Product Manager Interview: Design a Medication Reminder for Older Adults

Prompt and Applicable Context

Design a medication-reminder product for older adults who live alone, take several long-term medications, and occasionally miss a dose. How would you segment users, identify the core problem, define the MVP, manage medication-error and caregiver-collaboration risks, and validate the product?

This is a product-design question for product managers. It tests whether a candidate can select a user, problem, and boundary from an ambiguous, high-risk space, then turn the choice into a testable minimum product. Current public PM-interview material still describes product sense, product design, product cases, and product thinking as interview rounds or evaluation dimensions. That evidence establishes the format's present relevance. It does not establish that any company asks this exact prompt or support an interview-frequency claim.

Use this fictional practice scenario: the primary user lives alone, takes 4 long-term medications across 3 scheduled medication windows each day, uses a smartphone, lacks confidence with complex setup, and occasionally forgets a dose. The user owns the product. A caregiver may participate only after the user gives explicit consent.

The first release helps the user store a medication plan reviewed by the user or a qualified professional, receive reminders, record a self-reported status, and notify a caregiver when authorized. It does not diagnose, recommend a dose, decide whether a missed dose should be taken later, or treat a tap on “Taken” as proof of clinical adherence. Medication-dispensing hardware, care for severe cognitive impairment, and high-risk medications that require real-time clinical monitoring are outside the first release.

What the Interviewer Evaluates

The first signal is whether the candidate segments “older adults.” A person who lives independently and manages medication, someone who needs family assistance, and someone with severe cognitive impairment have different goals, permissions, and product forms. A weak answer puts all three into one persona. A strong answer selects a primary group and states which users need hardware, professional care, or a different product.

The second signal is causal diagnosis. A dose may be missed because the user forgot, could not read a label, had an outdated plan, manages too many medications, fears side effects, cannot afford the medication, or intentionally stopped. Reminders directly address only part of unintentional nonuse. If cost or side effects drive the behavior, sending more notifications will not solve it and may cause the user to disable every reminder.

The third signal is the safety boundary. The product must distinguish “a reminder was delivered,” “the user acted,” and “the medication was actually taken.” It must also handle stale schedules, duplicate confirmation, uncertainty about whether a dose was taken, caregiver overreach, and notification fatigue. A strong answer does not let the app invent clinical instructions such as “take one now” or “double the dose.” The prescription instructions, pharmacist, or another qualified professional owns that decision.

Finally, the interviewer looks for a complete product loop: research in the real context, a state model, an MVP, explicit exclusions, user-outcome metrics, and safety guardrails. Shipping the feature, enabling notifications, or accumulating taps on “Taken” is not sufficient evidence of success.

Questions to Clarify Before Answering

• Who are the primary user, payer, and caregiver? An independent older adult needs control and comprehensibility. If an adult child pays while the older adult uses the product, payment does not grant default access to all data. An institution-owned workflow would instead emphasize task assignment, audit, and professional operations.
• Where does the medication plan come from, and who updates it? Manual entry, label scanning, pharmacy import, and health-system synchronization have different error modes. If the first release uses manual entry, the user or a trusted helper must review the name, dose, purpose, timing, and special instructions, and the product must show when the plan was last updated.
• What does “missed” mean in the product? No response 10 minutes after a reminder, exceeding a scheduled window, and explicitly selecting “Skip” are different states. Thresholds need personalization, and no tap cannot be asserted as no medication.
• Which medications and risks are in scope? If a medication has a narrow timing window or dangerous missed-dose handling, software reminders may be insufficient. Clinical, pharmacy, and compliance owners should define the workflow first, or the pilot should exclude that population.
• What may a caregiver see, and when are they notified? Seeing only unresolved status, viewing the full medication list, editing the plan, and contacting the user are separate permissions. The recipient, fields, duration, revocation, and emergency rules change the product.
• Which devices and reminder modes does the user rely on? A phone-only user, a person who also uses a watch or smart speaker, unreliable connectivity, and differences in vision, hearing, and dexterity all change the notification and interaction design.
• What is the success objective? An initial round can test whether users configure the product correctly and make timely self-reports. Self-reported product data alone cannot establish reduced medical risk or improved clinical outcomes.

30-Second Answer Framework

“I would scope the primary user to an older adult who can manage medication independently, uses a smartphone, lives alone, and occasionally forgets a dose. I would exclude severe cognitive impairment and medications requiring real-time clinical monitoring. Through in-home interviews and task observation, I would separate forgetting from unreadable labels, an incorrect plan, and intentional stopping; the first release would address forgetting and plan management. The MVP would store a reviewed medication list and use personalized sound, vibration, and large-text reminders across the three scheduled windows. The user can choose Taken, Remind Me Later, or Skip. Only after an unresolved event exceeds the user's chosen window would an explicitly authorized caregiver be notified. The app would never recommend a replacement dose or dose change. I would validate setup and confirmation with a prototype, then run a limited pilot. The primary metric would be timely self-confirmation within the scheduled window, with guardrails for incorrect plans, duplicate confirmation, accidental-action reversal, disabled notifications, support requests, and privacy complaints.”

