How to write product descriptions that convert and rank

Wednesday afternoon, a brand selling office furniture online. The support rep answers the same question for the third time this week: does this chair fit under a desk with a low frame? The clearance is on the page. It sits on the ninth row of the specification table, as a number in millimeters. What the buyer is asking is not a measurement, it is a fit. Forty-odd messages a week go this way, and a few boxes come back every month marked not what I expected. The real bill never shows up in any report: the shoppers who could not be bothered to ask and simply closed the tab.

The product description is the most written and least considered field on a page. What follows starts with the job a description actually performs, the order in which a buyer's questions arrive, and why republishing the manufacturer's copy is an expensive shortcut. Then: the blocks a description is built from, how a feature turns into a sentence a buyer can use, why search fit is different from sprinkling keywords, what the machine-readable version of the same text has to say, which number tells you the description is working, and how any of this survives a catalog of four hundred products.

PromiseProofSpecsDecisionDescription

What job does a product description actually do?

The common assumption is that a description is persuasion. On-page behavior does not support it. Buyers look at the image first, then the price, and reach the description only once a hesitation has formed — and even then they scan for one answer rather than reading top to bottom. A description is less an advertisement than a help desk.

That desk has four jobs: speed up the decision, answer the objection, match the query a buyer actually types, and prevent the wrong purchase. The fourth is discussed least and pays the most, because copy that never says who the product is not for wins the sale today and pays for the return three weeks later.

So the question when writing is not how to praise the product, but what remains unresolved in someone's head just before they buy. That answer lives in the support inbox, not in the imagination. How on-page hesitation gets measured and removed is covered more broadly in our article on conversion rate optimization.

What does a buyer ask when they land on the page?

The outline of a description comes from your last fifty pre-sale questions, not from a copywriting framework. It takes an afternoon: pull the questions that reached support and chat for one product group, list them, count the repeats, sort them. That list is your heading order. In most catalogs the questions cluster into six.

• Will it fit me: Size, capacity, compatibility, whether it works with what the buyer already owns; they are looking for their own situation, not a number.
• What is in the box: The parts included and the ones sold separately; a page that stays quiet about the missing bracket manufactures disappointment at unboxing.
• How long will it last: Material, duty cycle, warranty scope and spare part availability; a durability claim tied to no measurable thing says nothing at all.
• What does assembly require: Time, tools, whether it needs two people; withholding this does not make assembly easier, it postpones the complaint.
• Who is it not for: Name the scenario where the product fails. This line does not cost conversions — it filters out the wrong buyer and visibly reduces what lands in your returns process.
• What happens afterward: Delivery window, setup support, exchange terms; the biggest uncertainty at checkout is often not the product but everything after it.

Why is the manufacturer's copy an expensive shortcut?

Publishing the supplier's text as received looks free. It bills you in two places. First, that text sits on ten other storefronts alongside yours, so a search engine has nothing to tell you apart with and the only remaining difference is price — a race the smaller seller does not win. Second, manufacturer copy describes the product, not its use; most of what buyers ask never existed as a question for the manufacturer.

There is an honest exception to this advice. For products people search by part number, where the specification is already standardized, rewriting the manufacturer's text earns almost nothing. On a bearing or a toner cartridge page, value comes from an accurate compatibility list and live stock, not from prose. Spend the rewriting effort on products buyers choose between, and keep the data flawless on products buyers merely verify.

Which blocks is a description built from?

A working description has five blocks, ordered not by reading sequence but by the order in which doubt appears. A one-sentence promise says what the product does and for whom. Three bullets close the three most frequent questions early. A use scenario places the product inside the buyer's own day. The specification block exists for verification. The scope block covers what is in the box, what is not, and who this is not for.

The promise line is the hardest sentence in the text, because it is the spot most exposed to adjective pile-up. Sleek, premium and comfortable produce no picture in anyone's head. We go through how frameworks like AIDA and PAS apply on a product page in our piece on copywriting; on a description they collapse into a single rule — put a measurable behind every adjective.

The fifth block is missing from most pages: the customer's own words. A short quote lifted from a review and dropped next to the relevant claim persuades far more than the same sentence from the brand. Letting social proof spread through the text instead of waiting in a block at the bottom means it answers doubt where doubt appears.

What stays above the fold?

On mobile the first two lines of the description are visible and the rest hides behind a link. Repeating the product name in those two lines is the most common waste on the page. What belongs there is the answer to the most frequent question about that product. For the chair, it is not the upholstery material — it is the clearance under a desk.

How long should it be?

