An in-home usage test (IHUT) is a CPG research method in which consumers use a product at home under normal conditions and report on their experience. Run with research agents, an IHUT collects photo and video evidence of use, prompts entries at each usage occasion, and probes the need state behind each use, then synthesizes findings across households.
Summary:
|
About 30,000 new CPG products launch each year, and NIQ's analysis of launches found that roughly 30% of U.S. launches maintain or grow their sales over the first two years. A product can test well and still miss, because surveys and lab sessions rarely show how it is really used at home.
This guide shows how research agents change that. Instead of a kit and a recall survey, the agent checks in at every use, collects proof of use, and asks why. If your team works on consumer insights, the framework below applies across categories and markets.
What Changes When Research Agents Run an In-Home Usage Test
An IHUT is a method in which consumers use a product at home under normal conditions and report on the experience. Our guide to in-home product testing covers the fundamentals, including kits, shipping, and logistics.
A traditional IHUT sends a product kit, waits, and ends with a recall survey. An agent-run IHUT checks in at each use, probes what it sees, and synthesizes findings household by household.
Dimension | Traditional IHUT | Agent-run IHUT |
Data timing | End-of-period recall | At each usage occasion |
Evidence | Self-reported answers | Photo and video entries plus answers |
Occasion capture | Reconstructed from memory | Logged as it happens |
Drop-off handling | Noticed at the end | Flagged and nudged during the test |
Time to insight | After fieldwork closes | Rolling, with per-household synthesis |
Adoption of AI is already broad. A McKinsey survey of 1,491 participants fielded in July 2024 found that 71% of organizations used gen AI in at least one business function. Moving fieldwork to agents is a natural next step for insights teams.
An AI research agent can run the check-ins, remember each household's earlier answers, and keep probing as the test unfolds.
Teams comparing consumer research platforms for this work should confirm the platform supports photo, video, and event-triggered prompts.
The Three Layers of IHUT Data: Evidence, Occasions and Need States
Every IHUT study can be read at three layers:
Evidence: what happened. The product used, how it was prepared, how much was used.
Occasion: when, where, and with whom it happened.
Need state: why this product was chosen at that moment.
The layers connect into one view of usage. Evidence proves a use occurred. The occasion places it in a routine. The need state explains the choice. Remove any layer and you are left guessing: evidence without a need state tells you what people did but not what to change.
End-of-test recall surveys flatten all three. People remember the typical use, not the awkward or unusual one, and they rationalize why they chose the product. This is the same gap explored in the say-do gap in CPG research.
Capturing Usage Evidence in Real Homes
Evidence is anything that shows how the product was actually used. In an IHUT it typically includes:
Unboxing and first impressions
Preparation steps
Application or serving
Portion or amount used
Leftovers and waste
Storage between uses
The agent requests photo and video at these key moments and probes what it sees. If a photo shows a larger portion than the instructions suggest, the agent can ask why. If the pack is stored somewhere unexpected, it can ask about that routine.
Short video entries are especially useful because they show the real kitchen, bathroom, or laundry room, not a described version of it.
Products are often shared. The agent can invite other household members to add their own entries and tag who used the product, so evidence reflects the household, not only the person who signed up.
Verifying Compliance and Data Quality
Good data starts with confirming that the right product was used as instructed and in the assigned order. The agent can ask for a photo of the product code or pack before each use, then check it against the participant's assigned cell.
It can also flag problems early:
Non-use: no entries after the expected usage window.
Backfilled entries: several entries logged at once, with matching timestamps.
Fraudulent participants: reused images, generic answers, or mismatched household details.
Our guide to detecting fraud in AI-moderated studies goes deeper on these signals. Flagged cases should go to a researcher, who decides whether to follow up, replace, or exclude the household.
Mapping Usage Occasions
An occasion is the context around a use. Capture these attributes for each one:
Time of day
Location
Who is present
The trigger that prompted use
What else is consumed or used alongside it
Event-triggered check-ins capture occasions as they happen, not from memory. The agent prompts the participant at each use, so you can see how often the product is used and how occasions vary within and across households. A morning use by a parent in a rush is a different occasion from a weekend use by the same person.