This skeleton establishes the user and boundary before describing research, the MVP, risk, and validation. A full answer must explain why the segment was selected, how each state changes, and what the metrics can and cannot prove.

Step-by-Step Deep Answer

Step 1: Express the goal as a user outcome and a product boundary

A useful goal is: “Help an older adult who still manages medication independently remember the plan with less cognitive load, confirm their own action, and deliberately seek help when needed.” Three limits are intentional:

1. “Manages medication independently” excludes people who require continuous supervision or direct dispensing by another person.
2. “Confirm their own action” identifies the record as self-report, not an observed ingestion.
3. “Deliberately seek help” preserves user control instead of turning a caregiver into a default monitor.

The product should not claim that it has already improved adherence. Clinical adherence is also shaped by side effects, cost, understanding, daily routine, and willingness to continue treatment. The first release can measure whether setup is correct, reminders are understandable, users produce credible self-reports within an agreed window, and escalation follows their permissions.

Step 2: Choose a primary segment and state who is not served yet

Start with three groups:

• Group A: independently manages medication but occasionally forgets or confuses a time window;
• Group B: usually manages independently but needs remote family assistance;
• Group C: has severe cognitive, visual, or motor impairment and needs direct management by a caregiver or device.

Choose Group A for the first release and offer limited, optional caregiver collaboration for Group B. Software reminders and simpler interaction directly affect Group A's problem, and the user can review and reverse actions. Group C's core need is not a reminder screen. It may require a locked dispenser, professional care, in-person verification, and a stricter accountability model. Including Group C in the same MVP would create false reassurance.

“Older adult” still needs segmentation by medication complexity, digital confidence, sensory and motor ability, living arrangement, and access to a trusted caregiver. The initial persona is a research hypothesis, not a substitute for real users.

Step 3: Find the missed-dose journey in the user's real environment

Interview and observe target users where they actually organize medication. With consent, separately interview caregivers, pharmacists, or relevant professionals. Begin with routines: where medication is kept, how time is recognized, what happens during travel, who updates a changed plan, and what the user does after forgetting. Do not ask only which features people want. Observe whether labels are readable, packages look alike, the phone is muted, and the routine is anchored to meals or sleep.

Place the findings into a task chain:

obtain the current plan → identify the medication → receive a prompt when due → decide and act → record status → seek help for an exception → update after a plan change

Each break requires a different intervention. If the user lacks a trustworthy current plan, a precise reminder amplifies an error. If the user skips because of side-effect concerns, the need is professional communication. If a cooking routine masks an audio alert, multimodal and delayed reminders may help.

Research should produce prioritized problem statements, not a feature wishlist. For example: “An independently managing user who lives alone misses a scheduled window while busy or away from home, then cannot determine whether the medication was forgotten or already taken, so they wait, repeat an action, or call a relative.”

Step 4: Use a state model so a notification is not treated as an outcome

Each scheduled medication event moves through at least these states:

scheduled → due → reminded → taken / remind later / skipped / unresolved → resolved or escalated for help

“Reminded” means the system attempted a notification. “Taken” is a user self-report. “Unresolved” does not automatically mean the medication was missed. Every state should retain the time, plan version, and actor, and an accidental tap should be reversible for a short period. When a plan changes, old events must stop using stale instructions; the new plan should display its effective time and review source.

The model also controls caregiver behavior. Escalation occurs only when an event is unresolved beyond the user's chosen window and that user has authorized the caregiver and notification scope. The caregiver receives “not yet confirmed; please contact the user,” not “the patient definitely missed a dose; tell them to take it now.”

Step 5: Select an MVP from three alternatives

Compare three paths:

1. Reminder-only list: inexpensive and fast, but it cannot represent uncertainty, plan versions, or a help path.
2. Software with confirmation and consented collaboration: covers the primary journey but still relies on self-report.
3. Connected dispenser or deep pharmacy integration: provides stronger dispensing or plan evidence but adds hardware, coverage, operations, and accountability.

Choose the second path for the first release and compress it into one testable journey:

• Create a medication list containing at least the name, strength, purpose, instructions, scheduled window, information source, and last review time.
• Activate the plan only after review by the user or a trusted helper.
• At the scheduled time, use large text, strong contrast, and cues that do not rely on color alone; let the user choose sound, vibration, and reminder cadence.
• Give each event three primary actions: Taken, Remind Me Later, and Skip, with an undo path.
• When the user is uncertain or chooses Skip, show a pre-saved professional help route instead of generating missed-dose advice.
• Notify an authorized caregiver only after the personalized unresolved window expires.
• Let the user pause, edit, and revoke caregiver permissions, with a clear record of every change.

