What to sell online: how to test an idea cheaply
No software picks your product, including ours. Four filters that decide a good idea, a table matching what you have to what to sell, and five cheap tests.
The short version
- This is the one decision no store builder or AI can make for you, and any tool offering a trending-products list is handing you a list everybody else already has.
- Judge an idea on four things: how cheaply you can reach the buyer, whether the unit economics survive shipping and returns, whether you know more than a generic seller, and whether you will still want to do it in 18 months.
- Test before you build. Ten real conversations and five hand-sold units tell you more than any keyword tool.
- Set a kill criterion in advance, because the purpose of a cheap test is permission to quit.
Every store builder has a product-name field waiting for you, and it is the one part of the software that cannot help.
So here is the honest answer first. Deciding what to sell online is not a software problem and no tool solves it — not a keyword tool, not a trend report, and not mAutomate, which will happily build a beautiful store around a product nobody wants. What you can do is make the decision cheap to get wrong. That is a different skill, and it is learnable in a weekend.
Why trending-product lists are worthless to you
A list of winning products is public. That is the entire problem. By the time a product appears on one, the people with capital and ad accounts are already bidding on it, the margin has been competed away, and the thing being described as "high demand" really means "high competition with no differentiation."
The useful question is not "what sells." Plenty of things sell. It is "what can I sell at a cost of acquiring a customer that I can actually afford." Those are completely different questions, and the second one depends on you rather than on the product.
The four filters that decide it
Distribution: can you reach the buyer without paying for every visit? This is the filter that matters most and gets asked last. An existing audience, a skill you can make content about, a local scene you are part of, a trade you already work in, a forum where you have been a member for years — any of these is worth more than a clever product. Without one, every single customer has to be bought at auction against better-funded competitors.
Unit economics: does it survive shipping, returns and the first customer? Heavy, bulky, fragile or cheap items fail here. So does anything where the return costs more than the sale earned. Illustratively, an $18 item with $9 of outbound shipping and $9 to get it back means a single return wipes out several sales.
Knowledge edge: can you answer the buyer's third question better than a generic seller? The first question is price and the second is delivery, and anyone can answer those. The third question — will this fit my old radiator, is this safe for my kid's eczema, which of these two do I need — is where a small store wins. If you cannot answer it, you are a worse version of a marketplace listing.
Endurance: do you want to be in this inbox in eighteen months? Ecommerce is repetition. The same questions, the same packing, the same photographs. Choosing a category you find boring on day one is choosing to quit in month nine.
Start from what you already have
| What you have | Where to look first | What usually goes wrong |
|---|---|---|
| An audience, even a small one | Ask them what they last bought and hated, then sell the fix | They like you, not necessarily your product; pre-sell to find out which |
| A craft or making skill | Small batches, made-to-order, commissions | Pricing your own hours at zero — see handmade selling |
| Local supplier or wholesale access | Brands in your region that nobody has put online yet | Exclusivity you were promised verbally is not exclusivity |
| A trade or profession | Dull consumables and tools for your own industry | Smallest audience, best margins, hardest to scale beyond you |
| Time but no money | Made-to-order, services, or digital products | You become the bottleneck and cannot take a week off |
| Money but no time | Stocked inventory in a niche you understand | You buy inventory before you have demand |
| None of the above yet | Pick an audience first, product second | You buy a course about dropshipping instead |
The pattern across that table is that the answer comes from your situation, not from a product category. Two people can pick the same product and only one of them has a business, because only one of them can reach the buyer.
Five tests that cost almost nothing
Run these before you open an account anywhere.
One: ten conversations. Not "would you buy this," which everyone answers politely. Ask what they bought last in the category, what it cost, what annoyed them about it, and where they looked. Ten real conversations beat any research tool because they tell you the words buyers use, which is also your copy and your SEO.
Two: hand-sell five. At full price, to people who are not your friends. A market stall, a direct message, a colleague's office, a stand at a local event. Five units sold by hand is the single strongest signal available, and it comes before a website rather than after one.
