Field guide

Lens User Manual

How to collect data with the Lens enumerator app, and how to manage forms, review data, and monitor fieldwork from the Lens admin console.

Worked example throughout: the JigawaUNITE Midline Evaluation form — its Oral Reading Fluency passages and its ICAN numeracy assessment.

A note on this document

Lens is still under active development. Screens, wording, and even some steps in this manual may change as the app improves — this manual will be updated to keep pace. If something on your screen looks slightly different from a screenshot here, don't worry: the overall steps and logic will almost always be the same.

This manual is written for two audiences: field enumerators who collect data using the Lens app on a phone or tablet, and administrators who manage forms, review data, and monitor fieldwork through the Lens admin console on a computer. It uses the real JigawaUNITE Midline Evaluation form as its working example throughout, but everything described applies to any form built on the Lens platform.

Every screenshot in this manual was taken from the real, live Lens system using a test account, so what you see here is what you will actually see in the field.

Watch: a complete video walkthrough

Prefer to watch rather than read? This eleven-minute, narrated video covers everything below in one sitting: signing into the admin console, building a brand-new form question by question in the Form Builder — every question type, plus hints, validation rules, choice randomization, skip logic, and a real random-assignment pattern — saving it live, then switching to the enumerator app to answer every question and submit a real response.

Recorded against a live Lens test form — every screen shown is the real, working product.
Part 1

The Enumerator App

Used out in the field, on a phone or tablet, to record a pupil's answers. It works fully offline — everything is saved on the device first, and sent to the server automatically once you're back online.

1. Signing in

Open the enumerator app at surveyplatform.newglobedata.education on your phone or tablet's browser. You'll see the sign-in screen below. Enter the username and password you were given, then tap Sign In.

The Lens sign-in screen
The sign-in screen.
  • One shared login per device. Most field tablets use one shared username and password that every enumerator using that device signs in with — your admin will tell you what it is.
  • You only need to sign in once. The app remembers you until you deliberately tap Sign Out.
  • No internet? No problem. Once signed in the first time, you can keep collecting data fully offline — you only need a connection again when you're ready to send finished responses.

Installing the Android app (APK)

Lens is also available as a downloadable Android app, for devices that prefer a proper installed app over opening a browser address:

Download the Lens app (.apk, ~2 MB)

  • Google Drive will first show a page saying it “can't scan this file for viruses” — this is Drive's standard warning for app files, not a sign of a problem. Tap Download anyway to continue.
  • Since this app isn't distributed through the Google Play Store, Android will ask you to allow “install from unknown sources” the first time — this is expected and safe for an internal field-data tool distributed directly by your project.
  • Open the downloaded file from your phone's Downloads/notifications to install it, exactly like installing any other Android app.
  • The installed app opens the same live Lens site described throughout this manual — every screen and step below works identically whether you're in the installed app or a plain browser tab.

Prefer not to install anything? You don't have to — opening surveyplatform.newglobedata.education directly in your phone's browser works exactly the same way, including offline. iPhone/iPad users (no APK support on iOS) should use this browser route, and can optionally add a home-screen shortcut for a more app-like feel: open the site in Safari, tap the Share icon, then tap “Add to Home Screen”.

2. The Home screen

Your starting point for everything: start a brand-new response, resume a paused one, see what's waiting to send, and see what's already been sent.

The Home screen
The Home screen: Fill Blank Form, Drafts, Ready to Send, and Sent.
  • Fill Blank Form — starts a brand-new response from question one. What you'll use most often.
  • Drafts — responses you started and paused (for example, after tapping Save & Exit). The number shown is how many are waiting.
  • Ready to Send — finished responses that haven't reached the server yet because you were offline. These send automatically once connected.
  • Sent — a personal log of responses this device has already delivered. Your data is already safe even if you never check this list.

A small status label at the top of every screen shows whether your work has reached the server: All sent (green), Sending… (blue), or Offline · N to send (amber) when there's no signal and finished responses are waiting.

It's completely normal for Ready to Send to hold items for a while with no signal. Nothing is lost — it sends the moment the device reconnects.

3. Starting a new response

Tap Fill Blank Form to see the list of forms you have access to. In this example, only one form is assigned to this login.

Choose a form screen
Choose a form — tap the one you want to start collecting data for.

Tap the form's name to begin. A short welcome screen describes the survey before the real questions start.

Welcome screen
The form's welcome screen.

4. Basic questions: names, places, dates

Most questions are simple: tap one option, or pick a name from a dropdown. A thin progress bar always shows how far through the form you are, and “STEP X of Y” in the corner tells you exactly which question you're on.

After picking your name as the Assessor and the LGA and School you're working in, you're asked which pupil you're assessing — already pre-loaded from the school's real roster.

Selecting the learner
Selecting the learner being assessed, from a pre-loaded list.

Right after this, some forms ask for a date. A plain date question opens Lens's own in-app calendar when you tap it — not your phone's own date picker — so it looks and works identically on every device.

The in-app calendar picker, open
Lens's own calendar — tap the field to open it, pick a day, or use Today/Clear.

In JigawaUNITE specifically, this step is a locked date: it fills itself in with today's real date automatically and can't be changed, since this particular question always means “the day of this interview,” never a date you need to recall or pick.

The locked, auto-filled date field
A locked date field — already filled in, greyed out, nothing to tap.