Occasions are also how category growth is read at scale. Kantar's Appetite for Growth research analyzes about 2.5 million consumption occasions a year from a proprietary usage panel across Brazil, Mexico, the UK, Spain, and France. A panel shows the pattern across a market. An IHUT shows the why behind one product's occasions.
Grouping households by occasion is the basis of occasion based segmentation, an approach related to behavioral segmentation.
Designing Occasion Prompts by CPG Category
Prompts should fit the category. Each one should be short enough to answer in the moment:
Snacks and beverages: where were you, who were you with, and what were you doing when you reached for it?
Beauty and personal care: which step of your routine was this, and how did it fit with your other products?
Household cleaning: what were you cleaning, and what did you try before this product?
Pet food: how did the pet respond, and how did it fit the feeding routine?
Prompts also differ by type of use. Planned routines call for scheduled check-ins, such as every morning wash. Spontaneous use calls for event-triggered prompts that the participant can start themselves.
Run this way, an IHUT is a CPG diary study with a product kit attached.
With AI agents, an FMCG diary study gains live probing, as covered in AI-moderated diary studies.
Uncovering the Need States Behind Each Use
A need state is the functional and emotional goal a person has at the moment of use. Two people can use the same product on the same day for different reasons, and the product wins or loses on how well it meets each need.
The agent probes "why this product, why now" at each occasion, then follows the answer. If someone says they wanted something quick, the agent can ask what quick meant and what they would otherwise have used.
The idea has a long history. In a well-known example from Harvard Business School's Working Knowledge, a one-day observation at a fast-food restaurant found that about 40% of milkshakes were bought first thing in the morning, mostly by commuters taking them to go. The product was hired for a job, and that job shaped what a good milkshake was.
Across households, the agent clusters similar need states. Researchers then name and size the clusters, such as "energy boost", "comfort", or "easy for the kids".
From Occasions and Need States to Demand Spaces
Pairing occasions with need states defines demand spaces. A demand space is a combination of a moment and a motivation where a brand can compete. An occasion by need state matrix makes the pattern visible. This one is illustrative, for a snack:
Occasion | Energy boost | Comfort | Reward | Convenience for kids |
Morning commute | Strong | Weak | Weak | Weak |
Afternoon slump at work | Strong | Moderate | Moderate | Weak |
Evening TV | Weak | Strong | Strong | Weak |
School lunchbox | Weak | Weak | Weak | Strong |
Cells with strong signals and little competition point to opportunities for positioning and claims. Cells where your product underdelivers point to formulation or pack changes.
Designing an Agent-Run IHUT
Start by mapping each study objective to a layer. If the goal is to confirm the product is used as intended, you need evidence. If it is to find growth moments, you need occasions. If it is to choose a positioning, you need need states.
Then design the cells and the check-in cadence:
Monadic: each household uses one product. The agent checks in at every use of that product.
Sequential monadic: each household uses products one after another. The agent resets with a new baseline for each product.
Randomized order: products rotate across households to cancel out order effects. The agent follows the assigned sequence and confirms it.
Recruit by category usage and household composition, since a family of five uses a household product differently from a single person. Close the test with liking and purchase intent measures so qualitative patterns can be checked against numbers. The wider CPG research methods and timelines show where an IHUT fits in a launch plan.
Keeping evidence, occasions, and need states together in one consumer insights platform makes cross-household analysis far easier.
Setting Test Length by Product Usage Cycle
Length should follow how the product is used. Single-occasion products, such as meal kits and snacks, can be tested over a short window, because each use is a complete experience. Repeat-use products, such as shampoo and detergent, need more time to show repeat usage and routine fit.
Length also affects the data you get:
Wearout: liking can fall as novelty fades.
Habit formation: routines take time to settle.
Drop-off: longer tests lose more households, so keep prompts brief.
Agent-Run IHUT Use Cases Across CPG Categories
The method fits many categories. CPG and FMCG brands can apply it wherever real-world use matters more than a first impression.
