Shopify product description length: it's not the median
We have already measured how long a Shopify product description actually runs to. If that is the question you came with, the answer is in product description examples vs live products — a distribution taken from a different sample, with the quartiles and the shape.
This piece asks the question that one left open. Does the answer depend on how big your catalogue is?
On 22 September 2026 we read the public product catalogue of 178 Shopify storefronts — 6,393 products — and split them by catalogue size. The typical description barely moves. Almost everything else does.
The typical store writes the same length either way
We read up to 50 products per store, which splits the set in two: 105 stores filled that read, so they hold 50 or more products and we are seeing their 50 most recently updated; 73 came back short, so we saw their entire catalogue.
| Filled the 50-product read | Whole catalogue seen | |
|---|---|---|
| Stores | 105 | 73 |
| Products measured | 5,250 | 1,143 |
| Median description (per store) | 668 | 696 |
| Products under 400 characters | 31.0% | 22.0% |
| Typical store's share under 400 characters | 20.0% | 0.0% |
| Stores with a duplicate description | 42.9% | 15.1% |
Read the third row first, because it is the one that surprised us: 668 characters against 696. Merchants with big catalogues are not writing shorter descriptions. Whatever they write, they write at about the same length as everyone else.
See what Arvio does with product descriptions
What changes is how many products get skipped
Half the stores in the left column have at least a fifth of what we saw under 400 characters; only 29 of 105 had none that thin. So the median store with a large catalogue is carrying one thin product in five, while writing perfectly normal descriptions for the rest.
The second chart is there because a median can mislead you here. On the right, 38 of 73 stores had nothing under 400 characters — which looks like the smaller stores are simply better. Read it with its context: those stores have a median catalogue of 8 products, and 42 of them hold fewer than ten. On a store with three products, "share under 400 characters" can only land on a round fraction, and 9 of them were under 400 on every product they had.
The smaller stores are not uniformly better. They are small enough that there is no long list to lose track of.
Duplicates follow the same curve, harder
In the left column, 42.9% of stores ship at least one description word-for-word identical to another product in the same store, against 15.1% on the right. Across the whole set that is 772 products carrying a copy of some other product's text, 702 of them in the left column.
Empty descriptions are not counted as duplicates of each other, so that figure does not overlap with the empty rate in the earlier piece. A duplicate here means two products whose descriptions match once you strip the HTML, collapse whitespace and lowercase.
That matters more than a short description does. A thin description is a weak page; two identical descriptions tell a search engine that two of your pages are the same page. If you want the mechanics of finding and merging them, we wrote that up separately in Shopify duplicate products cleanup.
How Arvio handles the products you never scroll to
Why a big catalogue does this
Nobody decides to leave a product blank. You write the early ones carefully, then a supplier sends a spreadsheet with a whole batch more, and you paste the same paragraph into all of them to get the store live. Those products are in your catalogue and in your sitemap, and you will never scroll to them again.
The problem is not the length you chose for the products you were thinking about. It is the products you stopped thinking about — and that group grows with the catalogue, not with the writing.
What to actually do
Sort by description length, not by date. Your admin sorts by newest, best-selling, or alphabetical, and none of those surface the problem. Export your products to CSV and sort ascending on the description column, then work from the bottom. One caution from our own method: an exported description is raw HTML, so a product that looks like it has 200 characters of markup may have almost no words in it. Strip the tags before you judge the length, or just open the short ones and look.
Look for duplicates before you look for short ones. Rewriting a duplicate is faster than writing from nothing, and it fixes the worse problem first.
If one product in five needs work, going through them one at a time is the same process that produced the list. The mechanics of doing it in bulk are in how to bulk edit product descriptions in Shopify.
Arvio was built for that shape of job: it finds the products with missing or duplicated descriptions and writes the fix rather than handing you a report. You choose the scope — one product, a batch, or the whole catalogue — and you approve or reject the changes before anything goes live, with any edit reversible in a click.
Check your own catalogue with Arvio
How we measured this
The sample frame is the Shopify Community "Store Feedback" board, where merchants post their own store URL and ask strangers for a review. We read the first post of 600 threads, posted between March 2025 and September 2026, and took every linked hostname. That gave 406 candidate domains, of which 185 returned a valid Shopify product catalogue. We then requested each one's public product endpoint once, anonymously — the same endpoint any visitor or crawler can already reach. 180 answered, two of those had no products, and that leaves the 178 storefronts and 6,393 products above.
What this doesn't show
- This is not a random sample of Shopify. It is a public, reproducible convenience frame. Merchants who post asking for feedback skew newer, smaller, and more likely to be struggling. Anyone can rebuild the same list from the same board, over the same date range, with the same rule.
- The larger stores were read one page deep. We see their 50 most recently updated products. If recently-touched products are better tended than forgotten ones, the tail in those catalogues is worse than what we measured, not better.
- Catalogue size here is a floor, not a count. A store in the left column holds at least 50 products; we cannot tell from this read whether it holds 51 or 5,000.
- Length is plain-text characters, not words. HTML is stripped before counting, so an empty paragraph tag counts as zero.
- Five stores could not be read and two had no products. They are excluded from every denominator rather than counted as zero. Unreachable and empty are different things.