How I do D2C market research before buying inventory

Surveys tell me what people say they might do. Product reviews show me what happened after someone paid. Before I spend money on stock, I want to know who actually pays, what disappointed them, and what they are forced to use instead.

Yesterday I picked a category. Today I am trying to understand the customer inside it.

I am not starting with a neat slide that says “women aged 25 to 35” or “urban customers with disposable income.” Those details can help later, but they do not explain why somebody buys on a Tuesday night, after a bad experience, when the current solution stops working.

I want to understand four things: the person, the situation, the struggle, and the current option that is not good enough.

Start with the situation, not only demographics

A useful customer description sounds like a moment in real life.

For example, a first time runner may have shin pain from ordinary shoes and be choosing between an unbranded pair at about ₹3,000 and an imported pair at about ₹12,000. That situation tells me more than a broad age and income segment.

The real question is not only who the customer is. It is what is happening when the customer starts looking for a solution.

I write the first version like this:

A specific person in a specific situation has a specific struggle and is using an imperfect current option.

This is not my final customer segment. It is a research hypothesis. It tells me what to look for, which reviews to read, and which questions I need to ask.

Read one, two, and three star reviews

I open the leading products on Amazon and Flipkart and read the lower rated reviews first. I am not trying to prove that every product in the category is bad. I am trying to find repeated friction.

I look for complaints about:

• fit or sizing
• quality or durability
• taste, smell, texture, or comfort
• unclear instructions
• packaging and delivery
• missing features
• price compared with the result
• support after the sale

A five star review tells me what customers value. A one star review may tell me what made the product unusable. A two or three star review is often useful because the customer saw some value but still found a problem.

One complaint is a story. A repeated complaint across products is a pattern worth investigating. I also compare the product promise with the customer experience. A brand may promise performance, while customers talk about pain, confusion, or poor fit. That gap may be an opening.

How I collect the notes

I keep one row for each useful review. I do not copy every sentence. I capture the product, rating, customer situation, complaint, current workaround, and source link.

The table is illustrative. The important point is to capture the situation and workaround, not only the complaint.

Mine public conversations

Marketplace reviews are useful, but they are not the whole market. I also search Reddit, Quora, YouTube comments, and comments under large brand posts.

I search for phrases such as:

• What did you buy instead?
• Why does this product not work for me?
• Is it worth the price?
• What do you wish it did better?
• What are you using because the proper option is too expensive?
• What happened after you tried it?

I want repeated customer language, not one dramatic story. Customers may not use the same words that marketers use. A brand may talk about performance, but the customer may talk about pain after thirty minutes. A brand may talk about premium ingredients, but the customer may talk about whether the product fits into a daily routine.

The customer language is usually more useful than the brand language because it shows how people experience the problem in their own words.

What can make review research misleading?

Public reviews are useful, but they are not a scientific survey.

People with very strong opinions may be more likely to post. Unhappy customers may be over represented. Some reviews may describe the same issue. A complaint may be caused by delivery rather than the product itself. A product may have changed since an old review was written.

So I label my evidence carefully:

• Observed: a customer wrote this specific complaint.
• Repeated: similar complaints appear across several products or sources.
• Directional: the pattern suggests a possible problem, but I need more evidence.
• Unknown: I do not yet know how common or commercially important it is.

This prevents me from turning ten loud comments into a claim about an entire market.

Turn complaints into unmet needs

A complaint is not automatically a product opportunity. I translate it into a possible need only after asking three questions:

1. Does the complaint repeat?
2. Does it matter enough to change a purchase decision?
3. Can a new brand solve it without creating a bigger cost or quality problem?

For example, “the bottle leaks” may become “people need packaging that can travel in a work bag.” “The instructions are confusing” may become “first time users need a simple routine and a clear result expectation.”

The need is more useful than the complaint because it tells me what I might test in a product, bundle, page, or service.

How I use the 200 anecdote exercise

For the Day 3 example, I want to work with 200 public anecdotes from reviews, forums, comments, and product discussions.

I will not treat 200 anecdotes as a scientific survey. They may be biased, repeated, or incomplete. They are useful for finding language, recurring complaints, and possible customer situations.

I first read a sample myself and create rough complaint groups. Then I give the cleaned notes to Claude and ask it to cluster the complaints by frequency, situation, and seriousness.

Grab the prompt I actually use for this below, in the download button.

The manual step comes first because I need to know what the tool is grouping. If the notes are messy, the output can look organised while still being wrong.

Turn the insight into a cheap test

Once I have the top unmet needs, I do not immediately place a large order. I choose the cheapest test that can change my mind.

• Create a simple page explaining the proposed solution and measure qualified signups.
• Speak to people who have the problem and ask what they tried before.
• Offer a small preorder only if the product, delivery timing, and refund terms are clear.
• Test a small batch with a narrow customer group.
• Show two versions of the promise and ask which one is clearer, without treating preference as a sale.

The test should match the unknown. If I do not know whether the problem is urgent, I need conversations or a paid action. If I do not know whether the product can be made consistently, I need samples and quality checks.

What this research cannot prove

This process can reveal repeated problems, customer language, current alternatives, and possible unmet needs.

It cannot tell me the exact market size. It cannot prove that people will buy my product. It cannot tell me whether a supplier can make the product well. It cannot replace a landing page test, a preorder, an interview, or a small batch.

It also cannot remove the need for judgement. If a complaint is frequent but the proposed solution is expensive, fragile, difficult to ship, or hard to explain, I still have an operating problem.

Evidence note: The 200 anecdote exercise is a research example, not a scientific survey. Public reviews and comments can over represent unhappy customers and may contain duplicates.

What repeated complaint in your category would you investigate before spending money?