A behavioural diagnostic, from holding a mobile wallet to saving in it every month

Context

Financial inclusion programmes have proven effective at opening accounts, but far less successful at driving active usage. Mobile wallets with savings features illustrate this challenge: an account holder enables the feature, makes one or two initial deposits, and then activity stalls. The standard operational response is to launch an informational campaign or financial training module—chosen well before anyone has determined which specific behaviour is missing and why.

A behavioural diagnostic resolves those two questions before resources are committed to an intervention. This article provides a comprehensive worked example based on synthetic data. It demonstrates every methodological choice required in a diagnostic and the explicit rules used to decide it: identifying the target behaviour, choosing the analytical framework, pinpointing attrition points along the journey, scoring and prioritising barriers, translating barriers into interventions, screening candidate solutions, and defining the expected outcomes. The interactive tool above allows you to adjust scores and weights to observe how the ranking and recommended intervention package adapt in real time.

1. One target behaviour, defined in advance

Dimension	Operational definition
Who	Adults (aged 18+) holding a mobile money wallet that was active within the past 90 days (at least one recorded transaction) and whose account tier (KYC level) permits access to the savings feature.
What	At least one customer-initiated electronic transfer from the primary wallet balance into the savings feature. Interest payments, technical reversals, promotional provider credits, and agent-assisted transfers are explicitly excluded.
When	Every calendar month.
Measurement	Extracted from the provider’s transaction logs using a pseudonymised account ID. Primary indicator: proportion of eligible wallet holders who make a qualifying deposit during the month. Secondary indicators: proportion depositing in at least 3 of 6 consecutive months; net savings volume (gross deposits minus withdrawals within 7 days).

Why focus on a single behaviour? Because behavioural barriers are highly context- and action-specific. The broad objective of "saving more" conflates activating a feature, making deposits, avoiding premature withdrawals, and maintaining a baseline balance. Each action involves distinct psychological and structural barriers; diagnosing them simultaneously produces a diffuse inventory of issues with no actionable path forward.

Four candidate behaviours were evaluated on a 1-to-5 scale across four equally weighted criteria:

Candidate behaviour	Direct impact	Measurability	Tractability	Policy relevance	Total
Activate the savings feature	2	5	4	3	14
Deposit into the savings feature every month	4	5	4	5	18
Maintain a balance equal to one week of expenses for 90 days	5	3	2	4	14
Attend a financial education workshop	1	4	4	3	12

• Direct impact: the extent to which the action immediately establishes a household financial buffer.
• Measurability: the degree to which the metric can be captured directly via administrative logs without relying on self-reported data.
• Tractability: the plausibility of moving the indicator using operational levers within a 12-month horizon.
• Policy relevance: alignment with the transition point where programme indicators typically stall (here, bridging the gap between account ownership and active use).

While feature activation is easy to measure, it does not constitute ongoing usage. Maintaining a minimum balance directly targets financial resilience, but it is heavily confounded by income shocks. Finally, workshop attendance reflects participation rather than an applied financial behaviour.

2. Framework: COM-B to diagnose, EAST to design

This diagnostic applies the COM-B model, which sits at the centre of the Behaviour Change Wheel (BCW) (Michie, van Stralen, & West, Implementation Science, 2011, doi:10.1186/1748-5908-6-42). COM-B posits that a given behaviour occurs only when an individual possesses the requisite Capability (psychological: knowledge, attention, memory, comprehension; physical: digital dexterity), Opportunity (physical: financial liquidity, time, environmental access, interface design; social: prevailing community norms), and Motivation (reflective: conscious goals, cost-benefit evaluations, beliefs; automatic: emotional responses, heuristics, habits). Each identified barrier is mapped to one of these six sub-components, which subsequently guides the selection of appropriate intervention functions (education, persuasion, incentivisation, training, environmental restructuring, modelling, or enablement).

To guide intervention design, we apply the Behavioural Insights Team’s EAST framework: ensure the desired action is Easy, Attractive, Social, and Timely (Behavioural Insights Team, 2014). COM-B clarifies what is missing; EAST guides the form and delivery of the solution.

