For a typical online survey experience, this would encompass the time from which they access the first (landing/instruction) page to when they submit their responses to the final page (typically the last survey question). LOI includes any time required to perform research "tasks", such as tasting products, reviewing concepts or media, taking and uploading pictures, etc. Time spent in the screening section is typically not included in LOI, although this practice is under discussion in the industry due to the rise of routers and screening for multiple studies at a time. LOI is usually measured as a median average time (vs. the mean which is not as accurate an indicator for how long a survey actually takes a respondent).
Ipsos Point Of View
Making surveys easier to take is essential for Ipsos to be able to successfully compete for respondents over the longer-term. LOI is one key driver of easier survey-taking.
LOI is a key trigger of respondent unengagement within the survey. In general, the longer the LOI, the more respondents speed through, straight-line questions, abandon, or decline to participate in the study at all. This behavior also differs by demographic group (e.g., young males are more likely to drop out or decide not to start long surveys). Thus, longer LOIs have several detrimental effects: skewing of results due to disproportionate sample representation, increasing time needed to complete the field, decreasing sample availability, etc.
Unengagement due to long LOIs can be mitigated by a very engaging topic or survey design, or by paying careful attention to the number of questions included that are known abandonment triggers (grids, open-ends). For the vast majority of studies, however, the longer LOI is, the higher the respondent unengagement.
Often clients attempt to collect as much data as possible in a survey, usually for cost reasons. While that approach may be cost-effective, it can lead to long, disjointed surveys that ultimately do not produce quality results/insights. Shorter surveys, focused on key research questions, are almost always better.
We recommend that online non-device-agnostic studies (e.g., for PCs) are limited to 20 minutes length. Interviewer-administered studies can generally be longer, but should remain a reasonable length.
We currently mandate that online studies using smartphone (SMP) respondents are limited to 15 minutes length.
Until recently, the industry "rule of thumb" on an acceptable online LOI has been 20 minutes. This is based on research-on-research and experience with how respondents behave at various interview lengths, across typical studies. When online surveys were only done on computers/PCs, and were mainly done in the more developed markets of NA and EU, this rule of thumb could hold across the vast majority of studies. However, the online environment is rapidly changing and these changes are impacting how respondents answer surveys and what they will tolerate in terms of LOI.
In less developed markets, for example, there may be infrastructure limitations that make the survey-taking process slower (e.g., slower internet in LATAM countries, Chinese censorship firewall). The rise of smartphones (SMPs), globally, being used to access the internet also has important implications for LOI, as surveys that are not designed especially for these devices will take longer to administer (and respondents who are not as familiar with taking surveys on SMPs will take longer to complete the survey). At Ipsos, LOI is one of the requirements for a study to be able to accept SMP respondents – LOI must be 15 minutes or under. This limit is applicable to all regions where Ipsos does business. (In addition to a 15 minute LOI, studies also must be device-agnostic in design in order to accept SMP respondents.)
It is worth noting that LOI will differ across regions – a 15 minute LOI in one country is not 15 minutes in all countries. In addition to the infrastructure differences already noted, language and respondent differences will contribute to differences in LOI. For example, when translated into another language a question may be longer or shorter (more/less words needed to explain the concept), respondents in one country may be less familiar with survey-taking and so take longer, etc. These differences should be factored in when running global studies.