Summarize with
Share

Most beginners spend three weeks “researching” and still pick the wrong product. The reason is simple: they confuse browsing for validation. They scroll Amazon, find something that looks promising, get excited, and order 500 units. Then returns pile up, or a price war starts, or the search volume was never there to begin with. We have onboarded 47 paying clients into Amazon launches since 2022, and the single biggest predictor of whether a launch worked was not the budget or the supplier. It was whether the product passed a hard checklist before a single sample shipped. This is that checklist, the same 9-step validation we run on every client product, written so you can run it yourself.

9
Validation checks every product must pass before we source it
$200K
Monthly Amazon revenue we manage for paying clients
47
Beginner clients onboarded since 2022, same research lesson in 8 of them

01Where research sits in the whole business (and why it decides 90% of the outcome)

The Amazon private label model has four phases. You do product research, you source and manufacture, you ship the inventory into Amazon’s fulfillment centers, and then you sell, market, and optimize. That is the whole thing. Four steps, idea to launch.

Here is what surprises most beginners: phase one decides almost everything that happens in phases two, three, and four. You can have a great supplier, clean photography, and a sharp PPC plan, and still lose money for a year if the product itself was wrong. We have watched it happen. The launch with a strong product and an average team beats the launch with a weak product and a great team, every time.

So the order of effort is backwards for most people. They rush research to get to the “real work” of sourcing and launching. We do the opposite. We spend the most disciplined time at the start, because that is the cheapest place to fix a mistake. Rejecting a bad product costs you an afternoon. Discovering it after you have ordered inventory costs you thousands of dollars and six months.

A quick map of where the money goes, so you see why the research stage carries so much weight:

  1. Product research: an afternoon of your time, near zero cash
  2. Sourcing and samples: roughly $150 per supplier sample, then a first production order
  3. First production order: around $4,500 landed for 500 units (about $9 per unit) on a mid-size product shipped DDP
  4. Photography and listing content: $200 to $800 depending on whether you use a full team or an AI-assisted workflow
  5. Launch PPC: around $3,000 over the first 60 days at the $50-a-day operator floor

Add it up and a realistic first launch deploys somewhere around $8,000 to $9,000 in cash. Phase one is the only phase where you can change your mind for free. Spend it well.

02Why most beginner research goes wrong (the lift-decline trap)

Neil came to us about two years ago with a wood-therapy massage tool already in market. Real product, decent listing, ranking on “lymphatic drainage massage tool.” We managed his account for 12 months and ran every legitimate tactic in our playbook: PPC keyword targeting, ASIN targeting on weaker competitor listings, listing refresh cycles, image upgrades. Each one produced a sales lift. Then the lift faded. We deployed the next thing. Lift, fade, lift, fade, with the baseline trending down each cycle. After a year we told Neil the truth: we could keep being the best at running a generic product, but a generic product would not survive long-term. He needed a differentiated successor or an exit.

The lesson cost Neil a year, and it is the most common research mistake we see across beginners. They evaluate a product on whether it sells today, not on whether it can be defended tomorrow. A generic product in a price-sensitive category does not blow up. It bleeds. For the first 60 to 90 days the metrics look fine, ranking moves up, ACoS lands reasonable, revenue trends right. Then a competitor enters at the same SKU and undercuts your price by 12 to 18 percent. You optimize. Sales recover for two weeks. Another competitor enters. You drop your price to match. Margin compresses. The pattern repeats until you are bidding harder for the same traffic on a product that no longer makes money.

The fix is not better management. It is a harder filter at the research stage. The point of validation is to reject products fast, before they cost you anything. Here are the three traps the 9 steps are built to catch:

Validation is filtering, not fixing. You filter the trap out at step one so you never have to fix it at month seven.

03Steps 1 to 3: build the idea list and prove there is real demand

The first three steps move you from a blank page to a shortlist of products with proven demand. This is the part beginners actually enjoy, so it is also where they cut corners. Do not. A wide, honest idea list is what lets you be ruthless later.

Step 1: Generate a list of 10 to 30 candidates. Two methods feed the same funnel. The first is the Amazon A-Z search method. Go to the Amazon search bar, type a broad one-word category like “kitchen,” “pet,” or “office,” then add a single letter of the alphabet after it. Type “kitchen c” and Amazon auto-populates the most popular searches starting with that combination. “Kitchen curtains” is too generic, but “kitchen cabinet organizer” is a real idea worth checking. Run the alphabet against a few categories and you will have 20 ideas in an hour. The second method is Helium 10’s Black Box, a database filter that scans Amazon for products matching your criteria directly. Use both. The A-Z method shows you what customers actually type, and Black Box shows you the products already winning under those terms.

