How to run a public survey campaign that people actually finish

Start from the decision, not from the questions

Before you write a single question, finish this sentence: "When this closes, we will use the results to ______."

If you cannot finish it, you are not running a consultation, you are running a survey for its own sake — and people can tell. The sentence does more work than any question-design tip in this article, because it tells you which questions to cut. Anything that does not feed the decision is a question you are asking someone to answer on a bus for no reason.

Write the sentence down. Show it to whoever asked for the consultation. You will frequently discover that two people in the room were expecting different decisions to come out of it, and it is much cheaper to find that out now.

The twelve-question rule

Twelve questions is roughly two minutes on a phone. Two minutes is roughly what a stranger will give you for a cause they broadly support.

Every question after the twelfth costs you completions, and the drop-off is not evenly distributed — the people who abandon are disproportionately the people who were least invested to begin with, which means the ones you most needed to hear from. A forty-question survey does not give you more data. It gives you the opinions of the people who were already going to turn up to the meeting.

Practical ways to get under twelve:

• Cut every question whose answer you could look up. You know the population of the ward.
• Cut every question that exists because a colleague wants to know, rather than because a decision needs it.
• Use conditional logic instead of "if not applicable, skip to question 9". Ask the follow-up only of the people it applies to, and it costs the other 80 per cent nothing.
• Put demographics last, mark them optional, and explain in one line why you are asking. People answer them far more often when they know the purpose.

Choose question types that earn their answers

• Single choice for the thing you will report as a headline. It is unambiguous and it charts cleanly.
• Rating scales for intensity, but only where intensity matters. A five-point scale on eleven attributes is a chore, not a question.
• Ranking when you need priorities rather than approval. "Rank these four" produces a decision; "do you support these four?" produces four yeses.
• Open text, exactly once or twice. It is where the quotes come from, and quotes are what change minds in committee rooms. But every open-text box adds analysis time, so budget for reading them.
• Address or a map pin when the question is spatial. "Where does this happen to you?" cannot be answered by a dropdown.
• Email, optional, at the end, with an honest label about what you will send.

Write questions the way you would say them at a door. "How often do you walk or ride along the riverfront?" beats "Please indicate your frequency of active travel usage on the riverfront corridor". The second one is not more rigorous, it is just harder to answer.

Be honest about what a public survey is

A self-selecting public survey is not a representative poll, and claiming otherwise is the fastest way to have your results dismissed by the one person in the room who knows the difference.

What it is: evidence of the volume, intensity and geography of opinion among people who chose to respond. That is genuinely valuable — it is often the only structured evidence anyone in the process has — but say what it is. A line in your report that reads "1,240 self-selected responses collected between 3 May and 14 June, promoted through X, Y and Z" is more persuasive than a confidence interval you are not entitled to.

Record who you promoted through, and track which channels brought people in. When someone asks whether you only heard from the usual suspects, you want a real answer.

Build the distribution ladder before you launch

Publishing is not distribution. Plan the ladder first, and plan it in rungs:

1. Your own list. Highest response rate, lowest effort, and it seeds the map and the idea board so later visitors arrive at something that already looks alive.
2. Partner lists. Ask three organisations to send it to theirs. Give them ready-made copy and an embed snippet so it takes them five minutes, not an afternoon.
3. Field. QR codes on flyers, posters, sandwich boards, table tents. A tablet at an event. This is where you reach people who are not on anybody's list.
4. Local press and community groups. A short piece with the link is worth more than a week of posting.
5. Paid, targeted narrowly. A small geo-targeted boost fills the gaps your other rungs missed. Watch which postcodes are quiet and spend there.

Do not fire all five on day one. Stagger them weekly so you have something new to say and can see which rung is actually delivering.

The first 48 hours

Watch the dashboard closely, then fix the obvious things fast:

• Completion rate under 60 per cent usually means a question is doing damage. Find where people stop and cut or rewrite it.
• A flood from a single source may be a supporter campaign or may be a bot. Rate limits and honeypots handle the crude attempts; the moderation queue handles the rest.
• Open text saying "none of these options fit" means your choices were written from inside the organisation. Add an option, or add an idea board and let people write their own.

Changing a question mid-flight is not cheating, as long as you say so in the write-up. A survey nobody finishes has no methodology to protect.

Moderate every day, briefly

If the public can post, moderate it. Ten minutes a day is enough for most campaigns: clear the flagged queue, hide anything abusive or identifying, restore the false positives.

Two rules that save arguments later: publish your moderation standard before you launch, even if it is three sentences, and hide rather than delete, so the record still exists when someone claims you censored them.

Close it properly, and publish

Set an end date, say it on the page, and honour it. Then do the part almost everyone skips: tell the people who responded what happened.

Publish the results — a public results page, a blog post, a slide at the meeting, all three. Show the counts, the map, and a handful of representative quotes including the ones that disagreed with you. Say what the decision was and how the responses fed into it.

This is not a courtesy. It is the single biggest determinant of whether people answer the next one. Communities do not stop engaging because they are apathetic. They stop because the last five times they answered something, nothing visibly came back.

The pre-launch checklist

• The decision sentence is written and agreed.
• Twelve questions or fewer, tested on someone outside the project.
• Completed the survey yourself on a phone, on mobile data.
• Every image has alt text; every field has a visible label; the map has an address-entry alternative.
• Distribution ladder booked, with owners and dates.
• Moderation standard published, and someone rostered to check daily.
• End date on the page.
• A plan for publishing the results, written before the first response arrives.