Offline Data Collection in Uzbekistan: CAPI vs Paper

A field team heads to a remote district of Surkhandarya, to a mountain village where 4G exists only at two spots by the entrance road. By noon the thermometer is past 40 °C, the phones are running hot, and the stack of paper questionnaires in the trunk has become a pile someone will spend a week typing into Excel. This is the Uzbek field as it actually is — and it's exactly here that it's decided whether your data reaches the server or not. This article unpacks why offline-first data collection (digital CAPI) in Uzbekistan is not an "advanced option" but the baseline setup, and how to run it so you never lose a single questionnaire.

Why offline isn't a choice in Uzbekistan — it's a condition

The paradox of connectivity in Uzbekistan is that the country is at once highly "mobile" and highly uneven in coverage. SIM subscriptions exceed the population, and smartphones have long been the norm even in villages — but that does not mean there will be a signal at every respondent's door. 4G holds confidently in cities and district centers, yet in remote mahallas (the neighborhood communities), in the mountains of Surkhandarya and the eastern valleys, in the Karakalpak desert, or simply in the basement or ground floor of a concrete apartment block, the signal vanishes.

Add to this the fragmentation across operators. Beeline, Ucell, Mobiuz (UMS), Uztelecom, Humans — each has its own coverage map, and no single network covers everything. A SIM that catches signal on the highway may go silent in a village ten kilometers off it. So the rule is simple: do not assume connectivity at the respondent's door. Build the process as if there is no internet at all, and treat sync as a bonus that happens when you drive back out to the highway.

This is exactly what AISurvey is built for: the interviewer downloads the instrument in advance, works fully offline, and the data goes to the server when a connection appears. Here offline-first is the default architecture, not a setting you have to remember to switch on.

CAPI vs paper: the honest tally

Paper still shows up in the Uzbek field — it's cheap up front and needs no devices. But that "cheapness" is deceptive, and you have to count the whole chain, not just printing the questionnaires.

• Double entry and reconciliation. Every paper questionnaire is keyed into a database by hand — a second round of fieldwork, with its own errors and its own cost. CAPI gives you digital data immediately.
• The fabrication surface. Paper is easy to "finish filling in" at a teahouse after the shift. Digital CAPI captures a timestamp, a GPS coordinate, and (where needed) a photo or audio — fabricating that is markedly harder.
• Feedback speed. Paper sits in the bag until the end of the week; you'll spot a logic error when re-running the field is already too late. CAPI syncs — and the supervisor sees the problem the same day.
• Quality control. Timestamps and GPS are the foundation of field control; without them you verify work on someone's word. More on that in our piece on field quality control.

The conclusion isn't that paper never happens, but that its hidden cost — entry, reconciliation, fabrication risk — almost always outweighs the device savings under Uzbek conditions. For most projects, digital offline-first CAPI is simply cheaper across the full cycle.

Device fleet and SIM strategy for the Uzbek network

Since no single operator covers everything, the device fleet is a methodological decision too, not a blind purchase. Here are a few practices that save field days.

• Mix operators within a team. If one interviewer carries a Beeline SIM and another an Ucell or Mobiuz, the chance that at a given point at least someone catches a signal to sync is markedly higher. For distant trips it's worth carrying a second SIM from another operator as a backup.
• Own phone or shared tablet. In Uzbekistan an interviewer almost always has a personal smartphone, and the "work on your own device" model is cheaper. But for distant multi-day trips, a shared, pre-prepared tablet with a large battery and verified storage is more reliable than a personal phone crammed with photos and messengers.
• Find sync points in advance. As you clear the route, mark where there's a reliable network along the way — a district center, a filling station, a roadside teahouse, a hotel. This turns sync from chance into part of the plan: the team knows the data will go out, say, on the way back through the district center.
• Minimize "heavy" media where the network is weak. If the design sends a photo and audio with every interview, the upload will drag in low-coverage districts. Decide at the questionnaire builder stage where media is truly needed and where a timestamp and GPS suffice.