Next, a quick confirmation of the learner's school, name, and grade — a check that you have the right pupil before continuing.

Learner details confirmation
The learner's information, shown back to you for confirmation.

Then a script for you to read aloud, introducing yourself. Anything in square brackets, like [YOUR NAME], is filled in automatically — you read the name, not the brackets.

Enumerator greeting script
The greeting script the enumerator reads to the pupil.
The pale box under a script, like “If the pupil is very quiet…”, is guidance for you — never read aloud to the child.

The pupil must agree to take part before anything else happens. If the answer is No, stop the assessment there.

Consent question
Consent — the assessment must stop here if the answer is No.

5. The Oral Reading Passage Assessment, step by step

The heart of the reading assessment. The pupil reads a short passage aloud for up to 60 seconds, while you follow along and tap any word they get wrong. The app times it, records the audio, and works out the reading speed automatically.

In JigawaUNITE, every pupil reads two passages: a common baseline passage everyone reads (“Our Pond”), then one matched to their own grade — “Our House” for Grade 1, “The Ant and the Grasshopper” for Grade 2, and so on through Grade 6. The app already knows which one applies, based on the grade entered earlier.

Step 1 — Read the instructions

Before every passage, the app shows a reminder of exactly how to run the timed reading.

ORF instructions
The Oral Reading Fluency instructions screen.

Step 2 — Get the right physical passage ready

The app tells you exactly which passage to hand the pupil from the printed stimulus booklet, and the exact words to say before starting the timer.

Passage instructions
The passage instructions — note which physical page to flip to.
The passage title shown (e.g. “Our Pond”) must match the page you hand the pupil from the physical packet. If unsure, check the title against the packet's own contents page.

Step 3 — Start the timer

Nothing happens until you tap the Start box — that's the moment the 60-second countdown, and an audio recording of the pupil reading, both begin. The single button under the timer stays in the same place throughout the task: Back before you start, Finish while the pupil reads, then the steps to confirm the last word.

Ready to start reading
Ready to begin — the Start box, with Back underneath until the reading starts.
Tap Start at the same moment you say “Begin”. If the pupil starts reading before you tap Start, the timer misses those seconds: the score comes out too high, and the recording misses the first words. Lens checks this from the audio afterwards and flags it as “Started late?” for your supervisor.

Audio recording starts automatically when you tap Start — nothing extra to turn on. Make sure the microphone isn't covered and the phone is close to the pupil. On surveys set to show it, a red Recording audio label appears at the top while it records.

If the microphone can't start (for example, permission was refused) or the recording comes out empty, a red Not recording — check the microphone warning appears at the top of the screen, whatever the survey's settings. Stop and fix the microphone before going on; the interview is flagged so supervisors know the recording is missing.

The Not recording warning at the top of the reading screen
The warning that appears when the microphone isn't recording.

Step 4 — Follow along and mark mistakes

Tap any word the pupil reads incorrectly, skips, or badly hesitates over. A tapped word turns red with a line through it; tap again to un-mark it.

Reading in progress
Mid-reading: the timer is counting down and three words are marked incorrect.
  • Stay quiet while they read, except to help when they get badly stuck.
  • If the pupil hesitates for about 3 seconds, supply the word, point to the next one, and say “Please go on” — mark that word incorrect. Don't end the reading because the pupil has stopped.
  • If the first 10 words are all marked incorrect, the app ends the reading automatically (early stop).
  • If the pupil reads the last word before time is up, tap Finish straight away. Letting the timer run on caps a fluent reader's score; Lens flags it as “Timer ran on”.
  • If the full 60 seconds runs out first, the timer stops on its own and asks you to tap the last word.

Step 5 — Finish and tap the last word read

When you tap Finish with time still left and the pupil hasn't reached the end of the passage, Lens first checks you mean it. The timer and the recording pause while the box is up, so the pupil loses no time.

The Time is still running reminder with End now and Continue
“Time is still running. If the pupil stopped, point to the next word.”
  • Continue goes straight back to the reading; the timer and recording carry on from where they paused.
  • End now ends the reading — for example, when the pupil has finished the passage, or can't go on.

Whether you ended the reading or the time ran out, Lens then asks you to tap the last word the pupil read. This only marks where the reading stopped — it never marks that word wrong. When time runs out, tap the word the pupil had reached when the timer hit 0, not where they are now. Then tap Confirm & Finish, which stays locked until a word is tapped.

Tap the last word read step
The reading has ended: tap the last word read, then Confirm & Finish.

The app then calculates the score and shows it on the same screen.

Finished score
The reading's results, shown under the passage once it is confirmed.

Tapping Next shows a clean summary: words attempted, words correct, and the final score in correct words per minute.

Oral reading summary
The Oral reading summary — a clear breakdown of the pupil's results.
This exact sequence repeats a second time for the pupil's own grade-level passage. Nothing new to learn — it works exactly the same way.

6. The ICAN Numeracy Assessment, step by step

After the reading passages, the form moves into ICAN — a short, spoken numeracy check covering number recognition and basic arithmetic. Like the reading passages, it's read aloud and answered aloud; you record whether each answer was correct.

ICAN intro
The transition screen into the ICAN assessment.

First, the pupil's preferred language — English or Hausa. Every ICAN question exists in both; the app shows the right version to read.

