Skip to content
English
Back to incquery.com
  • There are no suggestions because the search field is empty.

Smart Weights for Quality Control

Use weights to gauge and adjust the level of importance for individual quality checks.

Introduction to Weighting

When dealing with quality controls, different quality indicators might have different levels of importance. These levels can change from survey to survey.

For example, based on your audience, you might consider speeding through a survey a stronger indicator of low-quality responses than completing it at unusual hours.

Assigning a weight to each quality indicator helps ensure that critical issues have a bigger impact on the QC score than minor ones do. This improves data quality because it lets you identify and filter out unreliable responses that can distort your results from those that can provide useful information.

Smart Weights

Qsponse will assign a Smart Weight for each quality control flag. You don't need to change these weights if you don't want to—the defaults have been set to provide robust QC reporting and help ensure data quality.

Qsponse makes it easy to change the weighting if you wish to. Simply adjust the sliders to set your QC sensitivity, making each flag more or less strict.

Default Flags and Weights

These are the standard Qsponse QC checks and default settings:

  • qc_speeder (100): Flags respondents finishing in <1/3 of median time. A strong signal of low quality.
  • qc_time (25): Flagged ~11% of QC removals in our database; unusual to take a survey between 12am–6am (except for telephone interviews conducted by overseas phone operators). This calculation is based on the operator’s browser time.
  • qc_SingleS_50perc_str8liner (50): Flagged ~25% of QC removals in our database; unusual to straight-line >50% of matrices. *only checks non-looped matrices
  • qc_used_vpn (25): Flagged ~5% of QC removals in our database; unusual to use a VPN (except company VPNs).
  • qc_masked_ip (50): Flagged ~5% of QC removals in our database; suspicious to use a proxy server.
  • qc_country_ip_mismatch (25): Flagged ~9% of QC removals in our database; suspicious if country doesn’t match IP (except for travellers).
  • Attention flags (50): Flags that check for lack of attention e.g. checking age against birth year. Unlikely for attentive respondents to fail this check.
  • Contextual flags (75-100): Cheaters/bots unlikely to understand survey subject.