A product title looks like the simplest field on a page. It is a short string, somebody writes it in a minute, and it goes into a box.
It is also the only field on the page that appears in three completely different situations without changing a character, and the three situations want different things from it.
Three readers, one string
The title is read by a search engine, by a person scanning a grid of products, and by a person who has already arrived on the page. Calling all three “the customer” hides the problem.
• Search needs matchable substance: the category, the material, the defining feature. Word order matters to it far less than word presence, and it does not care whether the result reads well.
• The grid needs difference. Someone scanning a row of twelve similar garments is asking which of these is not like the others, and they resolve that question in a fraction of a second. They see the first few words, because the rest is cut off or skipped.
• The page needs confirmation. Someone who clicked is checking that the thing in front of them is the thing they thought they clicked. They read the whole string, and they read it against the image.
Those are not three descriptions of one need. Search wants completeness, the grid wants contrast, and the page wants correspondence with the picture. One string has to serve all three, and it cannot fully serve any two of them at once.
The three also disagree about repetition. Search tolerates a word appearing that also appears elsewhere on the page; the grid reader is actively harmed by it, since a word shared with every neighbor carries no information; and the page reader finds it merely redundant. So a word can be simultaneously worth including, worth removing, and neutral, depending on which reader is being served — and there is no way to include it for one and withhold it from the others.
Order is a decision about which reader to lose
Since the string is fixed and the beginning is the only part every reader sees, sequence carries the whole trade.
Put the category and material first and search is well served, the page is adequately served, and the grid reader sees words identical to the eleven items beside it. Put the distinguishing feature first and the grid reader can tell the products apart, while a search engine receives the same words in a less conventional arrangement. Put the brand first and both the grid and the page reader spend their most valuable characters on something they already knew.
There is no arrangement that satisfies everyone, which means every title is already making this choice. The only variable is whether anyone made it deliberately.
What belongs in the substance
Whatever the order, the words themselves should be real attributes rather than atmosphere. A title is not the place for the register that descriptions use, partly because there is no room and mostly because a title is treated as identification by every system that touches it.
The practical rule is that a word earns its place in a title if it is a property somebody could verify — the category, the fiber, a construction feature, a length, a neckline, an intended use with a real meaning. Those come from the same records that supply the rest of the page, and the discipline is identical: which product claims have a source.
There is a second reason to keep atmosphere out. A title is copied into contexts that strip everything around it — a feed, a link, an order confirmation, a line in somebody’s cart — where it arrives with no image and no description to steady it. Whatever the title says on its own is the entire account of the product in those places. A string built to be evocative alongside a photograph reads as vague when the photograph is gone.
Words that fail this test tend to be the ones added last, when a title feels thin. They add length without adding distinguishing information, which costs the grid reader the most, since length is exactly what they do not get to consume.
The scarcest reader is the one you have not met
Of the three, one deserves more of the opening than it usually gets.
The search reader has already been delivered to you; whatever happened in the ranking has happened, and the title’s job there is to be matchable rather than persuasive. The page reader is already in, and will read the whole string regardless of its order. The grid reader is the only one still deciding whether to arrive, working from the first few words, alongside eleven alternatives.
So the opening of a title is worth the most to the grid reader, and most titles spend it on search — stacking brand and category at the front, where they are identical to every competing item on the row. The reason is not carelessness. Search is the only one of the three readers with a dashboard attached, so it is the only one whose needs arrive in a meeting with evidence, and the reader who is scanning and leaving produces no signal at all.
The title is also checked against the picture
The page reader does something the other two cannot: they compare the title to the image sitting next to it.
That comparison is where a title with an unsupported word gets caught. A title saying oversized against a photograph of a close-fitting garment, or naming a fiber the fabric does not appear to be, creates a small unresolved contradiction at the moment of highest intent. Buyers do not adjudicate these — they discount the whole page, which is a larger response than the size of the discrepancy warrants: what happens when images and copy disagree.
The contradiction does not have to be large to do damage. A buyer who notices one small mismatch has learned something about how carefully the page was assembled, and they apply that lesson to the claims they cannot check. A title overstating one attribute quietly reduces the credibility of the measurements, the composition, and the care instructions, none of which the buyer has any other way to assess.
This is also why the title and the main image should be decided together rather than in sequence by different people. They are the two elements every visitor processes, they are processed within the same second, and they are usually produced by teams that never see each other’s work.
Titles are written once and never rewritten
The last property is the one that changes how much care the field deserves.
Descriptions get rewritten. Images get replaced. Titles do not, because the moment a title enters the systems around it, it stops being a sentence and starts being an identifier — referenced in feeds, used as a key in downstream tools, appearing in links, quoted in ads, carried into marketplaces. Changing it is not an edit to a page; it is a change that propagates to places nobody has an inventory of.
It also means titles accumulate errors rather than shedding them. A description that was written against an older version of a product gets refreshed when somebody rewrites the page; a title written against that same older version stays, because nothing in a refresh touches it and changing it feels riskier than leaving it. Over a few seasons a catalog’s titles drift further from its products than any other field on the page.
The practical consequence is that the cost of writing a title and the cost of fixing one are orders of magnitude apart, and nobody writing one is aware of that. It is another case of a decision made cheaply upstream becoming expensive downstream, with no signal at the moment of the decision.
Which argues for settling titles at the point where a page is being assembled and the images are in front of you, rather than as a standalone copy task: where the page comes together.
Questions teams ask about product titles
How long should a product title be?
Long enough to carry the real attributes and ordered so the first few words do the distinguishing, which is a structural answer rather than a character count. Truncation lengths differ by channel and device and they change, so the number has to come from where you actually sell. What holds everywhere is that the end of a title is read by fewer people than the beginning.
Should the brand name come first?
Only if buyers are searching for the brand, in which case those words are doing work. On a grid of your own products the brand is identical across every item and distinguishes nothing, so it consumes the most valuable position to no effect. The question is whether the word is separating this item from its neighbors.
Can we use the same title across every channel?
You can, and it means accepting whichever channel’s truncation is harshest. Channels differ in how much they show and in what their search rewards, so a single string is a compromise rather than an efficiency. Where the channels are very different, the compromise may cost more than maintaining two.
Should titles be generated?
The values in a title should come from records regardless of who assembles them, so generation is reasonable for assembly and unreliable for deciding what belongs. What it cannot do is make the trade-off in the second section, because that requires knowing which reader you are prepared to lose.
Our titles are inconsistent across the catalog. Does that matter?
It matters most in the grid, where inconsistent structure makes comparison harder precisely where customers are comparing. A shared pattern helps them scan even when no individual title improves. The pattern should be decided from your own channel mix rather than copied from a template.
What should we do about titles already live?
Treat them as expensive to change and audit before you edit, since a title is referenced in more places than a page. Fixing a title that is factually wrong is worth the propagation; reordering one that is merely suboptimal usually is not. New products are where the ordering decision is free.
Where this leaves you
A product title is one string doing three jobs for three readers who want different things, and its order silently decides which of them goes unserved. The reader still deciding whether to visit sees only the opening, produces no measurable signal, and is therefore the one most often given away to the reader who already arrived. Put verifiable attributes in it, decide the order against your own channel mix rather than a formula, settle it next to the image it will be read with, and treat it as expensive to change — because it is.
Read your titles in the grid, not in the spreadsheet
Open a category page and look at a row of your own products as a customer scanning it would. Note how many of the visible openings are identical to each other, and whether anything in the first few words separates one item from its neighbors. Then check whether every attribute word in each title corresponds to something in the images beside it. Decide the ordering for new products before the titles enter systems that treat them as identifiers.
Written by
