Last updated: 17 September 2026
Quick Answer: No single company covers CPG market research, so brands buy across four provider classes. Circana, NielsenIQ, SPINS and Worldpanel by Numerator sell continuous measurement; Ipsos, Kantar and Material run custom studies; Curion, Highlight and MMR Research Worldwide go deep on product; Alchemic, Attest, quantilope and Zappi field AI-moderated interviews. Scan and panel data show what sold, not why.
Across five waves of benchmarking since 1990, the five-year new product success rate has barely moved. It ran 58 percent in 1990, 59 percent in 1995 and 2004, 61 percent in 2012, and 59.6 percent in the 2021 survey of 651 firms across 37 countries run by the Product Development and Management Association.
Most CPG insights teams hold a measurement contract, an agency retainer and a testing subscription at once, inside an industry ESOMAR's Global Market Research 2025 report sizes at US$153 billion. All three report on what happened and leave the same gap on why.
What Does a CPG Brand Need From Research That Other Categories Do Not?
A CPG brand needs three things most categories do not: a continuous read on what left the shelf, a launch clock measured in weeks not quarters, and a way to separate demand from distribution.
Channel spread makes one retailer's read misleading. USDA Economic Research Service retail trends data put 2025 food-at-home sales at 56.2 percent through grocery stores, 26.4 percent through warehouse clubs and supercenters, 11.1 percent through other non-food retailers, and 1.2 percent through convenience.
The FDA's proposed front-of-package nutrition rule of January 2025 would require a Nutrition Info box in the upper third of the principal display panel. Pack research has to test comprehension, not appeal alone.
Which Market Research Companies Serve CPG and FMCG Brands in 2026?
Four classes of company serve CPG and FMCG brands, and each sells a different unit of analysis.
Retail Measurement and Consumer Panel Firms
Circana, NielsenIQ, SPINS and Worldpanel by Numerator. Circana is IRI and The NPD Group merged, under that name since 2023. The panel long sold as Kantar Worldpanel combined with Numerator in January 2025 and took the Worldpanel by Numerator name that July.
Custom Research Agencies for CPG Brands
Ipsos, Kantar and Material. Ipsos covers the full CPG decision set including innovation forecasting; Material sells strategy-led research and analytics.
CPG Specialist Boutiques
Curion has run product testing for five decades in its own consumer centers. MMR Research Worldwide specializes in food, beverage, personal and household care. Highlight ships product to a vetted tester community.
AI-Native Research Companies
Zappi automates concept, creative and brand testing and works with a large share of the Fortune 500 packaged-goods roster. quantilope automates advanced quantitative methods, conjoint, MaxDiff, TURF and implicit association tests among them. Attest sells a consumer research subscription with a designated research expert on the account. Alchemic runs AI-moderated interviews on web, WhatsApp and phone.
How Do the Named CPG Research Companies Compare?
The table reads each company's own public materials as of 17 September 2026, sorted alphabetically by company name.
| Company | What it sells | Evidence it produces | Coverage |
|---|---|---|---|
| Alchemic | AI-moderated interviews on web, WhatsApp, phone | Transcripts, voice clips, distributions from one field | 14 markets, USA and UK |
| Attest | Consumer research subscription, brand and creative | Survey data, research expert included | US, Europe |
| Circana | Point-of-sale data plus a receipt panel | Share, distribution, price and promotion response | US, international |
| Ipsos | Designed studies across the CPG decision set | Custom quant and qual, forecasts | Global |
| Kantar | Brand and creative research, tracking, ad testing | Scores against large norm databases | Global |
| MMR Research Worldwide | Consumer and sensory research, food and home care | Sensory profiles, product and pack testing | Global, UK |
| NielsenIQ | Retail measurement, consumer panel, BASES forecasting | Share and distribution currency, forecasts | 90+ countries |
| quantilope | Automated advanced quantitative methods | Conjoint, MaxDiff, TURF, implicit association | Global panels |
| SPINS | Point-of-sale data across natural, specialty and conventional retail | Attribute-level share and trend by claim and certification | US, multi-channel |
| Worldpanel by Numerator | Household and receipt panels tied to shoppers | Trial, repeat, loyalty, source of volume | 65 markets |
| Zappi | Automated concept, creative and brand testing | Normed scores against a consumer-goods database | Global panels |
What Do Syndicated Retail Data and Custom Research Each Answer?
