# Barrier scoring rubric

Worked example by Ahmed Beddi. Use it with the barrier register (`barrier-register-en.csv`) and the
notebook (`behavioural_diagnostic.ipynb`). Synthetic example; no real programme is described.

Target behaviour: adults who already hold an active mobile-money wallet make at least one deposit
into its savings feature each calendar month.

## 1. Scoring rule

Each candidate barrier receives three scores from 1 to 5. Its priority is

    score = P^wP × S^wS × T^wT

with default weights wP = wS = wT = 1, so the default score is the plain product P × S × T (range 1 to 125).
A product, not a sum, so a barrier that scores 1 on any dimension cannot reach the top on the strength of the
other two. Setting a weight to 0 removes that dimension; setting it to 2 counts it twice.

Ties are broken by severity, then prevalence (both higher first), then barrier id.

The top K barriers (default K = 5) are carried forward to intervention design.

## 2. Scales

### Prevalence (P): share of the target population affected

| Score | Share of eligible wallet holders affected |
|---|---|
| 1 | under 10 % |
| 2 | 10 % to 24 % |
| 3 | 25 % to 39 % |
| 4 | 40 % to 59 % |
| 5 | 60 % or more |

Source of the estimate, in order of preference: administrative records (for example, the share who
never opened the savings menu); the linked short survey (weighted); user testing or interviews only
when neither exists, in which case P is capped at 3 and flagged "provisional".

### Severity (S): how far the barrier stops the behaviour for the people it affects

| Score | Meaning |
|---|---|
| 1 | A minor irritation; people who meet it usually still deposit |
| 2 | Delays the deposit or makes it less frequent |
| 3 | Stops the deposit in some months |
| 4 | Usually stops the deposit at this step |
| 5 | Makes the deposit impossible until the barrier is removed |

Evidence: conversion at the journey step among affected vs. unaffected people (logs linked to
survey), task failure in user testing, and accounts in interviews.

### Tractability (T): can a feasible intervention reduce it within 12 months?

| Score | Meaning |
|---|---|
| 1 | Outside the programme's or provider's control (for example, income levels) |
| 2 | Needs regulation, network investment or multi-year change |
| 3 | Needs a product change with a real cost or a partner decision |
| 4 | A change in communication or process at modest cost |
| 5 | Cheap, quick, and testable in a randomised trial within the provider's systems |

Tractability is scored by the design team with the provider's product and operations staff, not by
the analyst alone.

## 3. Evidence rule

A barrier is "evidenced" only if (a) an administrative indicator supports it, or (b) at least two
different source types agree (survey, user testing, interviews). Barriers that do not meet the rule
stay in the register with their score in brackets and are not carried forward.

Self-report is never used to measure the target behaviour itself. Survey answers about saving are
compared with the respondent's own transaction record (with consent) to estimate over-reporting; that
rate is reported with the survey results.

## 4. Scoring process

1. Two people score each barrier independently from the evidence pack.
2. Differences of 2 points or more on any dimension are discussed and the reason recorded.
3. The final scores, the weights and K are fixed and dated before intervention design starts.
4. A sensitivity check is reported: the top K under wT = 0 (ignore tractability) and under wS = 2.
   A barrier that is high on prevalence and severity but drops out only because of tractability is
   reported as a structural finding, not discarded.

## 5. Sludge audit checklist (one row per journey step)

For each step from "open the wallet menu" to "deposit confirmed", record:

- screens or menu levels, and keystrokes
- time to complete (median of timed user tests) and the session time-out
- fees charged or displayed, and whether the fee is shown before confirmation
- agent visits needed (cash in, identity upgrade)
- documents needed (identity document, proof of address)
- error messages and where users abandon (session logs)

Each item is classed as necessary (a legal or security requirement), reducible, or removable.

## 6. Intervention screening (APEASE)

Every candidate intervention is rated on the six APEASE criteria from the Behaviour Change Wheel guide
(Michie, Atkins and West, 2014): Affordability, Practicability, Effectiveness, Acceptability,
Side-effects and safety, Equity. Ratings are high / medium / low with one line of reasoning each. An
intervention rated low on Equity or on Side-effects is not taken forward without a mitigation.