strong and a weak store isn’t the cost per click or the quality of the ads. It’s that in weak stores 72% of budget goes to products that don’t sell, and in strong ones 40%.
Note the reverse as well: in weak stores revenue depends more heavily on a few items (the top 5 give 24.4% against 14.3%). Fewer working products means a higher risk that losing one supplier or one out-of-stock will sink the month.
40.5% — the median share of budget going to products with not a single sale in 13 months. The highest value observed was 72%
On a budget of 100,000 UAH a month that’s 40,000 UAH going to products that produced no sales at all in 13 months. Not “spent inefficiently” — with no result whatsoever.
📊 Find out your own loss share
The sample median is 40.5%. Your figure could be 12%, or it could be 70% — overall ROAS won’t show it, it has to be calculated separately. Connect your account and it will be calculated from your data automatically.
How to apply this yourself: four steps
Step 1. Break your catalogue into five layers
Export the product report for 12 months and calculate three numbers: how many products produced at least one conversion, how many incurred spend without a single sale, and how many incurred no spend at all.
Typical mistake: looking only at products with spend. Then the largest layer — the one advertising gives no impressions to at all — stays invisible, and that’s more than half the catalogue.
Step 2. Calculate your loss share and compare it against 40.5%
The share of budget going to products with no conversions over the period is a single number that says more about account health than ROAS does. The sample median is 40.5%; the top quartile 39.8%, the bottom 71.8%.
Typical mistake: comparing yourself with “the market average ROAS”. Average ROAS depends on margin and category; loss share depends on how well the catalogue is managed.
Step 3. Set a test threshold for your category, not a blanket one
15–20 clicks for bags and apparel, 30 for furniture and car parts, 50 for hobby products. At a CPC of $0.20 that’s $3–10 per product — the price at which you buy knowledge instead of a guess.
Typical mistake: a single threshold for the whole catalogue. In categories with a long decision cycle it switches products off before they’ve had a chance to show themselves.
Step 4. Reallocate rather than switch off — and do it more often
35.4% of “losers” convert later, with a median of 2 months. Switching them off entirely guarantees you cut off that third. Reducing their budget share doesn’t. And since the median lifespan of a winner is 3 months, reviewing the distribution once a quarter is already too late.
Typical mistake: “set it up and leave it”. Three months on, the campaign structure describes a catalogue that no longer exists.
Benchmarks: find your store in the table
By average order value
Comparing yourself with the whole sample is of little use — a store with a $9 order value and one with a $263 order value live by different laws. Breakdown across 128 stores:
| Segment | Stores | Median AOV | ROAS | Loss share | Products with sales | Clicks to purchase |
|---|---|---|---|---|---|---|
| Low AOV ($0–15) | 20 | $9 | 300% | 32.6% | 32.2% | 34 |
| Mid AOV ($15–50) | 67 | $29 | 444% | 39.1% | 22.8% | 27 |
| High AOV ($50–200) | 34 | $75 | 418% | 53.4% | 16.6% | 22 |
| Premium ($200+) | 7 | $263 | 1118% | 48.0% | 13.7% | 31 |
Two observations. First: loss share grows with order value — 32.6% in the low segment against 53.4% in the high one. Second: the share of products that sell at all falls in the opposite direction — from 32.2% to 13.7%. Expensive products have a longer path to a decision, and budget has time to spread across a larger number of non-converting items.
Caveat: the premium segment is only 7 stores. A ROAS of 1118% was observed there, but you can’t draw a conclusion about the whole premium tier from seven data points.
By category
| Category | Products | ROAS | Loss share | CVR |
|---|---|---|---|---|
| Home and garden | 296,016 | 614% | 24.5% | 3.11% |
| Apparel and accessories | 250,730 | 702% | 21.1% | 2.85% |
| Health and beauty | 58,989 | 654% | 15.6% | 4.26% |
| Tools and building materials | 142,673 | 776% | 33.8% | 1.88% |
| Sporting goods | 108,583 | 793% | 24.1% | 3.62% |
| Car parts | 152,470 | 456% | 37.6% | 1.40% |
| Furniture | 50,969 | 488% | 23.4% | 1.21% |
| Pet supplies | 20,264 | 783% | 16.7% | 5.78% |
| Electronics | 62,795 | 768% | 24.2% | 4.93% |
The highest losses are in car parts (37.6%) and tools (33.8%). These are categories with enormous catalogues of interchangeable items, where it’s hard for the algorithm to work out which specific part number the buyer wants.