The causal chain from knowledge to behaviour. Knowledge is merely an input to psychological capability. It translates into real-world action only through an explicit behavioural sequence:

1. The user becomes aware of the savings feature and understands its purpose (psychological capability).
2. This awareness updates beliefs regarding key consequences, such as fee structures and emergency liquidity (reflective motivation).
3. These updated beliefs crystallise into a concrete intention to make a deposit during the current month (reflective motivation).
4. That intention converts into action only if a salient prompt arrives when disposable funds are present (attention) and immediate spending impulses do not take precedence (automatic motivation).
5. The digital transaction sequence involves sufficiently low friction to permit completion (physical opportunity).
6. Post-transaction confirmation provides immediate feedback, reinforcing the feedback loop for subsequent cycles (habit formation).

Traditional educational programs focus exclusively on links 1 and 2. The quantitative funnel below highlights what fraction of user drop-off occurs during these cognitive phases versus the substantial attrition that occurs after an intention is already formed—a phase where educational messaging has little effect.

3. Journey map and funnel

The target behaviour is segmented into five observable milestones. In an empirical diagnostic, Awareness and Intention are measured via a brief phone survey linked to account records with prior consent; Initiation is tracked through USSD or smartphone app session logs; Completion and Repetition are captured directly from transactional ledgers. The figures below reflect the synthetic event log generated for this analysis.

Key funnel findings: Overall, 15.5% of eligible wallet holders execute the target behaviour in month 1, and only 7.4% repeat the action in month 2. The largest single point of absolute attrition occurs between Awareness and Intention (5,826 users lost). Crucially, however, more than half of those who form a clear intention to deposit never initiate the transaction flow (4,311 of 8,950 individuals, or 48.2% attrition at this step).

This gap illustrates the classic intention-action disconnect: an informational marketing campaign might improve upstream awareness, but it would leave the largest underlying drop-off completely unaddressed. Furthermore, conversion in rural areas reaches only 12.0% compared to 18.0% in urban centers, demonstrating why disaggregated sub-group funnels are essential to the diagnostic method.

4. Barrier analysis

Fourteen candidate barriers were identified along the user journey. Each is mapped to a specific step, a COM-B domain, and an empirical data source. The complete barrier register, including explicit validation protocols, is provided in the accompanying CSV dataset.

Rank	Identified barrier	Journey step	COM-B category	Primary evidence source	P	S	T	Priority score
1	B01 Unaware of the savings feature or its function	Aware	C-Ps	Admin data, survey	3	5	5	75
2	B07 Complex path: 7 USSD screens and session timeouts	Complete	O-Ph	Usability testing, admin data	3	4	5	60
3	B04 Intends to save, but incoming funds are consumed immediately	Initiate	M-Au	Admin data, survey	5	3	4	60
4	B05 Forgets: no timely prompt when funds arrive	Initiate	C-Ps	Admin data, user interviews	4	3	5	60
5	B03 Fears deposited funds cannot be accessed quickly in emergencies	Intend	M-Re	User interviews, survey	4	4	3	48
6	B02 Believes savings accounts are meant only for large sums	Intend	M-Re	Survey, user interviews	3	3	4	36
7	B06 Zero wallet balance; cashing in requires a trip to an agent	Initiate	O-Ph	Admin data, survey	4	4	2	32
8	B09 Difficulty reading interface text or verifying transaction amounts	Complete	C-Ps	Usability testing, survey	2	4	4	32
9	B12 Absence of confirmatory balance feedback or savings progress	Repeat	M-Au	Usability testing, admin data	3	2	5	30
10	B08 Uncertainty regarding transaction fees for internal transfers	Complete	M-Re	Usability testing, survey	2	4	3	24
11	B13 Absence of visible peer usage or positive social norms	Intend	O-So	Survey, user interviews	3	2	4	24
12	B11 Income too low or irregular to commit to monthly deposits	Repeat	O-Ph	Admin data, survey	4	4	1	16
13	B14 Account tier limits access to the savings feature	Initiate	O-Ph	Admin data	1	5	3	15
14	B10 Visible balances invite financial demands from family	Repeat	O-So	User interviews, survey	2	3	2	12

Scoring rule. Every barrier receives three ratings from 1 to 5: Prevalence ($P$), Severity ($S$), and Tractability ($T$). The priority score is calculated as:

$$\text{Score} = P^{w_P} \times S^{w_S} \times T^{w_T}$$

With default equal weights ($w_P = w_S = w_T = 1$), the calculation simplifies to the multiplicative product $P \times S \times T$, yielding values from 1 to 125. Using a multiplicative formula rather than an additive sum ensures that a barrier scoring a 1 on tractability cannot achieve high priority based solely on high prevalence and severity. Tied scores are broken first by severity, then by prevalence.

