GeroStats

Survey Methods Handbook · Chapter 04 of 15

04 Digital methods of collecting survey data

In this chapter…

This chapter applies survey-method principles to web, mobile, messaging, app-based, mixed-mode and sensor-enabled collection.

By the end of this chapter, you should be able to…

  • choose a digital mode from the measurement and population requirements
  • plan for accessibility, digital exclusion, consent, security and mode effects
  • use recruitment controls and paradata without compromising privacy or data quality

Digital collection can include browser questionnaires, mobile-first surveys, invitations or short exchanges by SMS or messaging, app-based diaries, remote interviewer administration, and repeated or sensor-enabled measures from a participant’s device. These are survey modes, not shortcuts around survey methodology: the design still needs a defined population, defensible recruitment, tested measures, proportionate consent, secure data handling and a plan for error.

Choose the mode from the research requirement

Start with the construct, population, sampling frame, sensitivity and complexity of the task. Browser surveys avoid installation and can work across computers, tablets and phones, but an openly circulated link normally provides neither a known sampling frame nor a population response rate. Unique invitations can control access and support reminders, but identifiers must not expose personal information in URLs. Telephone or video retains an interviewer who can clarify instructions and manage complex routing, while also retaining interviewer effects and adding connectivity, recording and privacy risks.

SMS and messaging can invite sampled people, issue reminders or collect a few short responses. Make the sender and purpose clear, minimise message length, explain any participant cost, authenticate only as far as the study requires and provide a usable opt-out. Apps and ecological momentary assessment can capture experience close to the event; sensors can add location, movement or other passive measures. Installation, permissions, operating-system restrictions, battery use, device sharing and frequent prompts may change who participates and what is measured.

Digital-mode selection flow. Move from the measurement need to an inclusive, testable collection plan.
  1. DefineConstruct, population, frequency and sensitivity
  2. Check reachFrame, devices, connectivity and digital confidence
  3. Select modeWeb, mobile, messaging, app, sensor or interviewer support
  4. Add alternativesPaper, telephone or assisted completion where needed
  5. Test and monitorAccessibility, mode effects, paradata, break-off and security

Design for devices, access and inclusion

Do not transfer a paper questionnaire unchanged to a screen. Design for the smallest supported display, present one clear task at a time, use responsive controls and preserve keyboard, screen-reader, zoom and reflow access. Follow WCAG 2.2 and test with disabled participants and assistive technologies, not only automated checkers. Test routing, validation, error recovery, save-and-return, time-outs and transfer on the devices and browsers participants actually use.

Digital-only collection may exclude people with limited connectivity, suitable equipment, digital confidence, language support, private space or accessible interfaces. Mixed-mode collection can offer paper, telephone, face-to-face or assisted completion. It can reduce exclusion, but it can also create mode effects because presentation, interviewer presence and device alter comprehension and disclosure. Pre-specify how modes will be compared and combined; record mode and relevant design changes.

Recruitment, identity and data quality

Recruit through a frame or process that matches the intended inference. Advertising, social-media sharing and open links can be useful for exploratory or hard-to-reach work, but self-selection and unknown coverage must be reported. Use proportionate duplicate-response controls such as single-use tokens, server-side checks and review of completion patterns. IP addresses, device fingerprints and geolocation are personal or potentially identifying data; do not collect them by default or treat them as infallible identity evidence.

Paradata, including timestamps, contact attempts, navigation events, device information and break-offs, can diagnose usability, nonresponse and suspicious submissions. They can also reveal behaviour that respondents did not expect to disclose. Define each field and its purpose before collection, explain material uses, minimise and protect it, and set retention rules. Automated fraud rules should flag cases for review rather than silently deleting valid but unusual respondents.

Consent information must work on a small screen and remain available after participation begins. Distinguish what is necessary for the survey from optional passive or linked data. For sensors and apps, describe permissions, sampling frequency, background collection and how participation can be paused or withdrawn. Assess the lawful basis, supplier terms, international transfers and whether a data protection impact assessment is required.

Use encrypted transfer, minimise identifiers, separate contact details from responses where possible, restrict administrative access, maintain backups and recovery procedures, and test incident response. Document which organisation controls the data and what a platform supplier may retain or reuse. A free consumer platform is not automatically an approved research environment.

Summary of key points

  • Choose digital modes from the population, frame, measurement task and support required.
  • Design and test for mobile use, accessibility, routing, failures and secure transfer.
  • Provide proportionate alternatives and assess mode, coverage and self-selection effects.
  • Use paradata and identity controls transparently, minimally and with human review.

Further reading

References