Do not put photo-based pill identification, inferred dosing, open-ended AI medical advice, full hospital integration, or a connected dispenser into the first release. Those may have value, but they add accuracy, compliance, device, and operating uncertainties before the minimum pilot answers whether a simple software journey is understood and used reliably.

Step 6: Turn safety, accessibility, and privacy into product rules

Safety rules should appear directly in the interaction:

• If a user marks Taken and later becomes uncertain, allow reversal with an audit record and show how to contact a pharmacist or qualified professional. The product does not decide whether another dose should be taken.
• A plan change must show who changed it, when it takes effect, and who reviewed it. Old reminders become invalid immediately.
• Duplicate confirmation for the same event should be blocked, explained, and recorded rather than producing two Taken records.
• The medication list should include prescriptions, over-the-counter medicines, vitamins, and supplements so the user can maintain a complete record, while the product still avoids making its own interaction judgments.
• When travel or daylight-saving changes the time zone, ask whether the plan follows local time or the original zone. Do not silently move a high-risk schedule.

Accessibility cannot stop at “use a larger font.” Age-related changes may affect contrast perception, fine motor control, hearing, short-term memory, and attention. The first release therefore needs large targets, clear hierarchy, concise language, combined audio and haptics, screen-reader support, and user-selected reminder methods and volume. Color may reinforce status but cannot be the only distinction.

Use least-privilege sharing. The user selects each caregiver, visible field, and notification type. The full medication list is not shared by default. Consent can be revoked at any time, and a caregiver cannot change the plan without a separate confirmation path. If no caregiver is authorized, the product remains a personal reminder with a professional-help route; “safety” does not justify bypassing consent.

Step 7: Validate in stages instead of jumping to retention

Begin with task testing on a clickable prototype. Can the user create 4 medications, set 3 windows, understand the three actions, reverse an accidental tap, find help, and revoke caregiver access? Observe the task rather than replacing evidence with “Would you use this?”

Next, run a time-limited pilot with real reminders. The primary metric is the percentage of valid scheduled events for which the user makes a self-confirmation within their chosen window. It is a process proxy, not proof of medication ingestion. Also measure:

• the percentage of created plans that are reviewed and successfully activated;
• the percentage of unresolved events eventually resolved by the user or authorized caregiver;
• the number of reminders required before a response and use of Remind Me Later;
• whether users update the plan promptly after a prescription change.

Safety and experience guardrails include an incorrect plan being activated, duplicate confirmation, rapid reversal of Taken, user-reported uncertainty about duplicate dosing, system-level notification disablement, withdrawal from caregiver sharing, privacy complaints, and support volume. Any serious safety incident should pause the pilot and trigger review rather than wait for the primary metric's reporting cycle.

Expand the segment or investigate pharmacy import, connected dispensers, and more complex care collaboration only if users complete prototype tasks, pilot use is sustainable, the primary metric improves, and guardrails stay within bounds. If reminders are frequently disabled, determine whether the channel, frequency, or underlying problem is wrong. Do not simply send more notifications.

High-Quality Sample Answer

“I would first narrow the primary user to an older adult who can manage medication independently, uses a smartphone, lives alone, and occasionally forgets a dose. In the practice scenario, the user takes 4 long-term medications across 3 scheduled windows each day. The first release would exclude severe cognitive impairment, real-time clinical monitoring, and direct dispensing by a caregiver because those cases need a stronger care or hardware workflow.

I would not begin by drawing a reminder screen. I would interview and observe users where they organize their medication and separate four problems: whether the current plan is trustworthy, whether the medication can be identified, whether a reminder can be noticed, and whether the user knows what to do after a missed window. If side effects or cost cause intentional stopping, reminders are not the answer; the workflow needs professional communication.

The MVP would store a medication list reviewed by the user or a trusted helper, including the name, strength, purpose, instructions, schedule, source, and last review time. At the scheduled time, it would use large text, strong contrast, sound, and vibration. The user can choose only Taken, Remind Me Later, or Skip and can undo an accidental action. If the user skips, is uncertain, or exceeds the response window, the product provides a pre-saved route to a pharmacist or qualified professional; it never recommends a replacement dose or dose change.

Caregiver collaboration would be off by default. The user may authorize a specific person to receive a notification only after an unresolved event exceeds a user-defined window and may separately choose whether to share the medication list. The caregiver sees ‘not yet confirmed; please contact the user.’ They do not receive a clinical conclusion and cannot edit the plan by default.