Step 2: Set your Black Box filters so the database does the first cut for you. These are the exact filters we run:

  1. Category: start with ungated categories like Home & Kitchen, Sports & Outdoors, Patio, Lawn & Garden, or Office Products. Avoid Watches, Jewelry, and Cosmetics or Skin & Hair for now, because they need special approval and paperwork.
  2. Price range: a minimum of $15 and a maximum of $80. High enough for healthy margins after fees, low enough that the first order does not require a huge upfront investment.
  3. Review count: a maximum of 200. This is the most important low-competition filter, because it surfaces products succeeding without thousands of reviews.
  4. ASIN revenue: a minimum of $10,000 per month. This is your demand filter.
  5. Shipping size: check both Small Standard and Large Standard. This keeps oversized items out, and oversized FBA fees will quietly destroy your margin.
  6. Fulfillment: FBA only.
  7. Variation count: a maximum of 3, so you avoid products like clothing with dozens of sizes and colors that are expensive and complex to manage.

Step 3: Confirm the demand is real and beatable. Take each candidate keyword and run Helium 10’s Xray on the whole first page. You are checking three numbers. Average revenue among the top sellers should be at least $10,000 per month. Total market revenue across the page should be at least $200,000 per month, which signals a substantial market. Main keyword search volume should clear roughly 7,000 searches per month, because below that the demand simply is not there. Then the proof-of-concept check: filter Xray to show only sellers with under 200 reviews, and confirm that a real share of them, roughly 20 percent or more, are still making $10,000 a month or more. That is your evidence that a new seller can enter and compete instead of drowning behind established listings.

A worked example of why step 1 has to be wide. Type “kitchen c” into Amazon and you might get “kitchen curtains,” which is too generic and competitive to bother with, so you try another letter. Then “kitchen cabinet organizer” comes up, which looks like a real idea. Run it through Xray and the early signs are good: strong average monthly revenue and an average price around $28, which is exactly the range we want. Then the red flag, an average review count over 5,000. That market is far too saturated for a new seller, so it is a pass. This is the point of the A-Z method. It is not built to find a perfect product on the first try. It is built to generate a list of 10, 20, or 30 ideas quickly, so you can afford to throw most of them away at the validation stage. The seller who only sourced one idea had no choice but to force it. The seller with 30 ideas can be ruthless.

If a product clears step 3, you have demand. You do not yet have a winner. The next three steps are where most candidates die.

04Steps 4 to 6: the competition and quality filters that save your money

Demand alone is a trap. Plenty of high-demand niches will still bury a new seller, and a few high-demand niches will quietly return your product to death no matter how well you execute. Steps 4 through 6 are the filters that caught the mistakes we learned the hard way.

Step 4: Average first-page reviews under 600. Count the review totals across the first-page listings using the Helium 10 extension. If the average sits above 600, the market is too dense for a new listing to climb in a reasonable timeframe. We have a hard rule against entering niches where top sellers carry 1,000 to 5,000 reviews, because review volume is social proof, conversion rate, and algorithmic preference all at once. A new listing with 20 reviews cannot out-convert a competitor with 3,000 on the same search results page, no matter how good your product is. You would spend 12 to 18 months and a large PPC budget trying to catch up. (Our course teaches a slightly looser ceiling of under 700 as the outer limit; in practice we hold the line at 600 for new sellers because the extra room matters when you have no review history.)

Step 5: Top sellers’ average star rating must be 4.5 or higher. This is the filter that cost us a launch before we adopted it, so I will be blunt about why it matters. If the best sellers in a niche cannot get above 4.5 stars on their best efforts, the problem is usually the product category itself, not the sellers. The product type disappoints a chunk of buyers no matter who makes it, which shows up as a high return rate. And returns are a P&L line item exactly like ad spend. On a 30 percent margin product, a 10 to 15 percent return rate means roughly 1.5 of every 10 units sold comes back as lost product cost, return shipping, an Amazon return fee, a lost inventory unit, and downward star pressure. Your ad metrics can look healthy while your bank account tells a different story. Read the star rating before you order, not after.

Step 6: No brand or Amazon dominance on page one. Check that no more than 3 of the top 10 sellers come from the exact same brand, and that Amazon itself is not dominating the first page. If one brand or Amazon controls the results, the market is locked and too hard for a new seller to break into. While you are looking at the page, confirm the majority of first-page sellers are FBA rather than FBM, which tells you the serious money in the niche is running the same model you are.

