Data quality
Bad data is worse than no data.
A thesis built on careless responses can fall apart at the defence. So every Ximopi response passes five checks before it reaches you — and the ones that fail cost you nothing.
The problem with online surveys
When a survey is shared openly, some people click through without reading, some aren't the audience you need, and some give contradictory answers. Research on careless responding has estimated that around one in ten online survey responses can be careless (Meade & Craig, 2012, Psychological Methods).
Cleaning that by hand is slow, and hard to justify to a committee. Ximopi does it automatically, consistently, for every response.
Five layers of checks
Before the survey
1. Audience matching
Only people whose profile matches your criteria — country, university, field, year, gender — can open your survey. Off-target respondents never see it.
Example: You target year-3 marketing students: a year-1 IT student never sees the survey.During
2. Attention check
Every survey has an instructed-response question inserted midway, e.g. “please choose Agree for this one.” A wrong answer means the respondent wasn't reading.
Example: Respondent picks “Strongly disagree” on the check → rejected.During
3. Timing
We estimate how long your survey takes to read and answer, and time every response. Finishing in under 35% of that time isn't humanly possible while reading properly.
Example: A 3-minute survey submitted in 25 seconds → rejected.After
4. Straight-lining
When a survey has 5 or more scale statements and someone gives every one the exact same rating, they're almost always not reading — especially when some statements point in opposite directions.
Example: “Agree” to all 8 statements, including two reversed ones → rejected.After
5. Gibberish filter
Written answers like “asdf”, “aaaa” or a single character are flagged as meaningless.
Example: “What would make you buy more?” → “xxx” → rejected.
What happens to a rejected response
The respondent sees why
Immediately, with the exact reason. No waiting, no silent penalties.
You're not charged
Rejected responses don't count toward your target and aren't deducted from your deposit.
No manual review queue
Decisions are automatic and consistent. Respondents can ask for a human review if they think we got it wrong.
You can still check for yourself
Serious researchers want to apply their own criteria. Every export includes each response's completion time and its speed ratio (actual time ÷ estimated time), so you can set a stricter cut-off, run your own consistency checks, or report exactly how you screened the data.
What our checks can't catch
No automated system is perfect. Here's what you should know:
- Someone who reads carefully but answers dishonestly will pass every check.
- Profiles are self-reported. We act against duplicate accounts, but we can't verify every student's university or year. Add a screening question if your criteria are strict.
- A respondent using an AI tool to answer may produce plausible, well-timed responses.
- Reversed items only help if you include them — we recommend at least two in any multi-item scale.
We'd rather tell you this now than have your committee find it later.