Sayantani Mukherjee

Case study 01Amura HealthUser research and service design

The behavioral audit: closing the system and user alignment gap

Reorganising high friction clients into a six segment psychographic model that realigns how the service is delivered. The program is a chronic care service for 7,000+ clients.

Role
Service Designer and User Researcher, solo project.
Duration
44 days of research and modelling.
Scope
80 high friction clients audited in depth. 20+ contextual inquiry sessions with health coaches, doctors and escalation managers.
Impact
In comparable client cohorts across three care teams, escalations fell from 27.4% to 18.1% and refund requests from 16.2% to 4.7% after the model was introduced.
Methods
Artefact analysis, criterion sampling, contextual inquiry, affinity mapping, thematic analysis, psychographic segmentation, journey mapping and a behaviour failure model
00The problem

A belief about clients with no evidence underneath it

The business problem

Clients kept escalating and dropping out. Care teams had no way to see it coming: before the model, no early risk flags were being acted on.

What the founder believed

“Clients drop off because they are not serious. The protocol works, so the failure sits with the client.”

What the research showed

The same failure behaviour came from six different underlying mindsets. Each one broke down at a predictable stage of the program.

What I built

A way for care teams to read a client’s mindset early and act on the warning signs before they became an escalation.

What I led and conducted

  • The sampling that narrowed 7,000+ clients to 80
  • Artefact analysis of their chats, recorded calls and escalations
  • 20+ contextual inquiry sessions
  • Affinity mapping and thematic analysis
  • The six segment psychographic model
  • Journey maps and the behaviour failure model

Who I worked with

  • The founder, whose read of the drop offs was the claim I set out to test
  • Health coaches, doctors and escalation managers, who hold the why behind client behaviour

How the care team uses it

  • Coaches read every client against the six mindsets and revisit it through the program
  • Coaches and doctors tailor coaching and medical advice to the mindset
01My process

Reading the raw signal first

About 44 days, front loaded onto reading what clients had already said and done, because the answers were sitting in the record. Synthesis and modelling were fast once the signal was clean.

Frame and gather · 32 days
  1. Finding the logic gap between the founder’s belief and the evidence1 day
  2. Artefact analysis of 80 clients21 days
  3. Contextual inquiry, 20+ sessions10 days
Synthesise · 3 days
  1. Affinity mapping1 day
  2. Thematic analysis2 days
Model · 9 days
  1. Psychographic segmentation4 days
  2. Journey mapping2 days
  3. Behaviour failure model3 days
Figure 01The 44 days as one continuous timeline. Nearly three quarters went into gathering evidence before any model existed.
02Before we dive in

What counts as an escalation and why it is expensive

At Amura, a client complaint becomes an escalation the moment a manager gets tagged in a message or called into the group chat directly. Escalations burn expensive backstage resources. They emotionally exhaust the care team and their managers and quietly degrade the experience of every other client under that same care team.

A belief awaiting evidence

“Clients drop off because they are not serious. The protocol works, so the failure sits with the client.”

↓  but read on what evidence? Two gaps sat underneath it

The data gap

The record of what clients actually did, sitting in their chats, calls and escalations.

The behavioural gap

The reason underneath, held by the people who watch these clients daily.

↓  so each gap got its own method

Method 01

Artefact analysis

Reconstructs what happened, from 80 clients’ chats, recorded calls and escalation logs.

Method 02

Contextual inquiry

Surfaces the why, sitting with and questioning coaches, doctors and escalation managers.

Now the belief could be tested against real behaviour.

Figure 02The starting claim and what it was missing. Splitting the gap in two decided the research design: one method for what clients did, another for why.
03Artefact analysis21 days

Testing the claim against the record

The starting point

The founder read the drop offs as a client commitment problem. I treated that as a claim to verify and went to the record for the evidence.

What clients actually did was already recorded in their own chats, calls and escalations, so I started there. With 7,000+ clients I needed a defensible cut. The protocol and the systems had held steady for eight months, so behaviour was comparable across that window. I took the clients who had breached the escalation benchmark of three in a six month journey. That gave 120. I read 80 of them closely, the number I could do justice to in the time.