Number recognition

The pupil is shown a row of numbers (from the physical packet) and reads each aloud. You tick off exactly which ones they read correctly — not necessarily in order.

Number recognition blank
Before any numbers are ticked.
Number recognition with selections
Two numbers ticked as read correctly.
Important: this is tick-many-that-apply, not a single choice. Leaving a number unticked means it was not read correctly.

Addition, subtraction, multiplication, division

For each task, the app shows the exact problem to read aloud, with the correct answer already printed in grey for your reference — the pupil never sees this part. You mark whether their spoken answer was Correct or Incorrect.

Addition task
An Addition task — the correct answer shown to you in grey for quick reference.
  • Read the problem exactly as written, aloud, in whichever language the pupil chose.
  • Give them a moment, then compare their spoken answer to the reference shown.
  • A few tasks include a short word problem instead of a bare sum — read the whole scenario aloud.
ICAN finishes with a School Location step, capturing the device's GPS automatically — tap Capture GPS and wait a moment.

7. Reviewing and submitting a response

Once every question is answered, the Review screen lists every question and its answer, start to finish. Scroll through and check everything looks right.

Review screen
The Review screen — every answer, laid out for a final check.

When satisfied, scroll down and tap Submit. If any audio was recorded (the reading passages), it uploads at this point too — you'll see “Submitting… (X of Y recordings sent)”. Stay connected until this finishes.

Submitted successfully
Submitted — the response has been recorded successfully.
No connection right now? Tap Submit anyway. The response saves safely and moves to Ready to Send, delivering itself automatically once you're back online.

8. Drafts, Ready to Send, and Sent

  • Need to pause partway through? Tap Save & Exit (visible on every question screen). Your progress — including finished reading passages — saves as a Draft, resumable later from Home exactly where you left off.
  • Never start fresh over an unfinished response. Always resume a paused interview from Drafts, rather than tapping Fill Blank Form again for the same pupil.
  • Ready to Send and Sent are safe to ignore day-to-day. They're there so you can double-check your own work, not because anything requires action.

If you're assigned more than one form, these three lists show everything together, across every form — the same way a real field-collection device normally works, so you can send everything at once before you lose signal. A row of form-name chips above the list lets you narrow it down to just one form when you want to; tap All forms to see everything again.

Drafts list with form-filter chips, assigned to two forms
Drafts across two assigned forms — tap a chip to narrow the list to just one.
Only assigned one form? The chip row doesn't show at all — nothing to choose between.

9. Signing out and staying safe

Sign Out is on the Home screen and the list screens (Drafts, Ready to Send, Sent, and choosing a form). It is no longer shown on question screens, so it can't be tapped by accident in the middle of an interview — use Save & Exit to leave an interview first. Signing out clears the device's saved login — the next person needs to sign in again. On a personal device kept for the whole project, you generally don't need to sign out between interviews; on a shared device handed between enumerators, sign out at the end of your shift.

Signing out does not delete any Drafts, Ready to Send, or Sent items already on the device — that data is tied to the device, not to who is currently signed in.

Part 2

The Admin Console

The web-based control room for the whole project, used from a computer. This is where forms are managed, submitted data is reviewed and exported, and reading assessments get an experimental AI cross-check.

The admin console is for supervisors, M&E staff, and administrators — not for enumerators in the field, who only need the app covered in Part 1.

10. Signing in to the admin console

Open the admin console at surveyplatform-admin.newglobedata.education in any browser and sign in with your own admin username and password. Unlike the enumerator app, every admin has an individual account — never a shared login.

Admin sign-in screen
The admin console sign-in screen.
No admin account yet? Ask an existing administrator to create one for you from Team (section 22).

11. The Surveys list

The first thing you see after signing in — every survey this account can see, in one list, reached any time from Surveys in the left sidebar.

The Surveys list, with KPI tiles, status chips, and the survey table
The Surveys list — four KPI tiles across the top, then every survey with its own status, progress, and last activity.
  • The four tiles at the top total up interviews, live surveys, enumerators active this week, and submissions waiting for review — across every survey combined, as of the moment shown next to the search box.
  • The Active / Live / Draft / Archived chips filter the table by status; Project narrows it to one project, and Sort reorders it (by last activity, name, and similar). List and Cards switch between the table layout shown here and a card-per-survey layout.
  • Each row shows the survey's project, status, progress (submissions collected, and against a target where one is set — see Settings, section 20), and when it was last touched. Click Open to go straight into that survey.
  • The search box searches by survey name across the whole list, regardless of which status chip is active.

+ New survey, top of the sidebar, starts a brand-new one: a handful of ready-made templates (e.g. EGRA, EGMA, a simple survey), pulled from this account's own real surveys, alongside uploading your own XLSForm or starting from a blank survey.

12. Opening a survey

Clicking a survey's name or its Open button takes you into that survey specifically. A breadcrumb (Surveys / <name>) always shows where you are, with a one-click way back to the full list. Right below the survey's title sits its status chip (Live, Draft, Archived), then its project, kind, question count, version, and language.