Eight types of store
Breaking the sample down by the combination of “share of selling products × loss share × ROAS”, we got stable groups:
| Type | Stores | Median ROAS | Products with sales | Loss share |
|---|---|---|---|---|
| Star factory ⭐ | 34 (26.6%) | 548% | 14.4% | 13.4% |
| Balanced growth | 7 (5.5%) | 601% | 4.7% | 35.2% |
| Struggling with scale | 28 (21.9%) | 491% | 1.6% | 63.9% |
| Dark catalogue | 22 (17.2%) | 415% | 1.2% | 63.9% |
| Zombie farm 🧟 | 16 (12.5%) | 164% | 4.5% | 64.3% |
| Budget drain 💸 | 6 (4.7%) | 24% | 0.4% | 97.4% |
Note the “Star factory”: that’s 27% of the sample, with a median ROAS of 548% at a loss share of just 13.4%. Their catalogues are small — a median of around 2,200 products. A managed catalogue produces a better result than a huge unmanaged one.
And at the opposite pole, 6 stores with a 97.4% loss share. Almost their entire budget over 13 months went to products that sold nothing.
Frequently asked questions
“My ROAS is 600%, everything’s fine — why do I need this?”
Good ROAS and budget losses aren’t opposites; they coexist quite comfortably in one account.
Look. You put in 100,000 UAH and got 600,000 UAH of revenue. ROAS 600%, excellent. But inside that 100,000 UAH it could be like this: 40,000 UAH went to products that sold absolutely nothing, and the remaining 60,000 UAH produced all the revenue. Which means the real return on the working part isn’t 600% but 1000%. And the 40,000 UAH you simply gave away.
A good ROAS means the working part of the catalogue is strong enough to carry the result together with the ballast. That’s not a reason to relax — it’s headroom you haven’t collected yet. The sample contains stores with above-median ROAS and a loss share close to 70%.
“My catalogue is small, this study is about huge stores”
Catalogue size doesn’t determine the result at all — and that’s visible directly in the data. Compare two groups of stores from the table above:
| Store type | Catalogue | ROAS | Loss share |
|---|---|---|---|
| Star factory | 2,229 | 548% | 13.4% |
| Zombie farm | 2,524 | 164% | 64.3% |
The catalogues are almost identical — around two and a half thousand items. ROAS differs by more than threefold, loss share fivefold. It isn’t about the number of products but about which of them the budget goes to. On a small catalogue that difference is simply more noticeable in money per item.
“Shouldn’t Performance Max sort this out by itself?”
Performance Max is the dominant campaign type in the sample: 2,665 campaigns, carrying the bulk of all advertising budget. And even so the median loss share is still 40.5%.
The reason is simple: Google’s algorithm optimises for the goal you gave it, on the data it can see. It doesn’t know your margin — to it, a product with a 5% markup and one with a 60% markup are the same. It doesn’t know which items matter to you. And it won’t, on its own initiative, give a chance to the 55% of the catalogue currently receiving no impressions.
“I’m afraid to touch what’s working”
Sensible. That’s exactly why the first step here isn’t a change but a measurement. You can break the catalogue into layers in read-only mode, changing nothing in the account. After that you decide for yourself whether anything is worth touching.
“What if the product just hasn’t had time to take off?”
That happens, and we calculated it separately. Of the products that started the 13 months with zero impressions and still reached a first sale, the median path took 4 months. So patience makes sense — but within months, not years.
After that the survival curve statistics take over: after 20 clicks with no sale the odds fall to 39%, after 50 to 20%. Over a year of advertising isn’t “hasn’t had time”, it’s an answer. That said, the answer “this product doesn’t sell in advertising” doesn’t mean “switch it off”: 35% of such items convert later if you leave them a minimal budget.
What to remember
- 92% of products didn’t sell in 13 months, and 55% of the catalogue received no advertising budget at all. Average ROAS doesn’t show this.
- The median loss share is 40.5% of budget. It’s this number, not ROAS, that separates the top quartile of stores from the bottom (39.8% against 71.8%).
- A winner lives 3 months. A campaign structure reviewed less often than that describes yesterday’s catalogue.
- There is no hard switch-off threshold — 35.4% of “losers” sell later. Budget reallocation works; switching off doesn’t.
Look at your catalogue through the same eyes
All the numbers above are medians across 130+ stores. Your store is somewhere among them, but exactly where isn’t something overall ROAS can tell you. Connect your Google Ads account and you’ll see your own distribution: loss share, catalogue layers, the products eating budget and the ones being denied it. Access is read-only, we change nothing in the account, and you can disconnect in one click.