7,000+active clients

more than 3 escalations in a 6 month journey · protocol and systems unchanged for 8 months

120high friction clients

narrowed for depth within the time

80audited in depth

Chats and recorded calls surfaced

Confusions, friction points, resistance to protocol, fears, lifestyle issues, adherence gaps and reasons for deviation.

Escalations surfaced

The final tipping point, plus how much each client could tolerate before breaking.

Figure 03How 7,000+ clients became 80. Criterion sampling on escalations meant every audited client had already reached the breaking point the business cared about.
The remaining gap

The records explained most of the behaviour. The why behind the rest was not in them.

04Contextual inquiry10 days

Going to the people who hold the why

The records tell you what a client did. The why lives with the coaches, doctors and escalation managers who sit with these clients week after week. I ran more than 20 sessions, part observation of the real work and part structured conversation, to reach the why behind the behaviour.

Health coaches

The daily texture: what clients say between sessions and where they quietly bend the plan.

Doctors

The clinical read: which deviations are medical and which are behavioural.

Escalation managers

The breaking point: what a client sounds like just before asking for a refund.

Figure 04Who I observed and interviewed across 20+ sessions, with the part of the picture each one holds.
05Synthesis3 days

From a wall of notes to named patterns

Eighty clients across two methods is a wall of observations. Affinity mapping clustered that wall by what the notes had in common. Thematic analysis then named the patterns that kept recurring. Those patterns are what became the segments and the failure stages.

Step 01

Affinity mapping

Clustered a wall of observations into groups by what they shared.

Step 02

Thematic analysis

Named the patterns that recurred and turned them into the segments and the failure stages.

Output

Two models

Six psychographic segments plus a five stage failure model that places each friction in time.

Figure 05The synthesis sequence. The boards are not shown because they hold real client data, so the models that came out of them stand in for the working.

Client dataThe affinity and thematic boards contain identifiable client conversations and are withheld.

06Psychographic segmentation4 days

Designing for the belief, not the biography

Why not personas, or demographics? During the audit, a 20 year old student in the US and a 64 year old retiree in Kerala turned out to show identical adherence patterns.

Insight

In chronic care, age, gender and geography are poor predictors of behaviour. Mental models are the real signal.

Adherence is driven by internal mental models and a lifetime of health related stimuli, not anything trackable in a straight line. Psychographic segmentation made it possible to design for the belief system rather than the biography. It surfaced six distinct segments that predicted how a client would respond to the program long before they dropped off.

Mindset 01

The health as a side task client

24of 80 audited clients showed this mindset

The belief they arrive with

“This reversal should slot seamlessly into my existing schedule.”

What success actually requires

“My environment and schedule must be intentionally redesigned until this behavior becomes a habit.”

Operational failures
Time based restriction lapses. Program follow through. Meal planning consistency. Supplement adherence.
Behavioural friction
Health management is filed away as a background task. Business and family responsibilities stay the core priorities, leaving limited mental and physical bandwidth for execution. Program effort reads as a productivity tax.
Root belief
“If the program is truly effective, it should be manageable alongside my current lifestyle without requiring significant sacrifice.”
Mindset 02

The willpower only client

18of 80 audited clients showed this mindset

The belief they arrive with

“The intensity of my desire to change equals my capacity to execute.”

What success actually requires

“Sustainable change requires environment design, not just sheer willpower.”

Operational failures
Lapses in consistency, especially during social eating. High emotional fatigue. Regression to old habits when motivation dips.
Behavioural friction
The client significantly overestimates individual discipline and underestimates social and cultural friction. Relying purely on willpower creates a cycle of high initial effort followed by inevitable burnout.
Root belief
“Failure is personal weakness. If I truly want this outcome, I will force myself to make it work, regardless of the environment.”
Figure 06The two highest volume mindsets, each defined by the gap between the belief a client arrives with and what success in the program actually requires.
Health as a side taskProfiled above24
Willpower onlyProfiled above18
Personalisation is the key14
Expecting no discomfort12
This program will take care of me10
Fixing my body fixes the rest of life2

24 + 18 + 14 + 12 + 10 + 2 = 80 audited clients

Figure 07The full distribution: six segments accounting for all 80 audited clients. Showing every segment, not only the two profiled, is what makes the model usable as a complete screening lens.
07Journey mapping2 days

Tracing where each mindset starts to slip

The first real realisation here: it is not entirely anyone’s fault when a client arrives with the wrong mindset. Even so, it becomes the service’s job to recognise that mindset early and subtly correct course before it hardens into disengagement.