Three filters, three different ways a high-demand niche can still be a bad bet: too crowded to rank, too returns-prone to profit, too controlled to enter. A product has to clear all three.

05Steps 7 to 9: economics, differentiation, and the structural traps

The last three steps decide whether a validated, beatable niche can actually make you money, and whether you can defend it once you are in.

Step 7: The economics have to work before PPC. The average sale price should be at least $25, which is what gives you enough room to cover Amazon fees and advertising and still keep a real margin. The advertising cost-per-click for your main keywords should be under $1.50, so your launch does not bleed cash on every click. And you are aiming for a net margin around 30 percent after everything. The profitability math we use mid-launch: a product is working on paper when PPC ACoS lands around 20 percent and TACoS lands 10 to 15 percent, as long as the return rate is normal for the category (under about 5 percent). If the numbers only work in a spreadsheet at zero ad spend, the product fails this step.

Step 8: Differentiation potential, the make-or-break check. This is the most important point on the whole list, and it is the one beginners skip. Before you source, you have to find a way to stand out that a competitor cannot instantly copy by dropping their price. Acceptable differentiation gets baked in at the sourcing stage, not bolted on later: a bundle that adds a complementary item, gift-quality packaging where competitors ship in poly bags, a useful accessory that solves an adjacent problem, a genuine material or quality upgrade, or a real design fix for a complaint that runs through competitor reviews. The one constraint is that any add must not push the product into a higher FBA size tier, because that fee increase erases the margin the differentiation was supposed to protect. If you cannot find a differentiation move that survives the size-tier math, the product is generic, and you already read what generic did to Neil.

Step 9: Clear the structural traps. Two quick checks that kill otherwise-good products. First, the product must not be highly seasonal. Pull the sales history graph in Helium 10 and the search trend in Google Trends, and reject anything with massive predictable spikes and dips, because seasonal products tie up your cash for most of the year. Second, keep variations under 3, which you already filtered in Black Box, so you avoid the SKU-fragmentation hell of managing dozens of size and color combinations as a new seller.

That is the 9-step validation. Demand (steps 1 to 3), competition and quality (steps 4 to 6), economics and defensibility (steps 7 to 9). A product earns a sample order only when it clears all nine. The discipline is not in knowing the steps. It is in actually passing on the products that fail one of them, even when you have already fallen a little in love.

06A real validation walkthrough: the keyboard-tray niche we passed on

The 9 steps are abstract until you watch one product run the gauntlet, so here is a real pass we documented. While scrolling Black Box results, an under-desk keyboard tray caught our eye. The main keyword was “under desk keyboard tray.” We took it to Helium 10 Xray to analyze the whole market, the way step 3 says to.

On the surface it looked promising. It met roughly 80 percent of our criteria. The green flags were genuine. When we filtered Xray to sellers with a maximum of 200 reviews, 50 percent of those smaller sellers were still making over $10,000 a month, which is a strong proof-of-concept signal that a new seller could enter. And there was a clear differentiation opening: competitor listings showed an obvious bundling opportunity, like adding a mouse pad with a wrist rest, or selling a full keyboard-and-mouse wrist rest set. Two of the hardest steps, new-seller opportunity and differentiation potential, were checked.

Then the red flags. First, the main keyword’s search volume sat below our 7,000-per-month minimum. The demand just was not deep enough to support a launch. Second, many competitors carried star ratings below 4.5, the exact returns-prone signal from step 5. That combination, a small market full of products that disappoint buyers, is a slow-bleed launch in waiting.

The verdict was a pass. Not because there was nothing good about it, there was. We passed because we hold the criteria at 100 percent, not 80. A product that meets 80 percent of the checklist is not “close enough.” It is a product with two specific reasons to fail, and we already know what both of those reasons do to a launch. Passing on it cost us 15 minutes. Sourcing it would have cost us thousands and a year. That trade is the entire point of the process.

07Two launches, same situation, opposite outcomes

The clearest way to show what these filters protect against is two real client launches that faced the same setup, a category with multiple competitors selling similar core products, and ended in opposite places.

The first was a vegetable chopper in Home & Kitchen. We did almost everything right. We chose a saturated niche but built in real differentiation: a cut-resistant glove, an instruction manual, and premium gift-quality packaging. We optimized the listing weekly for the first month, then quarterly. It hit good ranking and decent revenue, and the early ACoS was healthy. Then it failed slowly over months 4 through 12. The miss was step 5. We had not screened hard enough on top-seller star rating, which sat below 4.5 across the niche. That was a category signal we ignored, and returns ate the margin advantage the differentiation was supposed to create. Even with PPC ACoS around 20 percent and TACoS 10 to 15 percent, the bank account told the real story. The lesson: even a 5-out-of-5 product execution cannot outrun a category that structurally drives returns.