Prepare devices before you leave

Half of all offline failures happen not in the field but the evening before, when the instrument was "more or less downloaded." In a country where the survey point can be half a day's drive away, you can't go back for a forgotten file.

• Download the instrument and media in advance. The questionnaire, showcards, images, reference lists — all of it must be on the device before you head out, while you still have a connection. Verify this not by assertion but by opening the questionnaire in offline mode on the device itself.
• Load every language version. The Uzbek field is multilingual: ru and uz always, plus a Karakalpak version in Karakalpakstan and, in Samarkand and Bukhara, often Tajik. All needed locales are downloaded ahead of time, because you can't add a language out in the desert. For how multilingual instruments work, see the AISurvey guide.
• Check storage. Photos and audio eat space fast. Before a multi-day trip, make sure free storage covers the full media volume, or by day three the device will simply stop saving.

Power and heat: the logistics people forget

The technical side of offline comes down to physics, and physics in Uzbekistan is harsh. In summer the south — Surkhandarya, Kashkadarya, Bukhara, the Karakalpak desert — routinely exceeds 40 °C in the shade. Heat hits the hardware twice: the battery drains faster, and the screen dims in the sun to the point of being unreadable.

• A power bank is a mandatory part of the kit, not an "if it fits." Size its capacity for a full day of GPS, screen, and photo capture, not for "normal" use.
• Shade and discipline. Train interviewers not to hold the device in direct sun, to let it cool, and to work in the shade of a courtyard or indoors where possible.
• Multi-day trips need a complete offline plan. A trip to Karakalpakstan or a distant mountain district is not "we'll sync at the hotel tonight," because the hotel may not have a usable internet either. The plan must assume data accumulates locally for several days and goes out as a batch once the team reaches a reliable network.

The golden rule of fieldwork: data only exists when it's in two places. While a response lives solely on the interviewer's phone in a village with no signal, it's at risk — and your job is to get it to the second place intact.

Sync at every opportunity, but don't touch the local copy

Offline-first does not mean "we sync once a week." It means we sync whenever it's convenient, but build the process so that the absence of sync breaks nothing.

The main rule: the local copy is not deleted until the server has explicitly confirmed it received the response. No "sent and forgotten." Until there is a confirmation from the server, the single source of truth is the device, and it must not be wiped.

The second rule is idempotency. The network on the road is patchy: a response may go out, the confirmation may not come back, and the device will try to send again. Sync must be built so that resending the same response does not create a duplicate, but simply confirms an already-received record. Without this, after a trip down a highway with a jumping signal you'll get a database stuffed with twins, and you'll spend days cleaning it.

Protect data integrity

Data integrity offline rests on three simple things that every response must carry.

• A unique response identifier, assigned on the device at the moment of creation — the basis for deduplication and for idempotent sync.
• A timestamp — when the response was actually collected, not when it reached the server. This is both quality control and a way to reconstruct the chronology.
• A device identifier — so you know which tablet records came from and can trace the source of suspicious data.

This same trio — identifier, time, device — feeds field control: the supervisor sees who submitted which questionnaires, when, and from where, and catches anomalies. And here the topic of personal data arises: records with GPS and photos are data about people, which means Uzbek legislation applies to them. Where and how they are stored, what must be registered, and how to handle consent — we cover separately in our piece on personal data protection in Uzbekistan.

Shake down the offline flow before you scale

The cardinal mistake is testing your offline logic for the first time in Karakalpakstan, 1,200 kilometers from the office. Any instrument should be shaken down in a near-live mode before mass fieldwork.

• Pilot in airplane mode. Turn on airplane mode on the device and walk the entire questionnaire end to end: open offline, fill in, photo, save, exit, and re-enter. If something won't save without a network, you'll learn it at the office, not in a village.
• Check every language version offline. Open ru, uz, and where needed the Karakalpak and Tajik versions with no network — confirm the showcards and the Cyrillic/Latin variants downloaded, not just the locale you happened to test in.
• Run the sync cycle. Collect a few responses offline, go online, sync, then deliberately sync once more — there should be no duplicates. This is idempotency checked in practice.
• Give the team a short one-pager. A single sheet: what to do if the phone dies, how to confirm the data went out, whom to call when there's a problem. In a remote district with no signal, that sheet is worth more than an instruction in a chat that won't load.

