Research · 2026-07

We scored 50 product pages against 13 buying-fact checks. Median: 41/100.

No survey, no opinion piece. We ran 50 public ecommerce product pages through the same rule-based checker behind the free CartGap Grader and recorded what came out. Here is the distribution, the checks pages skip most, and what to write instead.

In three lines

50 pages, 6 niches, 17 stores, scored by the Grader's own scoreText() function — the same code that runs in your browser when you paste a page into the free tool.
Median score: 41/100. Mean: 43.3. Only 4 pages (8%) reached the "Solid" band (70+). None reached "Buying-facts complete" (90+).
The three checks pages skip most: who the product is not for (0% of pages state it), a named buyer and use case (8%), and assembly or installation info (14%).

Method, and where it's weak

The checker is 13 pattern-based checks worth up to 100 points: numeric specs, named material, what's included, fit and compatibility, who it's not for, assembly info, shipping, returns and warranty, specific social proof, FAQ coverage, a named audience, text coverage, and a penalty for vague adjectives with no fact behind them. It looks for the pattern of a stated fact, not whether the fact is true. It cannot tell you your return window is generous, only whether the page states one.

We picked 50 live product pages from small-to-mid Shopify stores across 6 niches (apparel, beauty, gadgets, home, kitchen, pets), pulled the visible page text with a lightweight extractor, and ran it through the identical scoring function used in the public Grader. No manual score edits. Store names are recorded internally but not published; we refer to them by niche only.

Where this study is weak, stated upfront:

The checks are keyword and pattern heuristics, not semantic understanding. A page can state a fact in a way the checker misses (a false negative), or use a matching phrase without answering the buyer's question (a false positive on our side, not yours).
English-language pages only. We have not tested the checks against other languages.
50 pages is a small sample from stores we could find and read via public search, not a random draw across ecommerce. Treat this as a case study, not a market-wide statistic.
The text extractor does not strip CSS-hidden elements (for example display:none blocks), so word counts on a few pages may run slightly high.

Score distribution

Score summary across all 50 pages (0–100 scale)
Min	Median	Mean	Max
12	41.0	43.3	74

Pages by band
Band	Pages	Share
Guess-heavy (0–39)	22	44%
Partial (40–69)	24	48%
Solid (70–89)	4	8%
Buying-facts complete (90+)	0	0%

What's missing, ranked

Pass rate is the share of pages that earned full points on that check. Sorted worst first.

13 checks, ranked by miss rate (all 50 pages)
Rank	Check	Pass rate	Pass	Partial	Zero
1	#5 Who this is not for	0%	0	0	50
2	#11 Named audience and use case	8%	4	0	46
3	#6 Assembly or installation info	14%	7	0	43
4	#4 Fit and compatibility conditions	40%	20	0	30
5	#1 Numeric specs (dimensions/weight/capacity)	42%	21	0	29
6	#P Vague/unsupported adjectives (penalty check)	46%	23	23	4
7	#3 What's included	48%	24	0	26
8	#9 Specific social proof	50%	25	0	25
9	#8 Returns and warranty terms	52%	26	0	24
10	#10 FAQ or objection handling	60%	30	0	20
11	#7 Shipping information	66%	33	0	17
12	#2 Material named	76%	38	0	12
13	#12 Text coverage (word count)	94%	47	3	0

#12 uses linear partial credit for word count, so a "partial" result there means the page is short, not wrong. #P is a penalty check — "pass" means no vague-adjective penalty was applied, not that the page earned points.

Anatomy of the top 5 gaps

0 of 50 pages state this

#5 · Who this is not for

Buyers ask: am I the exception this will not work for? A page that only argues the yes case leaves the no-case buyer to find out after the box arrives — which is where returns and bad reviews come from.

Before(nothing — the page argues the yes case and stops)

After"Not for outdoor storage, heavy tools, or heat above 60°C. For garage use, see our steel bin line."

92% of pages skip this

#11 · Named audience and use case

Buyers ask: is this for someone in my situation? "Perfect for everyone" answers nobody's question.

Before"Perfect for everyone!"

After"For renters who cannot drill walls. Use case: adding two shelves of storage above a washer without tools."

86% of pages skip this

#6 · Assembly or installation info

Buyers ask: how much work is this before it works? "Quick and easy setup" is a claim, not an answer.

Before"Quick and easy setup."

After"Assembly: about 15 minutes. Two people recommended for the top shelf. Hex key included; no drilling."

60% of pages skip this

#4 · Fit and compatibility conditions

Buyers ask: will it work with what I already own? "Fits most" is a guess the buyer has to test at home.

Before"Fits most monitors."

After"Fits flat monitors 1-3 cm thick. Not for curved screens or bezels under 5 mm."

58% of pages skip this

#1 · Numeric specs (dimensions/weight/capacity)

Buyers ask: will it fit my space, bag, or desk? "Compact and lightweight" is not a measurement.

Before"Compact and lightweight design."

After"18 x 28 x 12 cm, 320 g. Holds 4.5 L."

Score by niche

Median and mean score by niche (n = pages scored)
Niche	Pages	Median	Mean
Pets	9	56.0	55.3
Home	9	54.0	55.1
Gadgets	9	40.0	37.6
Beauty	6	38.0	40.7
Kitchen	9	34.0	38.7
Apparel	8	24.0	30.0

Six niches, not a ranked "best industry" list — sample sizes are 6 to 9 pages per niche, small enough that one or two pages move the median.

Score your own product page

The same 13 checks, running in your browser. Paste your page text, get the score and the top gaps free. A $9 report adds a rewrite template for every gap.

Grade a page free →