A survey open, showing its header and the shared tab row
Inside a survey — the header at top, then the tab row shared by every survey.
ButtonWhat it does
PreviewOpens a real, working copy of the survey to test — nothing entered here is ever saved or counted anywhere (not Overview, Data, or Reading AI). Available for every survey, whether or not it has a pending draft.
Fill on this computerOpens the survey's real, submission-writing experience directly, styled for a desktop screen — a quick way to see or try it, or to backfill data, without a phone.
Share with enumeratorsJumps straight to the Share tab (section 19) — the QR code, phone/computer links, and APK download real enumerators use to start collecting on this survey.

Below the header, seven tabs hold everything else about this one survey — the rest of Part 2 covers each in turn: Overview (13), Build (14), Data (15), Reading AI (16, only shown on surveys with a scored reading passage), Reports and QA (17), Share (19), and Settings (20). Every tab stays scoped to this one survey until you go back to Surveys.

13. Overview

The tab a survey opens on — a real-time summary of its own fieldwork progress.

The Overview tab, KPI tiles, charts, and breakdowns
Overview, on the JigawaUNITE Midline Evaluation survey.
  • Six tiles across the top: Interviews (collected, and against a target where one is set — see Settings, section 20), Last activity, Enumerators who have submitted, Schools visited, Waiting for review (links straight to the Data tab, section 15, filtered to that status), and Quality flags (links to Reports and QA, section 17).
  • Interviews per day — a chart of submissions over the last 14 days.
  • Interview locations — a map plotting both the GPS a question declared and the device's own silent background GPS reading, for a fraud/quality cross-check between what was entered and where the device actually was.
  • By school and By enumerator — ranked breakdowns of submissions, each scrollable to show everyone, not just the busiest few. By school shows progress against a target where one is set.
“13 of 3,060” on the Interviews tile means 13 real submissions collected so far out of 3,060 expected across every school — set from the Targets card on the Share tab (section 19).

14. Build

Where a survey's questions themselves live — a visual, no-code way to build and change a Lens survey directly in the browser, with no separate spreadsheet or file to manage. An Outline column on the left lists every section and question in order; the Inspector panel on the right edits whichever question is selected.

The Build tab: toolbar, Outline, and Inspector
Build, open on JigawaUNITE — Outline on the left, Inspector (Question / Phone preview / Survey) on the right.
Toolbar buttonWhat it does
Download spreadsheetDownloads the survey's underlying spreadsheet — the original uploaded file, or an always-up-to-date one generated from its current questions.
Questionnaire PDFA formatted, printable version of every question — handy as a paper reference or for review by someone without admin access.
TranslationsManages the text shown in each of the survey's supported languages, question by question.
Update from spreadsheetRe-uploads a revised spreadsheet to replace this survey's questions — the spreadsheet route to changing a survey, as an alternative to editing questions directly here. Every update is snapshotted first, recoverable from History (Settings, section 20).

The Inspector's three sub-tabs on a selected question are Question (its text, variable name, hint, type, and logic — the fields shown in the screenshot above), Phone preview (exactly how it will look to an enumerator), and Survey (settings for the survey as a whole).

  • Question types cover everything a Lens survey can ask — single/multiple choice, text, numbers, dates, sections, calculated values, and the special timed reading/numeracy mechanics used in assessments like JigawaUNITE.
  • Skip logic (“Show this question only if…”) controls which questions appear based on earlier answers, without writing code.
  • Change type, in a question's own Inspector, switches it to a different question type in place (e.g. single-choice to multiple-choice) — keeping its label, hint, and logic rather than deleting and re-adding it from scratch.
  • Show as the branded welcome screen, available on the first question, styles it as the survey's own introduction screen rather than an ordinary question.
  • Save to library keeps a copy of a well-built question for reuse elsewhere.
  • Changes are edited as a draft first, previewed, and then explicitly deployed to go live — a live survey already collecting real data is never changed out from under enumerators mid-edit.
Every brand-new survey starts with a GPS Location question already added, pinned to the end — remove it if a survey genuinely doesn't need one. Separately, every real submission also silently records a background GPS reading of its own, regardless of any visible GPS question — a fraud/quality cross-check you can compare against what the enumerator declared, exported as a “PassiveGPS” column alongside the regular data (section 15), and plotted on the Overview map (section 13).
Building or changing a live survey's questions is a specialist task, usually handled by whoever designed the survey originally — a single mis-set condition or calculation can silently affect real data collection. This manual covers what Build looks like and does in broad strokes; day-to-day admin use rarely requires touching it.

15. Data

Where collected data leaves Build and becomes something you can review, audit, or analyse — a toggle at the top switches between Interviews (the submissions table) and Audio recordings (playback of recorded clips).

The Data tab, Interviews view
Data → Interviews — every submission, with status chips, a column picker, and per-column search.
  • Status chips (All, Waiting for review, Flagged, Approved, Sent back) filter the table; counts update live.
  • Show answer labels switches stored choice values between their raw codes and their human-readable labels.
  • Columns opens a searchable checklist of every real column on the survey — useful since a survey with repeat sections or timed reading passages can easily have hundreds.
  • Download CSV and Stata (.do) export the table as shown, and a companion file that imports it into Stata with variable/value labels applied.
  • Every row's own detail view carries a Validation status (Pending / Approved / Rejected) — see Reviewing, rejecting, and redo under Cases, section 21.
  • When a phone reported that a recording failed (the microphone didn't start, or the recording came out empty), a Recording problem column appears, naming the question and the reason. It only appears once at least one interview has a problem.
The Data tab, Audio recordings view
Data → Audio recordings — Question, Pupil, School, Class, Enumerator and Length for every clip, with a Pick for audit action.

