The Best Free Market Research You’re Already Sitting On (And How to Read It in an Hour)

Every business owner wants better data about what customers actually think, want, and struggle with, and the usual response is to go out and buy it, whether that means putting a survey into the field, booking a focus group, or paying a research firm that quotes six weeks and a number large enough to make you flinch. All of that can be worth doing when the question is genuinely unknown and the stakes justify the spend.

What makes it strange is that the best free market research most businesses have is the writing their own customers have already sent them, unprompted, over the course of years. Very few medium-sized companies run a help desk or a ticketing system, but almost all of them have an inbox, a contact form, and a phone that rings, and that is where the material is.

Where Free Market Research Hides in a Business Without a Help Desk

The first place to look is the shared inbox, whether that is info@, sales@, enquiries@, or simply the owner’s own account. Most businesses have years of it, and almost none of it has ever been read as a body of evidence rather than as a queue of things to answer.

After that come the contact form submissions, which usually land in the same inbox and get treated the same way, and the follow-up questions that arrive once a quote has gone out. Both are worth separating from everything else, because the questions people ask before they commit are the clearest statement you will ever get of what makes them hesitate.

Then there is everything that lives outside email, which for a lot of businesses now carries the majority of customer writing. That includes WhatsApp and SMS threads with regular customers, Instagram and Facebook messages, Google Business Profile reviews and the questions posted underneath them, and the search queries in Google Search Console that show the exact words people typed before they found you.

The Source Almost Nobody Reads: Your Sent Folder

Reading what customers wrote to you is the obvious half of this, and reading what your own team wrote back is the half that gets skipped. Your sent folder is a record of every explanation the business has had to give more than once, which makes it a direct map of where your website, your pricing, or your process is failing to explain itself.

Search your sent items for the phrases your team falls back on, things like “just to clarify,” “what that includes is,” or “the reason we.” Anything you have written twenty times is something your marketing should have said once, in public, before the customer had to ask for it.

What to Do About the Phone

For a lot of medium-sized businesses the richest material is spoken, and none of it is written down anywhere. Pretending the text sources cover everything would leave out the single biggest channel, so the honest answer is to go and get it from the people who answer the calls.

Book fifteen minutes with whoever picks up the phone most often and ask them two questions: what do callers ask that still surprises you, and what do you find yourself explaining over and over. Write the answers down in their words rather than your own, because the phrasing is the part that matters.

Why This Beats a Survey or a Focus Group

Surveys and focus groups carry a well-documented problem, which is that people behave differently once they know they are being studied. They soften opinions to suit the room, they pick from the options you handed them, and they tend to say what sounds reasonable rather than what is actually true, and you are paying for the privilege.

A message someone sent you mid-frustration was written with no audience in mind and no reason to be diplomatic, which puts it about as close to unfiltered customer sentiment as you are likely to get. It also costs nothing, because it has already been written and filed.

How to Use Claude to Find the Patterns in an Hour

The reason most owners never do this is volume, since nobody is going to sit down and read four hundred emails on a Tuesday. That constraint has largely gone away, because a tool like Claude can read the whole set in one pass, and the work becomes a matter of asking it the right things in the right order.

Start by exporting a year of messages into a single text or CSV file, and strip out names, addresses, and account numbers before anything goes near a model. Then run four prompts in sequence rather than asking for a general summary, because a summary will launder away the exact language you are trying to collect.

First, ask it to cluster the messages: “Group these messages by the underlying question the customer is asking. For each group, give me the count, and quote three examples word for word without rewriting them.” The verbatim quotes are the whole point, so insist on them at every step.

Then paste in your homepage and your main service page and ask where the language diverges: “Here is our website copy. Compare the words customers use in those messages with the words we use here, and list every place the two vocabularies do not match.” A gap between the two is a translation problem, and no amount of additional ad budget will solve it.

Next, go after assumptions: “What do these customers appear to assume we do, offer, or include, that our website never actually says?” Those assumptions reveal what your brand has implicitly promised, whether or not anyone meant to promise it, and they are usually cheap to fix once you can see them.

Finally, ask the uncomfortable one: “What do we talk about constantly in our marketing that no customer in this set has ever mentioned?” The answer tends to identify where a meaningful share of the marketing budget has quietly been going.

What Claude Will Get Wrong

It will invent frequency if you let it, so treat any number it hands you as a claim to be checked rather than a finding, and make it cite the message behind every pattern it reports. Asking for quotes at each step is what keeps the exercise anchored to evidence rather than to plausible-sounding summary.

It is also a poor judge of what matters commercially, because it cannot tell the difference between a complaint from your largest account and a complaint from someone who was never going to buy. The clustering is the machine’s job, and the judgment stays with the person who knows the customers.

A One Hour Exercise, Once a Quarter

Export a year of messages, run the four prompts, and spend whatever time is left reading the verbatim quotes rather than the summaries. Most people come away with one thing they already suspected and one thing they did not, and the second one is what pays for the hour.

This will not replace deeper research when you have a question your own records cannot answer, but it will show you, in your customers’ own words, where the real gaps are, well before you spend money trying to close the wrong ones. The most useful diagnostic in a business is often just a willingness to read what is already in front of you, carefully and in one sitting.

Frequently Asked Questions

How can a business do market research for free?

Start with the customer writing you already hold, which for most companies means the shared inbox, contact form submissions, social messages, and reviews. Free market research is largely a matter of reading what you already have with a research question in mind rather than a reply in mind. The cost is an hour of attention rather than a line in the budget.

What if we don’t have a help desk or a support system?

Nearly every business holds the same material in a less tidy form, spread across the inbox, WhatsApp, social messages, review sites, and the phone. Not having a ticketing system means the evidence is scattered rather than missing, and pulling it into one file is the only extra step involved. Businesses without a help desk often have richer material, because nobody has ever standardised it into templates.

How do you use Claude for customer research?

Export the messages into a single file, remove personal details, and ask for clustered patterns with verbatim customer quotes rather than a summary. Follow that by pasting in your website copy and asking where your language and your customers’ language diverge, then ask what customers assume you offer that you never state. The sequence matters more than the tool, because a general summary throws away the exact wording that makes the exercise worth doing.

Is it safe to put customer emails into an AI tool?

Remove names, addresses, phone numbers, and account details before you upload anything, and check what your provider’s terms say about training on your inputs. Business and enterprise plans generally offer stronger guarantees than consumer ones, which matters if the messages contain anything commercially sensitive. If the material touches regulated or health-related information, take proper advice before running the exercise at all.

Does this replace surveys and focus groups?

No, and it is not meant to. What it displaces is the reflex to commission research before you have exhausted the evidence already sitting in your own systems. Surveys do their best work once you know which question is worth asking, and reading your own messages is how you find that question cheaply.