POEMS Framework: Five Lenses for In-Context Observation

Interviews and session recordings often capture what people say they do. In the real setting, tools fail, rooms are noisy, help is hidden, and a prompt is easy to miss. Researchers then write notes about the interface and wonder why the same drop-off repeats.

Selective attention is the usual failure. One dramatic moment (a confusing modal, a frustrated comment) fills the notebook. The objects, the room, the messages, and the support system stay invisible. Later, the team cannot tell whether the problem is copy, hardware, Wi-Fi, or a missing human help path.

POEMS is a way to force the notebook to cover those five parts of the scene. It is not a promise that structured notes equal insight. It is a habit of finishing the observation before declaring a pattern.

POEMS was developed by Patrick Whitney and Vijay Kumar at the IIT Institute of Design. It is an observation structure, not a research-stage model.

The five lenses:

• People: who is involved, what they are trying to do, how they behave under the conditions you can see
• Objects: tools, devices, products, and other artifacts that enable or constrain the action
• Environment: the physical or digital setting, including comfort, attention, noise, and layout
• Messages: prompts, sounds, instructions, gestures, and other signals the person receives
• Services: the support system around the experience (onboarding, help, staff, policies)

Each lens is a clue. Together they form a picture of what users are doing in that setting. POEMS does not pick whether you should interview, test, or survey. It does not replace the later job of deciding which findings matter for the product.

A useful POEMS pass is a complete field record plus a cautious synthesis, not five empty headings.

1. Set the observation

Name the activity, the setting, and who you will watch. One activity per pass (onboarding at the gym, checkout at the store) keeps the five lenses comparable.

2. Record all five lenses

For each participant or session, write what you saw under People, Objects, Environment, Messages, and Services. Prefer observable actions over inferred motives. If a lens was not visible, say so.

3. Separate observation from interpretation

Keep “the help link is three menus deep” distinct from “users feel abandoned.” The first is evidence. The second is a hypothesis.

4. Synthesize across lenses

Look for repeats that only make sense when two or more lenses combine (noisy room plus long on-screen copy plus no staff). State those as patterns to test, not as proven product requirements.

When another lens fits better, or when you need a complementary view, these frameworks do different jobs. They are not interchangeable labels for the same question.

POEMS is the lens for complete in-context notes. Other methods help when the question is which research stage, which journey stage, or how a product should form a habit.

• Field studies. You need context in the real world, not only what people claim in a remote interview.
• Job shadowing. Workflows, workarounds, and habits show up in the room more clearly than in a self-report.
• Contextual inquiry. You want observation and questioning to stay consistent across participants.
• In-context observation. Environment, tools, and signals shape behavior as much as the interface.

Example: Fitness app onboarding in a gym lounge

A team wants to understand why new users abandon onboarding. They watch people in a gym lounge with their phones, not in a lab.

A possible POEMS picture (illustrative, not a measured conversion study):

• People: New members pause on the goal-setting screen. Some look around for a person to ask.
• Objects: Phone, nearby equipment that pulls attention, a watch that fails to pair.
• Environment: Noise, unstable Wi-Fi, people standing rather than sitting.
• Messages: Long instruction text, other-app notifications covering the flow.
• Services: No staff who know the app. Help is buried in a menu. A trainer, when asked, cannot assist.

Implication: The drop-off may not be “users do not want goals.” It may be pairing, noise, and missing support colliding with a long screen. The team still needs to decide what to change. POEMS gave them a complete scene instead of a single quote about confusion.

What POEMS can help with

• Keeping field notes complete across five parts of the scene
• Reducing selective attention to the loudest interface moment
• Making environment, objects, messages, and services visible next to people
• Giving later synthesis raw material that still contains context

What POEMS cannot replace

• A research-stage chooser. The Research Funnel matches methods to exploratory, strategic, or tactical work.
• A journey map. Experience stages over time are a different cut than five lenses on one scene.
• Surveys or lab usability scripts. Those methods answer different questions.
• Synthesis and product decisions. Structured notes do not decide which findings matter.

Honest scope: POEMS structures observation. It does not replace method choice or interpretation.