Audio recordings lists every captured clip with the real passage title it belongs to (e.g. “Passage 1 (Our Pond)”), plus who it was recorded by and for. Pick for audit marks a clip for a human quality listen; the text above the table tracks what share of interviews have been picked at random so far. The search box finds a clip by pupil name or interview ID.

16. Reading AI

Only shown on surveys with a scored reading passage, like JigawaUNITE. Every time an enumerator records a pupil reading a timed passage, Lens sends that same recording to an AI speech-recognition model in the background, producing its own independent reading score — shown here next to the enumerator's own live-recorded score, for comparison.

The Reading AI tab, Readings view
Reading AI — every scored reading, human WCPM next to AI WCPM, with a direction-aware gap.
  • Four stat tiles summarise the whole survey: readings AI-scored, readings still awaiting scoring, the mean WCPM difference, and the share of readings within ±5 WCPM of the human score.
  • A Readings / Analytics toggle switches between the one-row-per-reading table and an aggregate view.
  • Date, Pupil, Class, Passage, School, and Enumerator identify exactly which child and which reading a row belongs to — each pupil normally appears twice, once for the shared baseline passage and once for their own grade-level passage.
  • Human WCPM and AI WCPM are measured in the same units (correct words per minute) so they compare directly. Difference and Status flag agreement — Close match, Review, or Large gap — and the difference shows its direction (AI higher, or AI lower, than the human) rather than just a raw gap size, since the two kinds of disagreement usually have different causes.
  • The search box finds a reading by date, pupil, school, enumerator, status, or timing check. Download CSV exports the whole comparison table, including the AI's own transcript and a Timing check column.
  • The AI treats British and American spellings (“mom” for mum), digits for number words (“2” for two), and close spellings of names in the passage (“Moussa” for Musa) as the same word. It never accepts a different word.

Timing checks

Both the enumerator's score and the AI's use the reading time from the enumerator's timer, so a timing mistake changes the score whoever marks the words. After each reading, Lens compares the timer with the recording and flags two problems:

  • Timer ran on (finished at 36 s): the pupil read the last words of the passage well before time was up, but Finish was not pressed, so the full 60 seconds was used. The score is too low.
  • Started late? (opens at word 9): the recording begins part-way into the passage, so the pupil was probably reading before Start was tapped. The reading time is too short and the score too high.
Reading AI stat tiles including Timing to check
A fifth tile, Timing to check, counts readings with a timing flag. It appears only when there is at least one.
Timing flags under a reading's status
The flags appear under each reading's status. Type “timer” or “late” in the search box to list them.
Timing flags explained in a reading's detail
Opening a flagged reading explains what the audio shows.

Flags never change a score. They tell a supervisor which recordings to listen to and which enumerators need a reminder about Start and Finish (see section 5). Readings scored before 17 Nov 2026 were not checked.

Clicking any row opens its full detail: the recording itself (with an Open in Drive link), and the same passage marked twice side by side — once as the enumerator marked it live during the interview, once as the AI's transcript matched against it, word by word.

Reading detail: human marking and AI marking, word by word
A single reading's full detail — human marking on top, AI marking below.
Important: AI scores are an assistive signal, not a verdict — the scoring pipeline has no speaker diarization, so an enumerator speaking a stalled pupil's missed word aloud (per the “say the word, mark it incorrect” protocol) can still be credited as correctly read by the AI. Treat a flagged reading as a starting point for a listen, never a final answer on its own. Automatic speech recognition is also still noticeably less accurate on Nigerian-accented English than on the accents it was mostly trained on, so meaningful gaps are expected and don't by themselves indicate a problem with the enumerator's work.

The Analytics view

The Analytics toggle switches to an overall summary of how often the AI and the human agree, then two ranked breakdowns: By passage (sorted by how many readings have been scored, to see whether a specific passage is unusually hard for the AI to follow) and By enumerator (sorted by average gap, largest first — a starting point for a supervisor to pull up a real reading and listen, not a verdict on its own; a high gap is just as often the AI struggling with audio quality or background noise as anything about how the enumerator scored). Passages or enumerators with fewer than 2 scored readings are grouped into an “other” row rather than shown individually.

17. Reports and QA

Three sub-tabs, all scoped to this one survey: Summary report, Quality checks, and Downloads.

Summary report

Reports and QA, Summary report sub-tab
Summary report — a chart, a table, and a plain-language finding for every chosen question.

Choose questions opens a searchable checklist to pick which questions appear here and in the downloadable PDF; Download PDF exports the same view as a formatted document. Under every question's own chart and frequency table sits a short, plain-language line — a quick read of what the real numbers actually show, generated directly from the counts shown right below it, never a guess or an AI-written summary.

Quality checks

Reports and QA, Quality checks sub-tab
Quality checks — one row per automatic check, each with a count and an OK/REVIEW chip.

