What it is

Clients document their own lived experience in real time on their phones — text, photos, video, voice — usually over days or weeks, guided by daily prompts on a diary-study platform.

What it tells you and what it cannot

How it runs

Strengths, watch-outs and quality signals

Equity & consent

The biggest digital-access equity risk of the eight methods — it structurally excludes people without devices, data, or stable housing. Mitigate with loaner devices, data stipends, voice-first entry options, and multilingual prompts (major platforms support 30+ languages). Disable location tagging and brief participants on filming others; treat submitted media as PII.

When to choose it

Study snapshot

For two weeks in July, twelve participants kept a diary on their phones about what it takes to get help from us: one to three entries a day, in text, photo, or voice, whichever was easiest in the moment. Amharic voice notes were translated by the platform. Participants received a data stipend and a loaner device where needed, and location tagging was switched off for everyone.

Why a diary? Because most of the work of getting help does not happen at our counters. It happens on a lunch break, at a bus stop, and at the kitchen table after the children are asleep — hours no staff member ever sees. A diary is the only method in our set that records those hours as they happen, in the participant's own words.

This report follows one record closely: P07, the study's only household diary — two adults documenting into a single shared record. When the study began they were 34 days into a second eviction notice, with $3,400 in rent arrears, and had already visited 6 offices, retold their situation 8 times, and completed 7 forms, with 2 of their 9 needs met. Together they logged 61 of the study's 402 entries: 34 voice notes from Hossina, mostly during the day, and 27 text-and-photo entries from Aboul, mostly after his evening shift.

Method & sample note

A phone diary reaches only people who opt in twice: once to the study, once through a working smartphone. This household is reachable because the platform takes voice-first Amharic and the stipend covers the data; a family with no smartphone, no minutes, or no safe moment to record would never appear here. Nothing below is a rate — it is what twelve self-selected records chose to show.

Themes

Five themes recurred across P07's 61 entries. The chart shows how often each appeared; the sections below tell each one through the household's own entries, with the team's reading kept to a closing sentence. An entry can support more than one theme.

Theme 1 — Benefits work happens in the seams of the workday

Fourteen entries — the largest cluster in the record — show case work landing in the only minutes of the day nobody else owns: Aboul's 30-minute lunch break, and the hours after 11 pm. Calls happen from the warehouse breakroom, letters get photographed at the bus stop to be reread on the ride, and forms get filled at the kitchen table long after the offices close.

"12:10 I call the eligibility line. Break ends 12:40. Still the music. Eating standing up so a hand is free to write if a person comes. 12:29 — no person. Hanging up, dock is calling us back."

— Aboul, diary entry, Day 8, 12:29 pm

The hold queue does not know it is spending someone's lunch break. That day it took 19 of his 30 minutes and gave nothing back — and the next call has to come out of the same window tomorrow.

Theme 2 — One case, two half-records

The household divides the work the only way their schedules allow: "Hossina can go at 2. Aboul has the English and is at work." So Hossina sits at the counters, and eleven entries show what follows: each program meets one adult and files that adult as the whole case.

"The worker asks where my husband is. At work, like every day. The file shows only my name, so today the household is only me. I record what she said so it is not lost on the bus."

— Hossina, voice note, Day 8, 5:50 pm, translated from Amharic

Every counter's record is internally correct — and jointly wrong. Nobody's file holds the whole household, so the couple reassembles the case themselves each night. As they put it: "We need someone to see the household, not half of it."

Theme 3 — Translation is a second, unpaid shift

Nine entries follow the same after-hours routine: an English-only letter arrives, and the household converts it into night work — photograph it, run it through a translation app, redo the math by hand, decide what it means. The app translates the words but not the reasoning.

"The letter says no, but it is using last year."

— Aboul, text entry beside a photo of the denial letter, Day 3, 11:52 pm

That one line carries the theme: the denial reasons from last year's income while the household lives in this one, and its logic never arrives in a language built to be understood.

Theme 4 — The arrears math lives on an envelope

Seven entries show the household's arithmetic running ahead of the system's. They know the number that matters; no office has said it back to them yet.

"Rent owed 3,400. Assistance letter covers 2,100. Gap 1,300. I write it on the envelope again. No desk has said 1,300 back to us yet."

— Aboul, diary entry, Day 5, 11:58 pm

Fast action, divided labor, exact math, persistence — the household's competence is primary data in this study, not a silver lining. The unclosed $1,300 gap is the system's, not theirs.

Theme 5 — The relay the system cannot see

Eight entries document a deliberate relay: Hossina holds the daytime, Aboul holds the English, and the baton passes at the kitchen table after 9 pm, when the day's voice notes get replayed against the letters.

"I hold the day, he holds the English. At night we put the two halves on the table and make one case out of them."

— Hossina, voice note, Day 11, 9:24 pm, translated from Amharic

No program sees this relay, and no record credits it — but it is the machinery keeping the case moving. Every recommendation later in this report is, one way or another, about taking work off this table.

Between-visit timeline

One representative weekday from the household record — Day 8, a Tuesday.

Then the record goes quiet. Days 9 and 10 pass with no entries — the only zero-entry days in P07's otherwise nightly documentation. On the morning of Day 11 there is one four-second voice note.

Analyst flag. The gap is unresolved and stays that way. The diary shows the household's nightly rhythm broke and then resumed; it does not show why. The team chose not to probe a silence; no interpretation is offered.

Highlight-reel shot list

Stills pulled for the readout deck; all synthetic illustrations.

• Day 3, 11:52 pm — the denial letter on the kitchen table, phone propped against the sugar jar, translation app open; text unreadable.
• Day 5 — an envelope margin filling with ballpoint subtraction, numbers out of focus.
• Day 8, 12:14 pm - breakroom counter; active speakerphone hold beside an unopened lunch container and note paper.
• Day 8, 2:04 pm - phone queue status showing queue number and estimated wait in a waiting room, queue area blurred behind it.
• Day 8, 9:20 pm — two phones side by side on the table, one mid-playback.
• Day 11, 6:12 am — dark frame, audio only, four seconds.

Findings & recommendations

Findings tie to diary evidence; interpretation stays in the theme notes above; owners are roles, not names.

Limitations & quality self-check

• Small n, self-selected. Twelve records, one featured household: depth, not measurement, never a rate.
• Performed for the camera. A diary is self-curated; the tidy envelope math and steady 6:05 am lists may be partly composed for the researcher's eyes.
• Translation layer. Hossina's Amharic reaches analysts through platform translation, with back-translation spot-checks on 10 notes; register and nuance can flatten in transit.
• The silent nights. Days 9–10 produced no entries and one terse morning note. What those nights held is an explicit unknown of this record: the method captures only what participants choose to show.
• Unknowns preserved, unresolved (verbatim from the persona dossier):
  • Which denials are local practice versus state or federal rule
  • How often language access is offered before a client asks
  • Whether household-level records can be used without erasing individual consent
• Self-check. One household, one record throughout; both adults' voices present; finding, interpretation, and recommendation kept on separate labeled lines; geotagging off; no entry either adult flagged as private was used.