Scale	Prevalence ($P$): share of cohort affected	Severity ($S$): impact on target behaviour	Tractability ($T$): feasibility within 12 months
1	Less than 10%	Minor friction	Beyond programme scope or control
2	10% to 24%	Delays deposit timing	Requires regulatory reform or multi-year investment
3	25% to 39%	Causes abandonment in some months	Requires costly product redesign or partner approval
4	40% to 59%	Usually causes abandonment at this step	Addressable via modest process or messaging adjustments
5	60% or higher	Blocks the transaction entirely	Low-cost, rapid change testable via randomised trial

The top five barriers span four distinct COM-B categories: psychological capability (B01, B05; representing 45% of the cumulative score of the top five), physical opportunity (B07), automatic motivation (B04), and reflective motivation (B03). Notably, only two of these five barriers (B01 and B03) can be addressed through informational interventions.

Sensitivity analysis. If tractability is excluded ($w_T = 0$), the top-ranked barriers shift to B03, B06, B11, B01, and B04. In this scenario, structural constraints—such as traveling to an agent and income volatility—rise to the top. While these barriers reflect genuine operational challenges, they cannot be resolved via a behavioral package alone. They are highlighted as foundational constraints for core product and policy teams, and the tool flags them whenever they enter the top tier.

Sludge audit. The savings flow within the hypothetical USSD interface was evaluated screen by screen. The values illustrate metrics recorded during timed usability sessions:

Interaction step	Screens	Keystrokes	Median duration	Explicit fee display	Agent visit	Documentation	Optimization status
Deposit cash to fund wallet (if empty)	–	–	Variable (travel)	Cash-in fee only	1	None	Structural constraint
Dial operator short code	1	6	5 s	–	0	None	Incompressible
Financial Services → Savings → Deposit	3	3	20 s	–	0	None	Removable (direct shortcut)
Select savings pocket and enter amount	2	5	20 s	No	0	None	Reducible (pre-fill last amount)
Confirm and enter PIN	1	5	15 s	Absent	0	None	Reducible (display "Fee: 0")
Receive transaction confirmation SMS	–	–	–	–	0	None	Improvable (show total progress)
Entry-tier accounts: identity verification	–	–	–	–	1	National ID	Regulatory requirement, simplifiable

Navigating 7 screens and 19 keystrokes requires approximately 60 seconds for a digitally proficient user, set against a hard technical session limit of 90 seconds. This narrow margin directly explains why slower readers experience frequent timeouts (B07, B09). As Sunstein highlights in his case for routine sludge audits, transaction friction falls most heavily on the vulnerable users who stand to benefit most from financial inclusion.

5. From barrier to mechanism to intervention

Barrier	Underlying mechanism	Proposed intervention (BCW category; EAST)	Intermediate indicator	Evaluation methodology
B01 Unaware	Lack of awareness prevents intention formation	P4: Prominent "Save" entry on the main USSD menu with contextual taglines; structured agent scripts during cash-in transactions (Education, Environmental Restructuring; Attractive).	Increase in awareness and intention; higher rate of savings-menu pageviews.	A/B testing of menu copy, randomized at the account level, as detailed in the reminder trial.
B07 Buried path	High interaction friction and session timeouts	P1: Streamlined 3-screen pathway, pre-filled recurring deposit amounts, explicit "No Fee" confirmation (Environmental Restructuring, Enablement; Easy).	Higher initiation-to-completion conversion; lower step-level abandonment.	A/B testing of menu hierarchies using live session logs.
B04 Spends first	Present bias: immediate consumption outcompetes future planning	P3: Opt-in auto-transfer feature reserving a chosen percentage of every incoming deposit, cancelable at any time (Enablement; Easy, Timely).	Product adoption rate; monthly deposit frequency and net balance growth.	Measure consumer demand via a discrete choice experiment, followed by a randomized pilot.
B05 Forgets	Limited cognitive bandwidth when liquidity arrives	P2: Real-time SMS prompt triggered immediately upon account inflow, paired with a simple implementation intention (Environmental Restructuring; Timely).	Proportion of inflow months with a qualifying deposit; deposits made within 48h of inflow.	Multi-arm randomized controlled trial (four-arm reminder trial).
B03 Liquidity fears	Misconception regarding access rules; provider trust deficit	P4: Explicit withdrawal rules displayed on deposit screens; agents demonstrate rapid test withdrawals (Persuasion, Modelling; Attractive).	Improvement in verified user comprehension; higher intent without increased dispute volume.	Cluster randomized trial assigned across agent service territories (cluster trial).

