Survey Best Practices: Plan, Write, Preview, Host

TL;DR

• Decide the decision first. Then write the fewest questions that inform it.
• Clear wording, no leading or double questions, short length, preview on a phone.
• Host the survey on your website. You keep the traffic. Store answers once. Branding off unless you want a credit.

Why survey craft matters

You run a survey to choose something: a menu, a headline, a support hours change. Weak surveys get rushed answers (“satisficing”) or drop-off. Respect for time is the whole method.

Plan before you write

Write one sentence: what you will do with the answers, who should take it, and what “done” looks like. Example: “Survey people who used billing last month so we know whether the invoice PDF is understandable.”

Turn that into two or three objectives, not ten. Extra objectives become extra pages.

Choose data type on purpose. Open text is rich and slow to read — keep to one or two optional boxes, usually at the end. Closed lists are faster and easier to chart.

Question writing

Be short

No jargon. Put instructions in parentheses: “Select all that apply.” “Rank 1 = favorite.”

Don’t lead

Skip praise in the stem (“We’re proud of our quality — how satisfied are you?”). Ask the rating. Offer a full scale.

Don’t load

Don’t force an answer that assumes a purchase they never made. Screen first, or allow “I didn’t buy.”

One idea per question

“Price and quality” is two questions. Split them.

Avoid absolutes

“Do you always buy online?” with yes/no throws away the middle. Use frequency if you need it.

Sensitive items

Income, identity, health: optional, late in the form, with why you ask. Ranges beat exact income. Include a prefer-not-to-say path.

Question types

Pair a score (NPS, CSAT, ease) with one optional “why.” Multiple choice: say if one or many. Dropdowns help long lists on a phone. Ranking shows order, not intensity of like. Matrices save space and invite straight-lining — keep them tiny, or skip them on mobile.

Scales

Words vs numbers: words are clearer in a report (“55% said somewhat easy”); numbers are faster to tap. 5- and 7-point scales with a middle point are enough for most attitude items. Yes/no is for true binaries, not feelings.

Length and order

Aim for one sitting: often ten questions or fewer. Screen out people the form is not for. Skip logic beats a wall of “if this applies.” Put open text late. Put sensitive items late. Only require what you cannot decide without.

Incentives

Use them when you are asking a favor (research panel), not when the person already wants to complain or praise. Prepaid small thanks often beat a huge raffle. Incentives should not buy fake answers — keep the form short so they don’t race.

Before you send

• Readable type, contrast, one column.
• Say who you are, why you’re asking, and what you will do with the data.
• Preview on desktop and phone. Submit a test. Check skip logic.
• Have a colleague try to break it.

Ten-point checklist

1. Goal in one sentence.
2. Short, plain questions.
3. As short as the decision allows.
4. Few text boxes.
5. No leading or loaded stems.
6. No double-barreled items; no overlapping choices.
7. Sensitive questions optional and late.
8. Context: who, why, how data is used.
9. Incentives only when needed.
10. Preview and test vote / test submit.

Host it yourself

A survey that lives on your site keeps people in your story. You do not need a vendor skin. MyFormCapture is the inbox and the export. Use AI Form Builder or a survey generator , paste your UUID, test, then publish. Default: no branding. Credit line only if you want it.

Customer and employee notes

CSAT and NPS work when they sit next to a real moment (after delivery, after a ticket). Employee surveys get better answers when they can be anonymous and you say how results are grouped. Exit surveys: voluntary leavers, confidentiality, not the same form as a firing process.