There is no universal number, and there should not be. Length is set by how many questions the buyer has, not by the price of the item. For consumables and repeat purchases, long copy is an obstacle: the buyer already knows the product and only wants the right variant, fast. For a first-time purchase that needs assembly or depends on fit, short copy produces returns directly. The rule is simple enough: the text runs until the pre-sale questions for that product are answered, and stops there.

How do you turn a feature into a sentence a buyer can use?

Feature-to-benefit appears in most guides as a slogan; in practice it is a sentence pattern. Feature, what it therefore means, who cares. Holding all three in one sentence does not lengthen the text — it collapses two repetitive paragraphs into one line.

Feature	Which means	Matters to
Removable cover	Comes off and goes in the machine	Homes with small children or pets
Eight-hour battery	A full shift without hunting for a socket	Teams working in the field
Modular shelving	Extends without dismantling as you grow	Businesses that move within the year
Standard fitting size	Swaps into existing plumbing	Buyers upgrading without renovation
Spare parts available	A fault replaces a part, not the unit	Buyers planning years of use

The third column sets the tone as well. The same product is described in different sentences depending on whether the buyer is a business or an individual; the item is identical, the risks are not. A business buyer asks about continuity and support. An individual asks about ease and appearance.

Search fit: put the words into the questions, not into the copy

On a product page, search fit comes from the text naturally containing the words people use, not from keywords sprinkled through it. Most queries do not include the official product name; they include a color, a size, a purpose, a compatibility, or a description of a problem. If those words appear nowhere on your page, the product stays invisible to the right buyer.

The practical route is to catch queries with questions rather than headlines: if your subheadings and scope lines are already written in the buyer's phrasing, search fit arrives without extra work. Separating the words people genuinely search from the internal jargon only you use requires keyword research at product level; research done at category level never surfaces variant queries.

A product description is not a well-written advertisement, it is the pre-emptive answer to the question a buyer could not be bothered to ask.

The machine-readable version: structured data and feeds

A description serves two readers, and the second is not human. Price, availability, brand, model, size and rating exposed through structured data markup make the same page appear richer in search results. Those fields must not contradict the visible text; a product marked in stock in markup and sold out on the page is the fastest way to lose trust.

The same data drives the ad side. In shopping campaigns, matching is decided by the titles and attributes in the product feed rather than by ad copy, so product feed optimization and description writing should draw on one source, or you end up maintaining two versions of the truth. A third reader has become visible more recently: generative engines summarize product pages and cite them inside answers, but only pages structured well enough to be read. We covered why that reader skips your page in a separate article.

Which number tells you the description is working?

Conversion rate on a product page is not enough, because it measures everything at once, from the photography to the delivery estimate. Three indicators isolate the description. First, pre-sale questions per hundred sessions: when that falls, the copy is doing its job. Second, not the return rate but the distribution of return reasons — everything filed under not as described is the description's invoice. Third, the change in add-to-cart rate.

The limits of measurement deserve saying out loud. Most product pages never see enough traffic to compare two versions of the copy with any confidence, which makes an A/B test on a single product wasted effort for most small businesses. What works instead is testing at category level: move every description in one category to the new structure, leave a comparable category untouched, and watch the difference for eight weeks.

How do you manage four hundred products?

A method that is right for one product may be unworkable across four hundred. What scales is treating a description not as one block of prose but as a structured record filled field by field: promise, three bullets, scope, specifications stored separately and assembled on the page. When the delivery terms change, you update one field rather than four hundred texts.

On the catalog side this means product information living in one place and flowing from there to the site, the marketplaces and your quotes, which we cover in our article on product information management. Variants deserve one warning: publishing twelve separate description pages for twelve colors puts your own pages in competition with each other. Variants belong on one product page, the description lives on the parent, and the variant-specific difference gets a single line.

Where templating ends

Applied to an entire catalog, templating creates its own problem: four hundred pages become interchangeable and none of them tells anyone anything. The workable balance is writing the first hundred products by hand, the ones carrying revenue, and filling the long tail from a template. When a tail product unexpectedly starts selling, it gets promoted to the handwritten list.

Where should you start?

Launching a project to rewrite the whole catalog stalls at week six. The faster route is to pick the ten products that sell most and generate the most questions, pull the last fifty questions those products attracted, and rebuild their descriptions around them. Two weeks later, count the questions those same products generate. The difference tells you whether the method works before you scale it.

Rule out three frequent mistakes while you are there: repeating the product name in the first sentence, hiding what falls outside scope, and letting the specification table stand in for the text. A table is for verification, not decision; decisions are made by sentences.

Keeping this sustainable depends on recording which questions a description closes and which returns came from which gap. Rocketly keeps customer messages, product records and sales history in one place, so you can see which product generates which question — create a free account and build the loop for your own catalog.