Food and beverage. Capture preparation, formulation fit, and serving occasions. Agents can ask whether the recipe worked, who ate it, and what it replaced.
Beauty, personal care, and household. Test routine fit, performance claims, and wearout over repeated use. Photos can show application and results over days.
Pet food and wellness. Owners evaluate on behalf of the pet, so the agent probes what the owner saw and how closely the regimen was followed.
Turning IHUT Findings Into Product and Claims Decisions
The first move is from per-household narratives to cross-household patterns. AI thematic analysis can group entries, photos, and probes into themes while keeping each household's story traceable.
Then tie findings to decisions: reformulation, pack size, claims, and occasion-led positioning. If households repeatedly use less than the pack assumes, or keep products in a different place than planned, CPG packaging choices are worth revisiting. Claims validation works the same way: check whether the benefit people describe matches the claim you want to make.
Finally, validate qualitative patterns against the closing quantitative measures. If a need state looks strong in the entries but liking and purchase intent do not reflect it, treat that as a finding, not a mismatch to smooth over. The wider case for this approach is set out in AI-moderated research for CPG brands.
Limits of Agent-Run IHUTs and Where Researchers Step In
Agent-run IHUTs have limits:
Self-reported sensory detail: taste, aroma, and texture are hard to describe, and photos cannot capture them.
Category and shipping constraints: frozen, fragile, or regulated products may need special handling or may not suit a home test.
Interpretation: need states are judgments, and an agent's synthesis needs validation.
Researchers should name need states, review clusters against raw entries, and decide what counts as non-compliance. That human oversight is covered in human-in-the-loop AI-moderated research.
It also protects the investment. Gartner predicts that over 40% of agentic AI projects will be canceled by the end of 2027 because of cost, unclear value, or weak risk controls. Clear boundaries for the agent help avoid that outcome.
Reading Emotional Response at the Moment of Use
Ratings often miss the first reaction. A short video entry at first use can show a smile, a wince, or hesitation that a participant would not mention, or would soften afterward.
Decode by Entropik supports agent-run IHUTs with this kind of measurement. It is trusted by 150+ global brands and backed by 17 patents.
Facial coding reaches 90%+ accuracy across 62 facial expressions, so emotional response to the product can be read at first use and compared across households.
Eye tracking reaches 96% accuracy and measures attention on pack and instructions during unboxing. With 70+ languages supported, the same approach works in multi-market tests.
Frequently Asked Questions
1. How do research agents run an in-home usage test?
The agent guides participants through setup, prompts a check-in at each use, requests photo and video, and probes what it sees. It then synthesizes findings per household and across the sample, with researchers reviewing the results.
2. How long should an in-home usage test last?
It depends on the product's usage cycle. Single-occasion products need a short window, while repeat-use products need longer to show routine fit, wearout, and habit formation. Longer tests lose more households.
3. How many households do you need for an IHUT?
Set cell size by the decisions you need to make. Qualitative patterns can emerge in smaller cells, while liking and purchase intent comparisons need larger ones. Confirm sample sizes with your statistician before fieldwork.
4. What is the difference between a usage occasion and a need state?
A usage occasion is the context of a use: when, where, and with whom. A need state is the functional and emotional goal behind it, the reason this product was chosen at that moment.
5. Can an in-home usage test be combined with a diary study?
Yes. An agent-run IHUT is closely related to a diary study, with a product kit and event-triggered prompts at each use. Many teams run both in one program.
6. How do you verify that participants actually used the product at home?
Ask for photos or video at key moments, confirm product codes against the assigned cell, and check timestamps. Flag non-use, backfilled entries, and suspicious patterns for researcher review.
7. What products are suitable for in-home usage tests?
Most food, beverage, beauty, personal care, household, and pet products suit home testing. Frozen, fragile, or regulated products may need extra handling.
8. Can agent-run IHUTs run across multiple markets and languages?
Yes. Agents can run check-ins and probes in many languages, so a single design can run across markets. Researchers should still review translations and local context.