The evidence base for the two lowest-cost levers is well established: periodic reminders have been shown to substantially increase deposit rates in bank field experiments (Karlan, McConnell, Mullainathan, & Zinman, Management Science, 2016), while commitment savings products have proven effective at growing average balances among adopters (Ashraf, Karlan, & Yin, Quarterly Journal of Economics, 2006). However, because empirical findings do not transfer automatically across distinct market contexts, these studies establish hypotheses to be validated locally rather than definitive outcomes.

Intervention screening (APEASE). Candidate solutions were assessed against the six criteria of the Behaviour Change Wheel guide (Michie, Atkins, & West, 2014), aligned with the Public Health England framework adapted by the UCL Centre for Behaviour Change (2019): Affordability, Practicability, Effectiveness, Acceptability, Side Effects & Safety, and Equity.

Intervention package	Affordability	Practicability	Effectiveness	Acceptability	Side effects & safety risks	Equity considerations	Operational decision
P1 Simplify digital journey	High	Moderate (release cycle)	Moderate to High	High	Minimal risk	High: disproportionately supports users with lower literacy	Proceed
P2 Triggered inflow prompts	High	High	Moderate (empirically supported)	Moderate (message fatigue)	May encourage saving funds needed for subsistence; requires neutral copy and single-click opt-out	Moderate: requires interactive voice response (IVR) for non-literate users	Proceed as trial
P4 Inform and reassure	High	High	Moderate for B01; low if deployed alone	High	Minimal risk	Moderate: depends on variable agent compliance	Proceed as bundle
P3 Automatic savings allocation	Moderate	Moderate	Uncertain until baseline demand is confirmed	Needs evaluation	Illiquidity risk: requires hard caps and instant revocation	Moderate: potentially ill-suited for highly volatile incomes	Assess demand first
P5 Social proof messaging	High	High	Low	Moderate	Real adoption is only 15.5%; claiming "most users save" is misleading and risks backfiring	–	Do not proceed

Selection protocol: No intervention package proceeds if it receives a low rating on safety or equity without a defined mitigation strategy. Furthermore, a package must demonstrate at least moderate projected effectiveness supported by existing literature or an empirical pilot plan.

6. Behavioural theory of change

Logical framework for the integrated intervention bundle (P1 + P2 + P4):

Logical level	Expected result	Measurement indicator (Source)	Core hypothesis to test
Activities	Provider deploys a 3-screen shortcut with fee clarity; real-time inflow triggers enabled; agent scripts distributed.	Deployment milestone dates; prompts dispatched within 1 hour of qualifying inflow (gateway logs).	Technical infrastructure can process and trigger events from real-time core banking data.
Capability	Wallet holders understand the savings feature and the immediate withdrawal policy.	Verified awareness scores; correct belief indices (linked survey).	Users actively notice the revised menu entries during routine cash-in and peer-transfer transactions.
Motivation	Deposit intentions form; fear of losing account liquidity subsides.	Stated deposit intention rate (linked survey).	Clear fee and liquidity disclosures are trusted; provider mistrust is not the binding constraint.
Opportunity & Cues	Prompts arrive while cash is present in the wallet; transaction journey takes 3 screens.	Median elapsed time from inflow to prompt delivery; step-level drop-off rates (session logs).	SMS prompts reach the intended account holder rather than a secondary user on a shared mobile device.
Behaviour	Monthly qualifying deposit into the savings feature.	Monthly deposit conversion rate; 3-of-6-months regularity; net deposit volume (transaction ledger).	Deposited funds remain in the account rather than being withdrawn immediately or diverted from higher-yielding instruments.
Long-term impact	Accumulation of a liquid emergency savings buffer.	Median balance after 12 months; self-reported ability to cover unexpected expenses (follow-up survey).	Exogenous income shocks do not deplete reserves faster than they accumulate (see the financial capability results framework).