Stage
Discovery
GC call
Onboarding
Intake call
Clean eating
V1 phase
V2 phase
Customer actions
  • Explores Amura via ads and referrals
  • Consumes testimonials
  • Evaluates: can this work for me?
  • Discusses goals and history
  • Asks feasibility questions
  • Decides to proceed
  • Begins the program on WhatsApp with the care team
  • Reviews onboarding messages and initial guidelines
  • Shares detailed lifestyle and schedule
  • Understands next steps
  • Attempts strict diet adherence
  • Manages meal prep and timings
  • Follows daily instructions
  • Continues protocol
  • Adds fibre drink, salt drink and stricter structure
  • Continues program with modifications
  • Balances protocol with real life
Touchpoints
  • Website
  • Social media
  • Word of mouth
  • Guidance counsellor call
  • WhatsApp
  • WhatsApp with HC and doctor
  • Onboarding messages
  • Initial guidelines
  • Health coach intake call
  • Daily WhatsApp communication
  • HC nudges
  • Diet plans
  • HC follow ups
  • Daily reminders
  • Protocol updates
  • Periodic HC check ins
  • Progress tracking
Mood
Mood curve across the journeyHopefulMotivatedManagingLoadedFrustratedFatiguedStabilized
Experience
Hopeful and outcome focused. Effort feels manageable.
High motivation. Trust in the system and its authority.
Slightly overwhelmed. Trying to fit tasks in.
Engaged but cognitively loaded. Still optimistic.
Time pressure. Cognitive overload. Frustration emerging.
Fatigue. Reduced enthusiasm. Selective compliance.
Stabilized but diluted adherence. More control taken by the client.
Friction points
  • Effort is underestimated
  • No anticipation of time and social trade offs
  • Ease expectation gets anchored early
  • Overcommitment without lifestyle planning
  • Effort casually explained
  • Leaves with confidence, not preparedness
  • Instruction overload
  • Timing conflicts begin
  • Early minor non compliance
  • No explicit life redesign discussion
  • Schedule conflicts remain unresolved
  • Execution feasibility not stress tested
  • Meal prep clashes with the work schedule
  • Social eating disruptions
  • Skipping or delaying meals
  • Beginning of partial adherence
  • Forgetting additional steps
  • Dropping non visible tasks first
  • Increasing negotiation with the protocol
  • Self modification of rules
  • Reduced strictness
  • Plateau or inconsistent results

Scroll sideways to follow all seven stages.

Figure 08Journey map for the health as a side task mindset. The friction row shows the decline starting at the very first touchpoint, where effort is underestimated, long before any visible non compliance.
08Behaviour failure model3 days

Placing each friction in time

Mapping each segment against the actual protocol made the failure visible. It stopped being a vague non compliance label and became a specific stage in the journey where a designed intervention was missing, which a policy reminder could never have fixed.

Stage
Stage 01Leading indicators
Stage 02Silent deviations
Stage 03Trust erosion
Stage 04Compounding friction
Stage 05Escalation
What it looks like
  • Instructions feel more complex than expected
  • Time and effort feel higher than assumed
  • Early confusion around the why behind steps
  • Client starts modifying the program quietly
  • Questioning effectiveness
  • Doubting instructions
  • Increased need for reassurance
  • Emotional fatigue
  • Reduced responsiveness
  • Increased inconsistency
  • Complaint or escalation raised
  • Request for simplification
  • Disengagement or refund
Client frictions
  • Struggles to align meal timings with work
  • Confusion between fibre drink and salt drink timing
  • Surprise at the number of supplements
  • Partial adherence
  • Skipping inconvenient steps
  • Adjusting the protocol to fit lifestyle
  • No visible change in clean eating raises: is this working?
  • Hunger and headaches read as negative signals
  • Comparing effort against visible results
  • Ignoring reminders and messages
  • Following only the easy parts of the plan
  • Rising frustration with restrictions
  • Asking for diet relaxation
  • Dissatisfaction with effort versus outcome
  • Dropping off or threatening to quit
Client quotes

“This is a bit more complicated than I thought.”

“I didn’t expect it to be this strict.”

“Why are there so many things to follow daily?”