The second was a sports and fitness product for a client named Travis, same kind of crowded category. The difference was step 8 done properly. Travis’s product bundled the core item with a hand band, a wrist band, a resistance band, and a small carrying bag, and crucially, all four bundled items were visible in the main image, not buried in the title or bullets. He priced it 20 percent above the category average and did not flinch. The result: he held first-page rank from launch with no mid-launch drop, ran a month-1 ACoS around 25 percent and a TACoS around 13 percent, and was profitable from the first 30 days. A buyer scanning the search results saw four items in his thumbnail versus one in a competitor’s, and clicked his even at the higher price. The bundle moved him out of the price race and into a value comparison he won.

A few things separate the two outcomes, and they are all in the 9 steps:

There is a third launch worth mentioning, because it is the proof the process works when you apply it fully. After the vegetable chopper, the same client re-entered Home & Kitchen with a different mid-retail product, and this time we ran the full checklist before sourcing: star rating at or above 4.5, FBA-majority first page, no brand or Amazon dominance, average reviews under 600, more than 40 sellers under 200 reviews each making $10,000 a month or more, and under 3 variations. The result was a 30 percent net margin and a month-1 ACoS of 15 percent, and the product is still generating revenue. Same operator, same category type, different research discipline, opposite result.

One more thing the research stage gives you, which most sellers waste: the competitor reading you do here pays off again at launch. When you read first-page listings to check star ratings and differentiation in steps 5 and 8, you are also building a map of which competitors are weak. Later, when you run PPC, you can target a specific competitor’s product page with a sponsored ad. Most teams pick those target listings by sales volume. We pick them by buyer walkthrough: open the competitor’s page as if you were shopping, scroll to the sponsored slot, and ask whether your ad sitting there would pull you away. The best targets are the listings with weak photography and thin content that are still ranking, because the buyer is already half-convinced to leave. We found exactly those on Neil’s account, a competitor selling an 8-piece wood-therapy set with weak images, and the ASIN-targeted ads converted well. The research notes you take to validate a product become the attack plan you use to launch it.

08The 30-minute differentiation audit (the step-8 tool)

Step 8 is the make-or-break check, so it deserves its own workflow. We started running this on every research session last year, and it compresses what used to be hours of manual competitive analysis into about 30 minutes. The model does the pattern recognition. You make the call.

Step one: open Amazon, search your target keyword, and take a full-page screenshot of the first results page with all the visible listings, their thumbnails, their prices, and their review counts. On a Mac, Cmd+Shift+4 then spacebar grabs the whole window cleanly.

Step two: open ChatGPT or Gemini, paste the screenshot, and ask this: “Looking at these competing listings, what differentiation could I add to my version of this product that would NOT increase shipping dimensions but would feel high-value to a buyer comparing options? Keep it under $1.50 per unit at 500-unit volume, and keep it relevant to how the product is actually used. Give me 8 options ranked by likely impact on conversion.” The dimension constraint matters, so state it explicitly. The model will happily suggest something that bumps you into a higher FBA size tier if you let it.

Step three: read the 8 suggestions critically and cross off the impractical ones. Some will be too expensive, some will break Amazon category rules, some just will not fit the product. You will usually have 3 or 4 viable candidates left.

Step four: open the top 5 competitor listings and read their 1-to-3-star reviews. Look for the complaint that repeats. The AI suggestions plus the competitor weakness reading almost always converge on 1 or 2 strong ideas. Those are your candidates.

Step five: message your supplier on Alibaba with the idea and ask for a per-unit cost at 500 and 1,000 units. If the supplier can quote it and the math holds, you have your differentiation. If it kills the margin, drop to the next candidate.

The reason this beats brainstorming alone is that the model is genuinely good at spotting patterns across many competitor offerings and genuinely bad at judging which patterns move buyers or fit your constraints. You supply the judgment, the model supplies the options. That division of labor is the whole trick, and it is the same principle behind every tool we use: the software provides intelligence, the operator makes the decision.

09When to run research yourself vs hand it to an operator

You can absolutely run this process yourself. The 9 steps are not secret, the tools are available to anyone, and a careful beginner who refuses to skip step 5 or step 8 will already be ahead of most of the market. If you are early and learning, do it yourself first. You should understand why each filter exists before you ever pay someone to apply it for you, and our beginner’s roadmap for selling on Amazon FBA walks the full journey around this research stage.

