CPG Market Research: The Playbook for Brands Without an Agency Budget

Most new CPG products fail. CPG is consumer packaged goods, the everyday food, drink, beauty and household products that sell off a shelf. Most fail not because the founders were careless or the idea was hopeless, but because nobody checked, in any serious way, whether shoppers wanted the thing before the money went into production and shelf space. BCG puts the number at 76% of yearly product launches failing, with two-thirds never selling 10,000 units. CPG market research is how you climb out of that 76 per cent, and you do not need an agency, or an agency’s budget, to do it.

This is the playbook I would hand a brand with more conviction than cash. What to research, in what order, using methods you can actually afford, and the one rule that decides where the little budget you have should go.

What CPG market research actually is

CPG market research is the work of finding out what shoppers want, what they will pay, and how they choose, so you can decide what to make, how to package and price it, and how to sell it, before you commit to production or retail. That is the whole job. Everything after this is method.

It helps to separate it from two things it gets confused with. It is not a market-sizing report, the sort that tells you the global snack category will be worth some enormous number by 2030. That is useful for a pitch deck and useless for a decision. And it is not CPG marketing, which is what you do once you have something to sell. Market research is decision research: it exists to make a specific choice less of a gamble.

Practically, it answers four questions, roughly in order. Is there real demand for this? Will the product itself deliver once it is in someone’s hands? How should it look and what should it cost? And once it launches, is it working? A launch that has honest answers to those four is not guaranteed to win. A launch that has skipped them is mostly hoping.

Why so much of it fails, and what that tells you

The failure numbers are worth sitting with, because they point straight at where the cheap wins are. Alongside that 76 per cent, BCG found that 65 per cent of new product launches are now renovations rather than genuine innovations, the highest share in 30 years, even though 83 per cent of CPG firms rank innovation as a top-three priority. As BCG’s Nicol Zhou, who leads the firm’s consumer product innovation work, put it, there needs to be a wholesale reinvention of how these firms innovate to avoid losing more ground to challenger brands. Read that as an opening. The incumbents are playing it safe and getting worse at it. A smaller brand that actually listens to shoppers has room to move.

Now the honest counterweight, because the failure stats get badly overcooked. You have probably heard that 95 per cent of new products fail. NielsenIQ calls that an urban myth, and the academic paper it points to traces how the figure spread through the industry without much evidence underneath it. After reviewing more than 60,000 innovations across five countries and four years, NielsenIQ found that 52 per cent of new items with national distribution grow sales in their second year on the market. Drop the distribution threshold to include everything and it is still 31 per cent. Small businesses land at 46 per cent, close to the big players.

So the truth is somewhere sensible in between. Plenty of launches fail, but success is not a lottery, and the gap between the two is mostly research and execution rather than luck. That is the whole case for doing the work: the misses are largely predictable, which means they are largely avoidable, and the cheapest way to avoid them is to check your riskiest assumptions before the money is committed rather than after.

The agency question: what you are actually paying for

When people say they cannot afford market research, they usually mean they cannot afford an agency, and they are often right. A full-service agency concept test is typically a five-figure project that takes six to eight weeks from brief to final report. You are paying for a representative sample, a defensible methodology and someone senior to stand behind the numbers. For a national retail listing where a wrong call costs hundreds of thousands, that is money well spent.

For most decisions a growing brand faces, it is not. And here is the part worth internalising: more preparation genuinely does correlate with success, but preparation is not the same as an invoice. NielsenIQ found that top innovators spend, on average, four more months in pre-market prep than low performers. Four months of thinking, testing and refining. You can buy a lot of that with interviews and cheap tools and your own discipline.

There is a catch in the other direction, and I would rather name it than pretend cheap research is always fine. When NielsenIQ ran 91 concepts that a manufacturer had previously passed through a low-cost DIY system, it found 62 per cent had been misclassified. NielsenIQ sells the expensive alternative, so read that with a pinch of salt, but the underlying point holds: a bad test is worse than no test, because it gives you false confidence. The real choice is not agency versus nothing. It is rigour versus guessing. You can be rigorous on a small budget if you are honest about your sample and careful about your method.

The playbook: seven decisions, and how to research each on a budget

Here is the spine of the whole thing. CPG market research is not one activity, it is a sequence of decisions, and each one has a lean version and a pay-up version. The skill is knowing which decision you are actually making and matching the spend to the risk.

The three that earn back their cost most often are demand, concept and price, so those are the ones worth doing properly even when everything is tight.