Syndicated retail data answers what happened, at what price, in which store, to which household. Custom research answers why, and what changing something would do.
A scan-based or panel-based read evidences four things well:
- Share and its movement, against a category definition a retailer accepts.
- Distribution, splitting volume change into more stores or more per store.
- Price and promotion response, from one item at many prices.
- Trial and repeat, the launch diagnostic only a household panel produces.
Three things it structurally cannot evidence:
- Why a shopper chose, since the record holds a transaction, not a reason.
- What the pack communicated, including a claim misread rather than ignored.
- What the non-buyer thought, because a non-purchase generates no line item.
Where the Two Reads Disagree
The usual cause is coverage, not error. The National Academies' study Modernizing the Consumer Price Index for the 21st Century is cited because statisticians wrote it while deciding whether to trust scanner data for official inflation measurement. It records that the Nielsen Scantrack data the Bureau of Labor Statistics purchased "omitted convenience stores, bakeries, butchers, smaller grocery stores, warehouse stores, and gas stations."
Panels have measured characteristics too. The USDA Economic Research Service report On the Accuracy of Nielsen Homescan Data matched panel records against a grocery retailer's own data and found quantities recorded more accurately than prices, with mistakes correlated to household demographics.
So the answer is sequencing. The CPG launch research sequence sets out which stage answers which question.
Who Is In the Sample When a CPG Study Runs Across Markets?
Retail measurement sees the channels that report scan data, and panels see the households that enrolled and kept recording. Where most groceries move through small independent outlets, both read a minority of the trade.
A systematic review of informal food retail in low- and middle-income countries, published in BMJ Global Health in December 2025 and indexed on PMC, synthesized 30 studies. In Kenya, Nicaragua, Zambia, Thailand and Mexico, informal outlets hold at least 60 percent of the food retail market. Intercepting that shopper at the counter is fieldwork of its own, covered in the companies that run in-store shopper research.
Reaching the Shopper the Scan Never Saw
A browser-link study reaches people with a browser, a connection and a reason to click. A voice-first shopper who paid cash at a counter needs a different channel.
That is where Alchemic fits a multi-market CPG study. Respondents pick the channel: a web link carrying pack and video stimulus, a WhatsApp conversation with no link and no app, or an outbound AI phone call with consent captured on the first turn. Recruitment runs as managed fieldwork or bring your own across 14 markets including the USA and UK, and in-home use tests run where the pack has to be lived with rather than looked at. The packaging testing platforms comparison covers the alternatives.
When Is a Specialist Boutique or an In-House Team the Better Call?
A boutique wins when the method is the product and the method is physical. An in-house team wins when the question recurs and turnaround beats independence.
- Sensory and formulation work. A trained descriptive panel in a controlled booth settles whether a reformulated biscuit still tastes like the old one, and Curion or MMR Research Worldwide beats any interview.
- In-home use testing at scale. If a product must be used for two weeks before an opinion means anything, the hard part is shipping product to testers and chasing them through the usage period, which Highlight built its business on.
- Retailer-facing numbers. In joint business planning the only number both sides accept is the syndicated one, and no qualitative depth replaces a currency read from Circana, NielsenIQ or SPINS.
The Limits Worth Stating Before You Brief
AI-moderated interviewing is the wrong tool for sensory discrimination, for a claim that must be substantiated to a regulator under a controlled protocol, and for measuring share.
The Insights Association's Code of Standards, which US research firms attest to annually, requires participation to rest on accurate information, so ask how consent and incentives work per market. For software rather than suppliers, start with the comparison of research platforms and the guide to concept testing.