“I missed it today, but I’ll do it properly tomorrow.”

“I’m mostly following, just adjusting a little.”

“I couldn’t manage it with my schedule today.”

“I’m doing everything, but I don’t see any change.”

“Is this supposed to feel like this?”

“I’m not sure if this is working for me.”

“This is getting hard to manage every day.”

“I have too much going on to keep up with this.”

“I’m trying, but it’s not sustainable like this.”

“This is not working for my lifestyle.”

“I need something more practical.”

“I can’t continue like this.”

Scroll sideways to follow all five stages.

Figure 09The behaviour failure model. Escalation, on the right, is a lagging indicator. The leading indicators on the left show up earlier in ordinary conversation, which is where the care team can still act.
Insight

Escalation is the last of five stages. The model moves the moment of action to the first.

09How this was used

From a model to a working parameter

Once approved, the model became a living parameter on every client. The coach reads each client against the six mindsets and revisits it as they move through the program. When a client matches a high risk mindset, the matching interventions switch on.

Step 01

Read against the six mindsets

Every client, revisited as a living parameter through the program.

Step 02

Interventions switch on

When a client matches a high risk mindset.

Step 03

Care is tailored

Coaches and doctors adapt coaching and medical advice to the specific mindset.

Figure 10How the model runs inside the service today. The interventions that switch on at step 02 are the subject of case study 02.
10Impact

Catching risk before it becomes an escalation

  1. Business problemEscalations pulled in managers, exhausted care teams and ended in refunds. The team only saw them once they happened.
  2. Research findingSix mindsets fail in predictable stages, with leading indicators visible well before an escalation.
  3. InterventionCoaches read every client against the model and act on early risk flags.
  4. Observed outcomeIn the cohorts after the model, escalations and refund requests were both lower.
27.4%→18.1%of the cohort escalated within three months
16.2%→4.7%of the cohort requested a refund within three months
0→67early risk flags caught and acted on before an escalation
How this was measured

The three teams whose clients made up the 80 audited. For each team, the clients in the first three stages of the program at one point in time were followed for three months, counting escalations and refund requests. The before cohort started in January 2025, the after cohort in October 2025, after the model was in use. They are comparable cohorts, not the same clients.

Before the model: cohort from January 2025After the model: cohort from October 2025

Share of each cohort within three months, by team

Escalated

Asked for a refund

The numbers behind the chart
Clients in the first three stages of the program, followed for the next three months
TeamBefore the model: cohort from January 2025, followed to April 2025After the model: cohort from October 2025, followed to January 2026
ClientsEscalationsRefund requestsEarly flags acted onClientsEscalationsRefund requestsEarly flags acted on
Team 326718 (26.9%)12 (17.9%)0367 (19.4%)0 (0.0%)23
Team 165816 (27.6%)9 (15.5%)0438 (18.6%)3 (7.0%)19
Team 217220 (27.8%)11 (15.3%)0488 (16.7%)3 (6.2%)25
All three teams19754 (27.4%)32 (16.2%)012723 (18.1%)6 (4.7%)67
Figure 11Every team moved the same way on both measures. The mechanism sits in the table behind the chart: before the model no team was catching risk early, after it the three teams acted on 67 flags before they became escalations.

The model gave the team something it did not have: a way to see risk before it arrived. That is the change I would claim. Because the cohorts come from two different windows, other changes in the service over that period may also have contributed to the size of the drop.

11Limitations

What this study can and cannot claim

  • Built from high friction clients. The 80 were chosen for their escalations, so the segments describe clients at risk, not the whole client base.
  • 80 of 120. Time limited the audit to 80 of the 120 clients who met the criterion.
  • Comparable cohorts, not the same clients. The before and after groups started at different times of year and differ in size.
  • Three teams. The outcome was tracked in the three teams whose clients were audited, not across the whole organisation.
12What I learned

Professional and personal

Scaling through systems

Research as a business metric: user friction is effectively an operational tax. Translating user pain into wasted clinical bandwidth made the research land with stakeholders in their own language. Demographics, in this case, turned out to be noise. Mental models were the real signal.

Backstage empathy

Empathy has to extend beyond the client to the clinical staff delivering the service. If a service design burns out the doctors and coaches behind it, the client experience fails no matter how good the client facing design looks.