Good ideas are found in complaints, not in imagination. Mine negative reviews, returns language, support threads, supplier catalogs, and habits from adjacent categories, then score every source on the same sheet. Brainstorming that produces a list without a scoring method produces enthusiasm, which is the most expensive input in this business.
The short version
- Start from an unhappy buyer, never from a blank page. The complaint is the brief.
- Volume of ideas is worthless without a filter. Twenty scored beats two hundred collected.
- Supplier catalogs answer a different question. They tell you what can be made, not what is wanted.
- Adjacent categories are underused. A solved problem in one aisle is often an open one in another.
- Score the source, not only the idea. Some inputs reliably produce launches, and some reliably produce distractions.
Why ideas are found rather than invented
Product ideas that work on Amazon share one structural property: somebody is already paying for a worse version. That single fact should reorganize how you brainstorm, because it means the useful raw material is evidence of existing spending, not creativity.
An invented idea has to prove three things at once, that demand exists, that people will search for it, and that your execution is right. A found idea only has to prove the third. Demand and search behavior are already visible in the market you took it from, which removes most of the risk you would otherwise be funding with a production run.
That is why every method below is a way of listening to a market rather than a way of generating novelty. The creative work happens after the idea is found, in how you fix the specific thing buyers keep complaining about.
Where ideas actually come from
- Negative reviews on the top listings. Read fifty per candidate. The complaint that repeats and has a manufacturing fix is a product brief.
- Return reasons and their language. Returns tell you what disappointed a buyer who had already decided to buy, which is a stronger signal than a review.
- Question and answer sections. Repeated questions mean the listing is failing to communicate, and sometimes that the product is failing to deliver.
- Your own supplier's catalog. Useful for adjacency and manufacturing capability, dangerous as a starting point, since it tells you what is easy to make.
- Adjacent categories that solved the same problem. A material, closure, or packaging format that is standard in one category and absent in another.
- Bundles people build for themselves. When buyers routinely purchase two items together, that is a product waiting to exist.
- Trade shows and supplier fairs. Best used to find a manufacturing capability you did not know existed, not to find a product.
- Communities where your buyer complains for free. Forums, subreddits, and comment sections carry unfiltered language you can put straight into a listing.
- Your own repeat annoyances. Weak on its own, useful as a source of candidates that you then have to validate like any other.
Score the sources, not only the ideas
Most people score products. Almost nobody scores where their products came from, which is why the same unproductive habits repeat for years. Track your own sources over a few cycles and the pattern becomes obvious quickly.
| Source | Signal strength | Effort per usable idea | What it is good for |
|---|---|---|---|
| Negative reviews | High | Medium | Finding the fixable complaint you can build against |
| Return reasons | High | Medium | Understanding why buyers regret the purchase |
| Question and answer threads | Medium | Low | Spotting communication gaps and missing variants |
| Supplier catalogs | Low | Low | Testing whether an idea can actually be manufactured |
| Adjacent categories | Medium | High | Transferring a proven solution into an unsolved aisle |
| Bundle behavior | Medium | Low | Raising order value without a new product line |
| Trade shows | Low | High | Building supplier relationships and finding capability |
| Personal annoyance | Low | Low | Generating candidates, never conclusions |
The two columns to read together are signal strength and effort. Low signal with low effort is where most people spend their time, because scrolling a catalog feels productive. High signal with medium effort is where launches come from, and it involves reading a great many unhappy paragraphs.
Turning a pile of ideas into a shortlist
Do it in one sitting, with a rule you wrote before you started. Collect twenty candidates, kill anything gated, patented, fragile, or oversized on sight, then keep only the ideas where you can name the specific complaint you are fixing and the change your supplier would have to make. Whatever survives goes into proper research. Whatever does not gets deleted rather than saved, because a maybe list is where good discipline goes to die.
Ask who is doing this work for you
If you are paying someone for product ideas, the question that matters is not how many ideas they produce. It is who produces them and where those people sit. This work gets outsourced constantly, and a shortlist assembled by a subcontracted freelancer from a database is not research, it is retrieval.
Everything at Flapen is in-house, with no subcontracting at all. Sourcing runs through our own studio in Guangzhou, creative through our own studio in Dubai, and the frameworks we use were built across more than 500 brands. I am not saying in-house is the only workable model. I am saying you should know the answer before you pay, and you should ask for the name and location of the person who will actually read those fifty reviews.
What most agencies will not tell you
Idea generation is the cheapest part of the process and the part most heavily marketed, because it is the part clients enjoy. The expensive, decisive work is elimination, and nobody sells a service called saying no to nineteen of your twenty ideas. When a proposal is thick with ideas and thin on rejection criteria, it is optimized for how the meeting feels.
The second thing: many idea lists are recycled. The same databases, filters, and thresholds are available to everyone paying the same monthly fee, so the outputs converge. If your shortlist was produced by a filter anybody can run, assume several hundred other sellers received something similar this month, and expect to meet them in the search results about eight months from now.
Related answers
- Recommend a product research framework for Amazon
- How to use reviews to discover product gaps
- Niche ideas for first-time Amazon sellers
- What products to bundle for higher AOV
- Amazon seller roadmaps and capital: the complete guide
Bring a shortlist and we will tell you which ideas we would kill, and why, in writing at Flapen.