I would first test whether users can correctly create 4 medications, set 3 windows, understand the three states, undo a tap, and revoke consent in a prototype. Then I would run a limited pilot. The primary metric would be timely self-confirmation for valid scheduled events within the personalized window. Guardrails would include incorrect plans, duplicate confirmation, rapid reversal, disabled notifications, support requests, and privacy complaints. That metric can show whether the workflow helps users remember and record. It cannot prove ingestion or improved clinical outcomes.

If timely confirmation improves, users continue using the product, and safety guardrails stay within bounds, I would investigate pharmacy import and connected dispensers. If users frequently disable reminders, or the main causes are side effects, cost, and an untrustworthy plan, I would stop increasing notification pressure and return to the problem definition.”

The answer connects the product goal, clinical boundary, user control, and validation evidence. If the interviewer changes the population or risk, the design should change too.

Common Mistakes

• Treating all older adults as one user → Independent use, remote assistance, and professional care impose conflicting requirements → Choose one primary segment and state exclusions and later segments.
• Jumping straight to alarms, voice, and large text → The cause may be forgetting, unreadable labels, an incorrect plan, or intentional stopping → Map the task chain and gather evidence at each break.
• Equating “notification delivered” with “medication taken” → The system's capability is overstated and its metrics and safety logic become invalid → Separate delivery, user action, self-report, and clinical fact.
• Letting the app tell a user to replace a missed dose → Different medications and prescriptions require different handling, so the product crosses a clinical boundary → Show reviewed instructions and a professional-help route instead of generating dose advice.
• Sharing everything with adult children by default → Payment or family relationship is not continuing consent → Let the user choose caregivers, fields, notifications, and revocation.
• Building hospital integration, AI pill recognition, and a smart dispenser in version one → The team cannot identify which component solved the problem and inherits multiple accuracy and operating risks → Validate the critical hypothesis with the software state journey first.
• Optimizing daily active use, clicks, and notification opens → Users may be pushed into opening the product without completing the job more safely or confidently → Use an outcome-near proxy such as timely self-confirmation plus safety, experience, and privacy guardrails.
• Increasing frequency when reminders are disabled → Notification fatigue worsens usability and may indicate that forgetting is not the cause → Check segment, window, channel, and cause before iterating or stopping.
• Using one interview as validation → Stated preference does not prove that users can configure and sustain the workflow → Run task testing, then a limited real-world pilot.

Follow-Up Questions and Responses

Follow-up 1: If an adult child pays, why should they not see every medication record by default?

The payer, user, and data subject may be different people. Default full sharing reduces user control and may cause rejection or hidden behavior. The user should authorize each caregiver, visible field, and notification condition. If legal guardianship changes authority, it requires a separate identity and compliance workflow; “the child pays” is not sufficient.

Follow-up 2: Is the MVP sufficient when missing a particular medication can cause serious harm?

Exclude that medication or population from the ordinary pilot until clinical, pharmacy, and compliance owners define the time window, escalation, and accountability. The solution may require pharmacy import, a connected dispenser, telephone confirmation, or professional intervention. Stronger evidence and reliable escalation take priority over scale, and self-reporting software must not present itself as real-time monitoring.

Follow-up 3: What if the user taps Taken, then says a few minutes later that they are unsure?

Allow reversal while retaining the audit record, then immediately show reviewed help information or a route to a pharmacist or qualified professional. Do not infer from the elapsed time that another dose should be taken, and do not erase the previous record. Count the event in safety guardrails to determine whether the confirmation control is easy to trigger accidentally or lacks information.

Follow-up 4: Should reminders automatically change when the user travels across time zones?

Do not silently move every window. Detect the time-zone change, show the original and local-time interpretation, ask the user to choose based on professional instructions, and record when the choice takes effect. If the user cannot decide, route them to a pharmacist or qualified professional. High-risk medications should use a pre-defined travel plan.

Follow-up 5: Timely self-confirmation improves, but many users disable notifications. Is that success?

First check whether the denominator now contains only the most engaged users. Notification disablement is an experience guardrail and can create survivorship bias. Analyze it by segment, reminder count, and usage duration. If it crosses the pre-agreed boundary, pause expansion and retest the frequency, channel, or target segment even when confirmation rises among the remaining users.

Follow-up 6: Research shows many users intentionally take less medication because of side effects or cost. How should the product change?

Separate those users from the forgetting problem. The product may help record the reason, prepare questions, and connect the user with a pharmacist, clinician, or affordability resource. It must not use more reminders to coerce use or change the prescription. If intentional nonuse is the dominant cause, the roadmap should shift from reminders to communication and support, reducing the priority of the original MVP.