Runs automatically against every submission: duplicate respondent entries, submission timing outliers, custom rules (set per survey in Settings, section 20), fluency-vs-time-remaining flags, Reading AI-vs-human large gap flags, straightlining (an enumerator or pupil giving the identical answer across a whole battery of questions), GPS impossible-travel flags, the same pupil interviewed more than once, recordings that failed (Sheet 12 of the Excel report), and reading timing checks (“Timer ran on” and “Started late?”, Sheet 13; see section 16). View next to any row drills into the flagged rows themselves. Counts here always match the downloadable QA report in Downloads.

Downloads

Reports and QA, Downloads sub-tab
Downloads — every export this survey offers, in one place.
DownloadWhat it is
Data (CSV)Every interview, one row each, with the Enumerator and background GPS columns.
Stata (.do)Imports the CSV into Stata with variable and value labels.
Summary report (PDF)Charts and a one-line finding for each chosen question.
QA report (Excel)All quality checks, plus a daily breakdown by school and enumerator.
Audio recordingsEvery recording, filed by interview and question, as a real index of clip links.
Questionnaire (PDF)A printable copy of every question.
Spreadsheet formThe survey as a standard spreadsheet form, always up to date.
Reading AI comparison (CSV)Human and AI scores with the AI transcript (reading surveys only).

18. Lens AI Assistant

A floating helper, available from any screen in the admin console — look for the Lens AI Assistant bubble in the bottom-right corner. It never covers what you're working on: clicking it opens a panel over the top, and whatever was behind it is still there once you close it again.

The Lens AI Assistant launcher bubble
The launcher — visible on every screen, drag it out of the way if it's covering something.

The panel has three tabs:

  • Ask Your Data — type a plain-English question about a survey's real submissions (“Which school has the most submissions?”, “How many submissions this week?”) and get a real, computed answer, not a guess. You can ask it to break a count down by more than one thing at once (e.g. “tabulate submissions by LGA, school, and enumerator”), and it resolves real choice labels (like an LGA name) rather than showing raw stored codes. A “Show query” link under each answer discloses exactly what was counted, for anyone who wants to double-check it.
  • How do I… — ask a general question about building or using Lens, grounded in this same manual's own content. If you ask to see a worked example (e.g. “show me how a repeat group works in XLSForm”), it can also sketch a real worked spreadsheet example alongside its plain-English answer.
  • Draft a question — describe a question you want to add in plain English, and it drafts it two ways at once: a live preview card you can add straight to the survey you're building, and the equivalent raw XLSForm spreadsheet rows, for anyone editing a survey that way instead.

Here it is answering a real question about JigawaUNITE's own data:

Lens AI Assistant answering a real question
A real answer, computed from the survey's own real submissions — not an AI guess.
Important: nothing the Assistant drafts or suggests is ever saved or applied automatically — a drafted question only becomes real once you explicitly add it to the survey, and an answer about your data is just that, an answer, never a change to anything. Treat it as a fast way to ask a question you'd otherwise have to do yourself in a spreadsheet, not as something that acts on your behalf.

19. Share

Everything needed to get a survey in front of real enumerators and track who's collecting it, in one tab.

The Share tab: QR code, links, APK, who collects, roster, and Targets
Share — a card for each way of reaching enumerators, plus who's assigned and the survey's roster and targets.
  • Open on a phone — a real QR code and the plain mobile link, with a one-tap Copy link.
  • Fill on a computer — a desktop-styled link for office-based data entry or backfilling; it writes real submissions, exactly like the phone link.
  • Android app — a direct link to download the Lens APK, for devices that prefer an installed app over a browser tab (see section 1 for install steps).
  • Who collects — lists the people and the shared device login with access to this survey, and “Manage people” jumps to Team (section 22) to change who.
  • Name list on the phone — one name per line, the roster an enumerator picks from on the “who is collecting this response” step; leave it empty to skip that step entirely. Type it in and Save roster, or Upload CSV for a long list.
  • Targets — set a real “expected” total that feeds the Overview tab's progress tiles and by-school bars (section 13): either one flat number, or a per-school list (typed in, or uploaded as a two-column CSV) for when different schools genuinely expect different amounts.

20. Settings

Configuration for this one survey, grouped into a handful of cards — each with its own Save button, so changes only take effect once saved.

The Settings tab: General and Audio recording cards
Settings — General (name, project, kind) and Audio recording, open on JigawaUNITE.
  • General — the survey's name, its project, and its kind (e.g. Assessment, Survey), each saved independently.
  • Audio recording — whether enumerators see a red recording indicator while a clip is captured; the consent statement read before recording; and a searchable checklist of which questions record a clip at all (shown here as “7 of 61 selected”). The same checklist is also where a question gets tagged as the source of the Pupil, School, or Class columns shown on Data → Audio recordings (section 15).
The Settings tab, Advanced section expanded: version History
Advanced, expanded — History, with every past version and a one-click Restore.

An Advanced section, collapsed by default, holds the more technical, rarely-touched configuration: Respondent ID (which answered questions get joined into an audio clip's filename), custom QA rules on top of the automatic checks (section 17), and History — every past version of this survey's questions, each restorable with one click.

Below Advanced, a Duplicate card creates a completely independent copy of this survey — same questions, roster, audio/QA settings, and preload datasets — that you can edit freely without touching the original or any of its collected data. A Danger zone card holds Archive (hides the survey without touching its data or configuration; reversible) and Delete (permanently removes the survey and every submission ever collected for it — no undo).

