So you've put your blood, sweat, and tears into your latest online survey. Once you finally got to the point where you could publish and send your initial invitations, your sigh of relief could be heard across the organization. Unfortunately your lull was short-lived because the completes rolled in quickly. After the second week of fieldwork, momentum has slowed and you are now ready to take the next step with this project. What is the next step?
Thinking it's the report? Nope.
Thinking it's analysis? Think again.
It's time to get your hands dirty. Or you can wear gloves. Your choice.
A critical piece of the any market research project is data cleaning. Depending on the quality of your panel or your list that you used to send invitations this could be a minor task or a major one. This step ensures that you gave the data a full scrub and you can be 100% confident your results are clean and accurate. Also keep in mind, garbage in, garbage out. A well-programmed survey can eliminate work on the back-end of a project.
Here are 9 data cleaning items you'll want to address in your survey preferably throughout fieldwork or once fieldwork is complete:
Tip 1: Double-check for duplicate entries.
This can range from those stuffing the ballot box to minor technical snafus where submit is clicked twice. Either way, this is a one-complete per person show in market research.
Tip 2: Review and code all other(s).
Oftentimes respondents will select other but will specify something they overlooked from the list. Recode these if appropriate. If it fits something in your list, this will tidy up your 30% of those who selected 'others.'
Tip 3: Check for straightlining (AKA flatlining.)
These survey speeders select the same response throughout (easily noticed in grid questions.) Run a quick sort or create a formula in excel to identify and flag those who selected the same number for all grid responses.
Tip 4: Check time to complete.
If the average time to complete is 10 minutes, trim the fat on those who completed it unreasonably fast. Those who can complete a 10 minute survey in 2 minutes are either super-human, a robot, or just seeking a quick reward.
Tip 5: Check the number of skipped questions.
It's okay to skip a question here or there but 12 skips? You are better off removing the case. A blank open-end might make sense if the respondent feels a response is not applicable. Tread lightly here.
Tip 6: Review the open-ends for erroneous responses.
These types of responses usually indicate the task was not taken seriously. I often wonder what language "xacawetaetascasdf" is in open-ends. Thanks but no thanks online panel respondent.
Tip 7: Review your red herring questions.
If you placed a test in the script (e.g., select "3" for this question) you've identified some easy removals. Red herrings are a great way to quickly identify speeders. This can be the first data cleaning step in your process. It likely opens up a can of worms to others on this list.
Tip 8: Review logic and sequences.
If a person said they were never a customer and then mentioned they purchased a product, dig deeper. You'll likely find inconsistencies elsewhere. Keep a close eye on continuity.
Tip 9: Run your normal spell check and grammar.
This should be your 1st or last step. The choice is yours. Either way, it's a no-brainer and a requirement when it comes to data cleaning.
Now, you are ready for analysis and reporting. Using this list of data cleaning tips will help you create a better prepared data file for next steps. Drive Research is a market research firm in Syracuse, NY. Call us at 315-303-2040 with any questions, or email us at [email protected].