For demand, start with what is free and already public. Search-volume trends tell you whether interest in a category is climbing or dying. Retailer best-seller lists and, more usefully, the review sections under competing products are a running feed of what shoppers love and what they complain about, which is where the gaps hide. Read a few hundred one and two-star reviews of the incumbent and you will have a product brief. Syndicated data from the big measurement firms is excellent and expensive, so save it for when a category is worth real money or a buyer asks for numbers.

For the shopper, the highest-return thing a small brand can do is talk to actual humans. Eight to twelve proper interviews with people in your target audience will teach you more than a thousand-person survey, because you can ask why and keep pulling the thread. Mine reviews and social comments for the language people use, then run a short survey to put rough numbers on what the interviews surfaced. This is the heart of a voice-of-customer programme, and it costs mostly time.

For the concept, the discipline that matters is monadic testing: show each respondent one idea and ask a purchase-intent question, rather than lining up your options side by side and asking which they prefer. Side-by-side flatters whichever concept is loudest; monadic tells you whether an idea stands on its own the way it will have to on a shelf. Keep the purchase-intent scale consistent so you can compare across tests over time. If you want the full method and the tools that run it, I have written a guide to concept testing platforms for exactly this.

For price, do not guess and do not just undercut. Two cheap, well-established survey methods will get you a defensible range: the Van Westendorp price sensitivity meter, which asks four questions about what feels too cheap and too expensive, and Gabor-Granger, which tests willingness to buy at specific price points. Run the price in context, next to the pack and the competitors, because a number in the abstract is not a number anyone actually pays.

Product, pack, name and claim each follow the same logic: a lean test that gives you a directional read, and a paid version for when that single decision carries the whole launch. Sample the product to a recruited group before you trust your own palate. Test pack renders for shelf standout before you print thousands. Put a claim or a name in front of the target shopper before the boardroom’s favourite becomes the label.

One thread runs under every lean method above: your sample. A cheap test is only as good as who you ask, which is where small brands either save the day or quietly fool themselves. The fastest sample is your own email list, but it leans towards people who already like you, so use it for language and warmth, not for a verdict. For a cleaner read, recruit category buyers through a panel such as PickFu, Prolific or Respondent, or intercept real shoppers at an event or in front of the shelf. Whichever you use, screen hard for people who actually buy the category, because fifty of the right shoppers beat five hundred of the wrong ones.

A word on the other column of that table, the pay-up one, because most of its triggers come down to a retail buyer wanting proof, and it helps to know what actually moves one. A buyer is not swayed by a pretty concept score. They want comparable-item velocity, a credible read on how fast the thing will sell per store per week, a concept or claim test on a sample that resembles their shopper, substantiation for anything printed on the label, and ideally a demand signal like rising search interest. If a national listing is the goal, research to that bar or keep your money until you can.

After launch, the clock is faster than you think

Research does not stop at launch, and the window to react is shorter than most brands assume. NielsenIQ found that the divergence between winners and losers shows up as early as the fourth week in market. Yet plenty of teams wait eight weeks or more before they seriously look at performance, by which point the story is largely written. Track velocity, repeat rate and fresh reviews from week one, and treat the first month as the diagnostic period it is. If distribution is strong but sales are soft, your message is not landing. If reviews are glowing but reorders are thin, something in the path to purchase is broken. That is cheap research too, and most brands ignore it.

A worked example: a challenger drink on a small budget

Say you are two people with a functional soda, a strong hunch and a few thousand to spend, not fifty thousand. Here is how the playbook actually runs.

Week one is demand and shopper, and it costs close to nothing. You check search trends for your category and the specific benefit you are selling, confirm it is rising, and read two hundred reviews of the three incumbents, tagging every recurring complaint. Too sweet comes up constantly. You book ten interviews with people who buy functional drinks weekly and ask what they reach for and why, and you notice the same thing the reviews told you: the existing options taste like medicine or like syrup, nothing in between.

Week two is concept and price. You write three positionings around the too-sweet gap and run each as a monadic test to a small online panel of category buyers, with a purchase-intent question. One clearly outperforms. In the same field you run a Van Westendorp block to find the price band where people stop calling it cheap and start calling it expensive. You now have a concept that stands on its own and a price range you did not invent.

Week three is pack. You have a designer mock up two label directions and run a first-click shelf test: here is a crowded shelf, which one would you pick up. One design wins the glance. You have spent a few hundred on panels and a designer, and perhaps three weeks, and you have evidence on the four decisions most likely to sink you.

What you deliberately skip is telling. You do not buy national syndicated data, because you are not pitching a national buyer yet. You do not run eye-tracking, because a first-click test is a good-enough proxy at your stage. You are not trying to do everything a large brand does. You are spending your small budget only where being wrong would be expensive, and being honest that the rest is directional.