A half-hour run-through like this on each device removes most field surprises. For more on accepting interviewers' work and the signals of fabrication, see our piece on field quality control.

Plan for failure scenarios in advance

Reliability isn't "everything works" — it's knowing in advance what happens if something breaks. Walk through these questions before the field, not after losing data.

1. The phone dies mid-interview. An auto-saved draft must kick in so the interviewer returns to the right question instead of starting over.
2. The app closes or the device reboots. Data in the local database must survive both — a baseline requirement for offline collection.
3. A device is lost or stolen. Here you need the ability to remotely mark the device as compromised and wipe its data, so respondents' personal data doesn't leak.
4. The team is three days in the desert with no signal. The process must calmly hold the accumulated offline responses and send them as a batch at the first reliable network — with no loss and no duplicates.

The interviewer's device kit and pre-field checklist

Everything above distills into a single sheet the interviewer runs before the car leaves reliable coverage. In the Uzbek field, where the survey point can be half a day's drive away, a forgotten detail costs an entire trip, not five minutes. Build this checklist once and make it a mandatory team ritual.

What must be on hand

• A charged device and a power bank sized for a full day. Plan battery life not for a "normal" day but for one with constant GPS, a bright screen, and photo capture. In the south — Surkhandarya, Kashkadarya, the Karakalpak desert — past 40 °C the battery drains faster than your estimate and the screen has to be cranked to maximum, so for distant trips carry two power banks per device.
• Free storage for media. Before you start, confirm there's room for all the photos and audio of the whole trip, not just day one. Clean the device in advance — out in a village with no signal you can neither download nor offload anything.
• The instrument and all media pre-downloaded. The questionnaire, showcards, images, reference lists, and every needed language version (ru/uz always, Karakalpak in Karakalpakstan, Tajik in Samarkand and Bukhara) must be on the device while you still have signal. Verify it by opening the questionnaire in offline mode on the device itself.
• A spare device or a second SIM across operators. No single network — Beeline, Ucell, Mobiuz, Uztelecom, or Humans — covers everything on its own: not the remote mahallas, not the Karakalpak desert, not the mountain valleys. A second SIM from another operator, or a spare phone in the team, turns "there's no signal here" into "someone has signal."

Field discipline and a plan for distant trips

• Sync at every signal. The moment a network appears, the data goes out — don't wait for the evening. And the iron rule: never delete the local copy until the server has confirmed it received the response.
• Map sync points along the route in advance — a district center, a filling station, a teahouse, a hotel — so upload is part of the plan, not luck.
• For distant regions, a multi-day plan. Build a trip to Karakalpakstan on the assumption that data accumulates offline for several days and goes out as a batch at the first reliable network; decide in advance which day and which point hold your first guaranteed sync.

This short pre-field ritual removes most field surprises — it costs minutes and saves whole trips.

Where offline meets the reality of the Uzbek mahalla

Technology is half the job; the other half is access. Even a perfectly prepared device is useless if the interviewer isn't let onto the doorstep. In Uzbekistan the gateway to a household is the mahalla and its active members (the chairman — raisi — the hokim's assistant, and the "yettilik," the seven), and an official letter plus a courtesy visit to the raisi decide whether the survey happens at all.

This ties directly to offline: by clearing your route with the mahalla in advance, you not only gain access but also learn where you'll be working without a network, where there's a point to sync, and where it makes sense to stay overnight. We cover access logistics and working with local authorities in detail in our piece on field logistics and the mahalla.

And one last thing. Offline-first isn't distrust of connectivity — it's respect for the field. The Uzbek field is large, hot, multilingual, and uneven in coverage, and an instrument that accounts for this from the start pays for itself with every questionnaire not lost. You can build such an instrument, with all its language versions and offline logic, in the AISurvey builder.