Marketplace search behaves differently from web search in one important way: the shopper has usually already decided to buy something, and is comparing a handful of specific listings. That makes keyword research more about matching how people describe your product than about reaching a broad audience. Getting the product data and images right still matters more than any individual term.
Start with what shoppers already tell you
The cheapest research sits in front of you. Read the customer questions and review text on comparable listings, noting the words people use for the problem, the feature and the occasion. Look at what your own customers call the product in enquiries and support tickets, because that language is closer to the search than anything a tool will produce. Then check the platform's autocomplete suggestions and the related-search prompts, since those are generated from real behaviour.
Understand the intent behind each term
Group terms by what the searcher wants. Some indicate a clear purchase intent with a specific product type. Some indicate a problem, such as a replacement part or a fix for something that broke. Some are broad category words where you will not compete easily. Some are brand-led, where the shopper already knows what they want and will not look at alternatives. Record which group each term belongs to before deciding where it goes on the page.
Separate visible fields from hidden fields
Terms that matter most belong in the title, because the title carries the main matching weight and is read by the customer. Supporting terms belong in the feature points and description. Anything that is a genuine synonym not already stated can go in the backend search term fields. Never place a competitor's brand name anywhere on your listing, and check the current prohibited-claims and keyword guidance in the official seller documentation before a bulk edit, since the rules differ by category.
Judge terms on whether you can win them
Volume is not the only measure. Open a term and look at the listings ranking for it. If they are well-established brands with heavy promotion and large review counts, a new listing may not get a look, however well written it is. Terms with fewer results, more specific intent and products that are imperfectly presented are usually more realistic starting points. In many cases the strategy that works is to win a handful of specific terms rather than chase one broad word.
Keep a catalogue-wide picture
If you sell many related products, avoid targeting the same term across the whole range. Build a simple map of which product owns which term and variant language, so two listings do not compete with each other for the same search. Check for products with thin content that duplicate each other, because consolidating or clearly differentiating them usually helps more than adding new terms.
Build the research once and reuse it
Saving the work matters more than the tool you use. Keep one sheet or document per category, listing the terms you found, where each one came from, which listing currently uses it and which product it describes. When a new product joins the range, the sheet shows you immediately whether the term is already owned by something else.
Read the ranking pages properly
For each candidate term, look at what actually appears. Are the results the same large sellers every time? Do the listings have heavy promotion behind them? Can you genuinely serve the searcher? You are looking for terms where the existing listings are imperfect and the customer need is real.
Use competitors honestly
Looking at what others in the category have written is normal research, and it tells you the vocabulary and the gaps. The line is between learning from the language of a category and copying a competitor's listing text or using their brand name to attract traffic. Copy the structure, not the sentences.
Test and read the results
Change one group of listings at a time and give it a fair period before judging. Track impressions, clicks, the conversion rate and the questions customers still ask. A listing that gains impressions but not clicks probably needs better images or a clearer title rather than different terms. A listing that gets clicks but not orders needs price, delivery or a description fix. And accept that some terms will never work for you; knowing which ones to stop pursuing is part of the work.
What to ask a keyword service
Ask to see the terms they would target for one of your products and why. Ask which terms they would deliberately avoid. Ask how they found them, and whether they would use the platform's own tools or only third-party estimates. Ask what happens to the term map if you add products later, and confirm you will keep the account and the map yourself.
Our marketplace guides cover how these terms get used across the listing. If you want the research and the writing done together, our Amazon and Flipkart listing work starts at INR 1,500 per product, with marketplace commissions, platform subscription fees, shipping and ad spend separate and set by the platforms. You can send us your product list for a view of which terms look realistic.
Common questions
Should I use the same keywords on every listing in a category?
No. Each product should target the terms that describe it and the way its own customers search. What repeats should be the accurate product type, brand and attribute structure, because those belong in the title and the visible copy. What differs should be the use case, the material, the audience and the variant language. Identical keyword sets across a catalogue usually create pages that compete with one another.
How do I know which terms are worth targeting?
Look for terms where you could realistically satisfy the search, not just terms with volume. Check what the ranking pages for a term look like: if they are large brands with thousands of reviews and heavy promotion, that term may be a poor target for a new listing. Longer, more specific phrases with real intent usually offer a better starting point than the broadest word in the category.
Is it worth using paid keyword tools?
Sometimes. The platforms' own search and autocomplete data plus free third-party estimates are enough for a small catalogue. Paid tools become worthwhile when you have many products, when you need to manage terms across a large catalogue consistently, or when you need volume estimates you can trust for planning. Buy one when you know which decision the data will change.