Three: check the search demand in the buyer's words. Use a free keyword tool and your own search bar's autocomplete. Zero search volume is not automatically fatal — new categories have none — but it tells you the product must be sold rather than found, which is a much slower business.
Four: do the shipping and returns dry run. Box the thing, weigh it, get a real quoted rate to the places your buyers live, and price a return both ways. Many ideas die honestly at this step, which is a good outcome in week one and a disaster in month six.
Five: the one-page offer. Photos, price, shipping cost, a buy button. Show it to the ten people from test one and count how many ask how to pay. Note that an email signup is a much weaker signal than someone trying to give you money — interest is free and purchases are not.
Categories that are harder than they look as a first product
People run excellent businesses in all of these. They are simply poor places to learn.
- Heavy or bulky goods, where shipping is a larger line than the product
- Fragile items, where breakage and claims become a part-time job
- Sized apparel, which is returned more than anything else you could choose
- Anything eaten, drunk or applied to skin, which brings labeling rules, insurance and liability
- Electronics with lithium cells, which many carriers restrict
- Products needing certification — children's items, safety equipment, toys
- Perishables, which punish every forecasting error immediately
- Anything with a sixty-day supplier lead time, because you are forecasting two months blind
- Licensed or fan-art products, where a takedown ends the business overnight
What software decides here, and what it does not
Be clear about the division. A platform builds the storefront, writes the product copy, handles the on-page SEO, schedules the marketing and answers the repeat questions. Ours does those things and nothing publishes without your approval. None of it constitutes an opinion about whether your product is a good idea.
Where AI genuinely helps before you commit is narrower and more useful than a trend list. It can summarize what buyers complain about in hundreds of existing reviews, which is test one at scale. It can draft three different positioning angles for the same product so you can show them to real people. It can write your one-page offer in an hour rather than a weekend, which makes test five cheap enough to run three times.
What it cannot do is tell you whether you will still care in eighteen months, or whether the person you plan to sell to trusts you. Both of those decide the outcome more than the product does.
Make the decision today, not eventually
Score yourself on the four filters and pick the idea where you are strongest on distribution, because that is the filter you cannot fix later. Run tests one and two this week.
Then write a kill criterion before you start: "if I cannot sell three units at full price to strangers within two weeks, I choose a different idea." The point of a cheap test is not to prove you right. It is to give you permission to stop, which is the thing most first-time sellers never grant themselves until they have spent far too much.
When an idea survives, the next two problems are pricing it properly and finding the first ten buyers. The platform decision comes after both, and ecommerce platform for beginners covers it. If you want to see what is available by business type, the solutions pages are organized that way.
Questions people ask
Can AI tell you what to sell online?
It cannot make the judgement, and tools that claim to are usually surfacing a public trend list that everyone else has too. AI is genuinely useful one step earlier: summarizing complaints in existing product reviews, drafting several positioning angles to show real people, and producing a test page quickly. Choosing the product remains a human decision about distribution and interest.
How do you validate a product idea without building a store?
Have ten real conversations about what people last bought in the category and what annoyed them, then sell five units by hand at full price — at a market, by direct message, or in person. Add a shipping and returns dry run with real quoted rates. Those three steps cost almost nothing and kill bad ideas in a week.
What products should a beginner avoid selling online?
Heavy or fragile goods where shipping and breakage dominate, sized apparel because of return rates, anything ingested or applied to skin because of labeling and liability, items requiring certification, and products with sixty-day supplier lead times. Good businesses exist in all of these categories; they are simply expensive places to make your first mistakes.
Is it better to pick a product or pick an audience first?
Audience first, in almost every case. The hardest cost in a small store is acquiring each customer, and an audience you can already reach for free removes it. A product chosen without a route to buyers has to be advertised against competitors with deeper pockets, which is the most expensive way to learn ecommerce.
- Getting started
- Product research
- Validation
Ready to start your store?
Let the AI build, market, and run it — you ship the product.
Get started free