21. Cases

Reached from Cases in the left sidebar, separate from any one survey. Cases is for fieldwork where you already know, in advance, exactly which households, schools, or respondents need visiting — a real sample, not just whoever an enumerator happens to find. It links that known list to one or more surveys, assigns specific cases to specific enumerators, tracks who's done what, and adds a reviewer workflow (approve, reject, redo, reassign) on top of ordinary submissions.

Cases: case list, linked surveys, and Assign cases
Cases — case lists on the left, linked surveys and assignment on the right.
  • Case lists — a case list is a real sample you upload as a CSV: one row per case (household, school, respondent — whatever your fieldwork unit is), identified by a column you choose (e.g. case_id). + New case list walks you through naming it and picking that column; + Add more cases (CSV) appends another CSV of new cases to an existing list later, without disturbing anything already assigned.
  • Linked surveys — check which survey(s) should pull their case picker and quota from this list, then Save linked surveys. One list can back more than one survey (e.g. a main visit survey and a separate back-check survey for the same sample) — see Back-check below.

Assigning cases

Once at least one survey is linked, the Assign cases table lists every case in the list, scoped to whichever linked survey is selected in the dropdown above it (assignment and status are tracked separately per survey, since the same case can be at a different stage on each one). Tick the cases you want, choose an enumerator (or “— Unassign —”) from the dropdown, and click Assign selected.

Below the assignment table, a Quota panel shows a live progress bar per individual enumerator account — completed vs. remaining, computed directly from real assignment and submission data. Cases assigned to the shared enumerator login appear as one pooled, unassigned-looking count instead of a personal bar, since Lens can't tell two people apart on a shared login (see Team, section 22).

Back-check

A back-check is a real second, independent interview for a random share of a survey's completed cases — a standard fieldwork quality check, confirming an enumerator's visit actually happened and was recorded honestly. Pick another survey already linked to this same case list as the back-check destination, a rate (e.g. 10%), and a supervisor (only real individual accounts with the Supervisor role appear here) to assign it to. That supervisor then sees the selected cases in their own ordinary case picker on the destination survey. Once they've submitted, a View comparison link appears next to that case, showing the original and the back-check answer side by side — mismatches highlighted.

The enumerator's side: picking a case

On a survey linked to a case list, an enumerator sees a case picker before the first question — a separate screen from the roster (section 2), since “which case is this” and “who is physically interviewing” are two different questions. Only that person's own assigned, not-yet-completed cases are listed; a completed case drops off the list automatically.

Enumerator app case picker, one case showing Needs redo
The case picker — a rejected case comes back labelled “Needs redo”, separate from an ordinary new “Pending” case.

A case sent back for a redo (see below) pre-fills the enumerator's own previous answers, so they only have to correct what was actually wrong rather than starting over — each pre-filled field is marked with a small note to that effect, editable like any other answer.

Reviewing, rejecting, and redo

From a case-managed survey's Data tab (section 15), opening any record's detail shows a Validation status control — Pending, Approved, or Rejected. Setting a case to Rejected immediately reopens it in the enumerator's case picker as “Needs redo”, with their old answers carried over as a starting point. A Redo goes to dropdown then appears, letting you send that redo to a different enumerator instead of the one who submitted it originally — useful when the original enumerator isn't available, or the mistake suggests someone else should take another look. A reviewer comment box lets you note why, for your own records.

Record detail modal showing Rejected status and the Redo goes to dropdown
Rejecting a record reopens it for redo and reveals who it's going back to.

Because a redo creates a brand-new submission rather than editing the old one, the Data tab hides earlier, superseded attempts by default once a case has been redone — a note at the top says how many are hidden, with a Show history link to reveal them again (and Hide earlier attempts to put them away).

Preview (section 12) skips case selection entirely on a case-managed survey — cases are assigned to real individual enumerator accounts, and a preview session has no real case of its own to pick. You'll land straight in the questions, with nothing pre-filled.

22. Team

Reached from Team in the left sidebar — everyone who uses Lens, what they can do, and which surveys they see, in one unified table.

The Team tab: people table and Shared device login
Team — every person, their role and surveys, plus the shared device login below.
  • Role chips (Everyone, Enumerators, Supervisors, Admins, Clients) filter the table, each with its own count. The table itself shows each person's name, role, which survey(s) they see, whether they have their own login, and when they were last active.
  • + Add person creates one individual account at a time — a personal login for one data collector, supervisor, admin, or client, letting you restrict that person to only the survey(s) you assign. Pick a password yourself, or let the person set their own via a one-time link shown right on the page after creating the account (share that link with them however you like; Lens doesn't send it automatically).
  • Add many (CSV) creates a batch of accounts at once. A batch can mix different survey assignments in one upload: add a forms column naming the real survey title(s) for that row (semicolon-separated for more than one) — any row with that column filled in gets exactly those surveys instead of whatever's checked by default. You can upload a new batch at any time, including after fieldwork has already started — it only ever creates the new accounts, never touches anyone already working.
  • Clients are outside clients given a restricted, read-only view limited to their own project's surveys.

Below the people table, the Shared device login card holds the single username/password most field tablets sign in with, plus which surveys are visible on it. Update login changes the username or password; the checklist underneath picks which surveys that shared login can see.

Because the shared login can be used by anyone with the password, Lens cannot tell two people apart on it — to restrict data collection to specific individuals, give each person their own personal account instead.