Where people hand it off is when the cost of a wrong pick gets larger than the cost of expert help. A first launch puts $8,000 to $9,000 of cash at risk on the research decision. If your time is worth more spent on your existing business, or you have been burned by a bad pick before, or you simply want a second set of trained eyes on the star-rating and differentiation calls, that is when an operator earns the fee. Our team runs this exact 9-step validation, plus supplier vetting, a Freightklan shipping quote, a competitor audit, and a listing outline, as a single done-for-you research package. It is the same process above, run by people who have applied it across client launches since 2022 and watched it succeed and fail enough times to read the edge cases fast.

Whichever path you take, the discipline is identical. Pass on the 80-percent products. Hold the criteria at 100. The research stage is the cheapest place you will ever fix a mistake, so spend it like it matters, because it decides almost everything that comes after.

Frequently asked questions

What is the Amazon product research process, step by step?

The Amazon product research process is a validation checklist you run before sourcing any product. Our version has 9 steps grouped into three stages: prove demand (generate a candidate list, set Black Box filters, confirm revenue and search volume), check competition and quality (average reviews under 600, top-seller star rating at or above 4.5, no brand or Amazon dominance), and confirm economics and defensibility (price at least $25 with sub-$1.50 CPC, a real differentiation move, and no seasonality or excess variations). A product only earns a sample order when it clears all nine.

How long does Amazon product research take?

Generating a list of 10 to 30 candidates with the A-Z search method and Black Box takes about an hour. Running the full 9-step validation on a single promising candidate takes 30 to 60 minutes, including the differentiation audit. The point of the process is speed of rejection. Most candidates fail in the first few minutes on demand or review-density checks, which is exactly what you want. Passing on a bad product in 15 minutes is the whole value.

What tools do I need for Amazon product research?

You need Amazon’s own search bar (for the A-Z method, which is free) and a product research tool like Helium 10 for Black Box database filtering and Xray market analysis. The Helium 10 browser extension reads review counts, revenue estimates, and star ratings directly on the search results page. For the differentiation step, ChatGPT or Gemini plus a screenshot is enough. You do not need a large stack to start. You need one solid data tool and the discipline to apply the criteria honestly.

What makes a good Amazon product to sell in 2026?

A good product clears all 9 validation steps: real demand (top sellers above $10,000 a month, market above $200,000 a month, search volume above roughly 7,000 a month), beatable competition (average first-page reviews under 600, with a real share of sub-200-review sellers making $10,000 a month or more), healthy quality signals (top-seller star rating at or above 4.5), workable economics (price at least $25, CPC under $1.50, around 30 percent net margin), and a defensible differentiation move that does not increase shipping dimensions. The differentiation is the part that decides whether the product survives once competitors arrive.

Why do I need to check star ratings during product research?

Because the top sellers’ average star rating is a category signal, not just a quality signal. If the best sellers in a niche cannot get above 4.5 stars on their best efforts, the product type itself disappoints buyers, which shows up as a high return rate. Returns are a direct P&L cost: lost product, return shipping, Amazon fees, and downward star pressure. We learned this the expensive way on a vegetable chopper launch that looked healthy for 90 days, then bled margin to returns for the rest of the year. Now star rating at or above 4.5 is a hard filter we apply before sourcing.

Can I do Amazon product research without paid tools?

You can do the first pass without paid tools using the Amazon A-Z search method and reading the search results page manually, but you will be guessing on revenue, review counts, and search volume. A paid research tool turns those guesses into numbers, which is what lets you apply the criteria at 100 percent instead of by feel. For a first product, the cost of one month of a research tool is small compared to the $8,000 to $9,000 a wrong pick puts at risk. Most serious sellers run a paid tool for at least the research phase.

The bottom line

Product research is the cheapest phase of the Amazon business and the one that decides the most. The 9 steps are not complicated: prove demand, confirm the competition is beatable and the category is not returns-prone, and check that the economics and differentiation hold before you spend a dollar on inventory. The hard part is the discipline to pass on products that clear 80 percent of the list, because those are the ones that quietly cost you a year. We run this exact process on every client product, and across 47 client launches since 2022 the pattern is consistent: the launches that passed all nine steps tend to work, and the ones that skipped step 5 or step 8 tend to fail in slow, expensive ways. Run the checklist honestly, hold it at 100 percent, and you will reject your way to a better product than you would ever browse your way into.

Filed under Product Research research Amazon product research process