Where AI-simulated testing fits, and where it does not

There is a newer option worth understanding, because it changes the maths on the lean column of that table. AI-simulated audiences let you put a concept, pack, price or claim in front of simulated shoppers built to stand in for your target market and get a read back in days, without recruiting a live panel for every question. Used well, it is a fast, cheap first filter: a way to kill the weak ideas and sharpen the promising ones before you spend on live testing or stock.

This is the one place our own tool fits, so I will name it plainly rather than dress it up. With TestFeed you can test an idea against your target audience, a concept, a product, a pack, an in-context price, a claim or an ad, before you commit stock or budget, and get back a purchase-intent read, the shoppers’ reasons in their own words, and a clear next move, in days rather than weeks. It is the job we built TestFeed for, working with challenger brands like Bae Juice and Sol Bevi.

Be clear-eyed about the limits, because they are real. It is a pre-spend, directional signal, not a sales forecast, and not a substitute for a representative study when the stakes demand one. It does not judge taste, texture or smell, so it will never tell you whether the soda is actually nice to drink. Use it to decide which ideas are worth putting real money behind, then validate the survivors with live shoppers and, once you launch, with the till.

What CPG market research cannot do

The fastest way to waste research is to ask more of it than it can give. Three limits are worth holding onto.

It cannot guarantee a hit. It shifts the odds, which over enough decisions is the entire game, but any single launch can still miss for reasons no test would have caught. Treat every result as a probability, not a promise.

It cannot fully close the gap between what people say and what they do. Shoppers overstate how much they will buy, especially when they are being asked nicely in a survey. This is why purchase intent is a comparative and directional measure, better for ranking options and spotting duds than for predicting exact volumes, and why a small, clean sample beats a large, sloppy one. The misclassification problem NielsenIQ found in cheap testing is real, and the defence against it is method and sample quality, not spend.

And it cannot replace distribution and execution. A concept that tests well but has no shelf presence and no marketing behind it still fails. Research tells you what to build and what to say. Getting it in front of people and keeping it there is a separate job, and one worth planning before you launch, not after. If you are at that stage, my product launch strategy guide picks up where this one leaves off.

Frequently asked questions

What is CPG market research?

CPG market research is the process of gathering and interpreting data about shoppers, competitors and categories so a consumer-goods brand can decide what to launch, how to package and price it, and how to sell it, before committing to production or shelf space. It covers demand research, concept and product testing, pack and price testing, claim and name testing, and post-launch tracking. The goal is to replace expensive guesses with evidence.

How do you do CPG market research on a small budget?

Start with the free data you already have access to: search-volume trends, retailer best-seller and review pages, and your own sales. Add eight to twelve customer interviews and a short survey to understand the shopper. Then run low-cost, focused tests only on the decisions that are expensive to reverse: a monadic concept test with a purchase-intent question, a simple price-sensitivity study, and a pack or shelf-standout test. Spend on an agency only where the risk justifies it, such as a national retail listing.

How much does CPG market research cost?

It ranges from almost nothing to tens of thousands. A full-service agency concept test is typically a five-figure project that takes six to eight weeks or more. The same core decision can be de-risked with a lean study for a few hundred to a couple of thousand, using online panels and DIY testing tools. The honest catch is that cheap is not automatically good: quality depends on the sample and the method, not the invoice.

What are the main types of CPG market research?

The main types map to the decisions a launch has to make: demand and category research (is there an opportunity), shopper and voice-of-customer research (who buys and why), concept testing (does the idea appeal enough to buy), product and sensory testing (does it deliver in use), pack and shelf testing (does it stand out and get chosen), price and willingness-to-pay research, claim and message testing, and post-launch tracking. Most brands need a subset, not all of them, chosen by where their biggest risk sits.

Do you need an agency for CPG market research?

No. An agency buys you scale, a representative sample and a defensible methodology, which matters when the decision is high-stakes or a retailer wants proof. For most early decisions a disciplined brand can get a directional read far cheaper and faster with interviews, panels and low-cost testing tools. The real choice is not agency versus nothing, it is rigour versus guessing, and you can be rigorous on a small budget.

The research stack to run this quarter

If you take one thing from this, make it the order of operations. Before your next launch, spend a week on free demand and shopper research: trends, competitor reviews, and ten real interviews. Then run three cheap, focused tests on the decisions that would hurt most to get wrong, which for nearly every brand means concept, price and pack. Track the first four weeks after launch as if the result depended on it, because it does. Save the agency for the one decision a year that is genuinely worth five figures.

The brands that climb out of that 76 per cent are rarely the ones who spent the most on research. They are the ones who asked the right questions in the right order, and who were willing to hear a no while it was still cheap to change their minds.