23. Org Settings

Reached from Settings in the left sidebar — configuration for the whole organisation, not any one survey.

Org Settings: Data backups, AI features, Storage, and Organisation cards
Org Settings — Data backups, AI features, Storage, and Organisation.
  • Data backups — every real survey's submissions are automatically copied to a dated Google Drive folder once a day, separate from the system's own database, as insurance against an accidental mistake. This card shows when that last happened, whether it succeeded, and how many surveys were backed up, with an Open in Drive link straight to it. Run backup now triggers an extra one on demand.
  • AI features — organisation-wide toggles for Reading AI (section 16) and the Lens AI Assistant (section 18), plus a live count of how many AI calls have been made today against the shared daily free-tier cap.
  • Storage — confirms audio, photos, and signatures are stored in a Google Shared Drive, with a Change button to point somewhere else.
  • Organisation — the logo and colours shown on every survey's welcome screen, editable from here.
Backups run automatically — you never need to click anything for them to happen. Run backup now is purely optional, for an extra on-demand backup or a quick confidence check.

Appendix — Frequently asked questions

Q: What happens if I lose internet connection partway through an interview?

Nothing changes about how you work — keep answering normally. Everything saves to the device as you go, and the finished response moves to Ready to Send until a connection is available.

Q: What if I accidentally tap the wrong word during a reading passage?

Tap the same word again to un-mark it — this works at any point before you tap Finish.

Q: Why did the app say “Time is still running” when I pressed Finish?

You pressed Finish before the time was up and before the pupil reached the end of the passage. If the pupil has only paused, tap Continue, point to the next word and say “Please go on”. If the pupil has finished or genuinely can't go on, tap End now. The timer is paused while the box is showing.

Q: I saw a red “Not recording — check the microphone” warning. What should I do?

The phone isn't recording. Check the app is allowed to use the microphone and nothing is covering it, then tell your supervisor. The interview is flagged so they know its recording is missing.

Q: My supervisor says a reading was flagged “Timer ran on” or “Started late?”. What does that mean?

Lens compared your timer with the recording. “Timer ran on” means the pupil finished early but Finish wasn't pressed, so the score is too low. “Started late?” means the pupil was probably reading before you tapped Start, so the score is too high. Tap Start as you say “Begin”, and Finish the moment the pupil reads the last word.

Q: Can I redo a reading passage if the first attempt went wrong?

Once a passage is finished, that attempt is what gets recorded. If a genuine mistake happened, use Save & Exit and speak to your supervisor about the correct process for your project, since this can vary by evaluation.

Q: Why does the pupil's name sometimes not appear in the learner list?

The list is pre-loaded from the school's official roster ahead of time. If a pupil is missing, double-check the correct School and Grade were selected earlier — the list is filtered by both.

Q: Does Reading AI change a pupil's official reading score?

No. It's shown purely for comparison. The enumerator's own live-recorded score is what always counts as the official result.

Q: Do I need to refresh the page to see new submissions or changes come in?

No — whichever admin tab you're currently viewing (Overview, Build, Data, Reading AI, Reports and QA, Share, Settings, Cases, or Team) quietly checks for updates every 20 seconds on its own. It pauses automatically while a modal like Build or the Data table is open, so it never disrupts work in progress, and while the browser tab itself isn't in front.

Q: Who can access the admin console?

Only accounts created by an existing administrator, from Team. Field enumerators do not need — and are not given — admin console access.

Q: Can I test a survey without affecting real data?

Yes — every survey's Preview button (section 12) opens a fully working copy that never saves or counts anything, on Overview, in Data, or in Reading AI. Use it as freely as you like. On a survey linked to a case list (Cases, section 21), Preview skips case selection entirely and goes straight into the questions, since a preview session has no real assigned case of its own.

Q: What's the difference between a survey's phone link and computer link?

The phone link (Share tab, section 19) is the plain link real enumerators use on a phone. “Fill on a computer” opens the same survey styled for a desktop screen — handy for office-based data entry — and still writes real submissions, exactly like any other response.

Q: Is there an APK to download for the enumerator app?

Yes, for Android — see section 1 for the download link and install steps, or the Android app card on the Share tab (section 19). iPhone/iPad has no APK support; use the browser address directly instead (also covered in section 1).

Q: What's a case list, and how is it different from a roster?

A roster (section 2, or the “Name list on the phone” card on Share, section 19) just names who's physically collecting a response. A case list (Cases, section 21) is a pre-loaded, known sample — specific households, schools, or respondents — that you assign to specific enumerators and track to completion. Most surveys don't need one; it's for fieldwork where the sample is fixed in advance rather than discovered in the field.

Q: When a case is rejected and sent back, does the enumerator lose their original answers?

No — a case marked “Needs redo” pre-fills the enumerator's own previous answers, so they only correct what actually needs fixing rather than starting over. The old, rejected attempt isn't deleted either; it's kept as a hidden earlier attempt in the Data table, visible again any time via “Show history.”

Q: Can a rejected case be redone by a different enumerator than the one who submitted it?

Yes — the “Redo goes to” control in a rejected record's detail view lets you send it to any individual enumerator account, not just the original one.

Lens is under continuous development. If something in the field doesn't match this manual, please let your project's technical contact know so both the app and this guide can be improved.