By assigning an auditable metric to every structural link, a null result in the final evaluation can be traced directly to the specific failure point in the causal chain.

7. Implementation requirements for a field diagnostic

Phase	Methodology	Target sample	Primary deliverable
1. Behaviour selection (Weeks 1–2)	Collaborative alignment workshop with provider leadership; multi-criteria barrier scoring.	6–10 key stakeholders	Formal target behaviour definition and indicator specifications.
2. Administrative funnel (Weeks 2–5)	Retrospective analysis of 12 months of pseudonymized session and transaction logs.	Complete eligible user base	Disaggregated funnels by customer segment; step-level drop-off heatmaps.
3. Qualitative interviews (Weeks 3–6)	Semi-structured interviews stratified by gender × geography × recent saving history; agent interviews.	32 account holders (4 per cell) and 8 field agents	Comprehensive barrier register with verbatim qualitative evidence.
4. Usability testing & sludge audit (Weeks 4–6)	Timed think-aloud usability testing across basic handsets and smartphones.	15–20 users across diverse literacy levels	Formal sludge audit; step-level friction analysis.
5. Quantitative survey (Weeks 6–9)	12-minute phone survey linked to user accounts with prior informed consent.	~1,100 respondents	Statistically verified prevalence estimates for each candidate barrier.
6. Scoring & prioritization (Week 10)	Blinded, independent dual scoring using the rubric; sensitivity tests.	Core research team	Validated, prioritized barrier ranking.
7. Intervention design (Weeks 10–12)	APEASE screening; theory of change formulation; RCT design.	Interdisciplinary team	Intervention design specifications and registered pre-analysis plan (PAP).

The qualitative interview sample follows empirical evidence indicating that core conceptual themes reach saturation within the first twelve interviews in relatively homogeneous populations (Guest, Bunce, & Johnson, Field Methods, 2006). The survey sample size of 1,100 ensures that estimated barrier prevalence remains within a margin of error of $\pm3$ percentage points at a 95% confidence level ($n = 1.96^2 \times 0.25 / 0.03^2 = 1{,}068$, assuming maximum variance where $p = 0.5$). Sub-group analyses at a $\pm5$ percentage point margin require a minimum of 385 respondents per cell.

Mitigating self-report bias. The primary target behaviour is never assessed via self-reported answers; it is calculated exclusively from core administrative ledgers. Self-reported survey data regarding saving activity are matched directly against administrative records to quantify social desirability bias, and this divergence rate is published alongside final findings.

Furthermore, behavioral intention is measured before any discussion of actual saving history; sensitive topics (such as intra-household financial demands) are quantified using list experiments; qualitative interviews focus on the most recent financial transaction rather than abstract habits; and interface usability is evaluated via direct observation. A barrier is classified as empirically evidenced only if it is corroborated by administrative transaction logs or confirmed across two independent qualitative methodologies.

Methodological limitations

This worked example serves an illustrative purpose. The transition probabilities in the funnel and individual barrier scores are model assumptions designed to showcase the analytical framework; empirical analyses in live markets may yield substantially different priorities.

Additionally, the synthetic event log models awareness and intention as if they were fully observed across the entire customer base, whereas field implementations rely on sampled survey estimates subject to standard sampling error.

Although the priority scoring formula provides an objective, transparent framework, individual inputs retain an element of qualitative judgement—necessitating dual scoring protocols and regular sensitivity testing.

Finally, while the COM-B framework classifies behavioral barriers effectively, it cannot predict the exact effect size of an intervention package on its own. Only randomized trials can establish quantitative impact. In all cases, macro-level constraints—particularly severe income poverty and physical distances to cash-in agents—place absolute limits on mobile savings adoption that behavioral interventions alone cannot overcome.