Plain-language guide to open-ended and closed-ended survey questions: what each measures, advantages and disadvantages, and how to pair them on the same survey
This guide explains the two question types in plain terms. What each one measures, when to use which, and why the survey that drives a program decision tends to pair them rather than pick one. Examples come from workforce training, healthcare feedback, and member-feedback surveys. No prior background in research methods required.
"How was the support-services help during the program?"
Average: 3.4 / 5 · Down from 3.9 last cohort. The number tells you something dropped. It does not tell you what.
"What got in the way of the support that would have helped?"
"The childcare hours did not match my class schedule. I had to leave early three weeks in a row."
"Career counselor changed twice. By the time I got to know the third one the program was almost over."
"It was good but the bus route to the satellite campus took ninety minutes one way."
The clearest way to see what each type does is to write the same question stem twice and look at what each version produces. Same program, same respondents, same survey moment. The form changes; the data changes with it.
"How was the program for you?"
A count distribution across five buckets. Average: 3.6 / 5. Slot in a dashboard, segment by cohort, compare to last cycle.
A column of paragraphs. Each one in the respondent's own words. Read or coded against a scheme; counts only after coding.
"Mostly good, except the Friday sessions ran late and conflicted with my pickup time at the daycare. After week four I switched to skimming the recordings."
"The instructor for the second module was the best teacher I have had since high school. The first instructor I never figured out."
"Helped a lot but the bus to the satellite campus was ninety minutes one way. I started carpooling with another participant which is the only reason I finished."
Prevalence, distribution, and trend across cohorts. The score the executive brief leads with. The number that segments cleanly by demographic.
Reasoning, context, and themes the survey designer did not anticipate. The sentence the funder will quote. The barrier you forgot to put on the closed list.
The why behind the score. The barrier that did not appear on the predefined list. The reason the OK ratings stack up at the middle.
Comparable trend. Speed to insight. The clean roll-up the dashboard needs. Most respondents on mobile do not type a paragraph.
Each type's gap is the other type's strength. The honest survey writes the closed-ended question and the open-ended question side by side, on the same survey row, asking the same respondent both at once. The score and the why arrive together.
The two columns above are the same forty-five respondents, the same program, the same survey moment. The form decides the data shape, and the data shape decides what you can do with it. Closed for the dashboard, open for the brief, both for the program decision.
Five questions worth answering before designing any survey. The same answers a research methods course covers, in plain language and with examples that match the surveys teams actually run.
An open-ended question is a survey question that lets the respondent answer in their own words rather than picking from a list. The response is free text, anywhere from one sentence to several paragraphs. Open-ended questions capture reasoning, context, and themes the survey designer did not anticipate. They are slower to analyze than closed-ended questions but they are the only way to hear the response in the respondent's own words.
The phrase "open ended questions meaning" gets searched almost as often as "open ended questions" itself, which is a sign that the term sounds technical. It is not. An open-ended question is a question that does not box the answer.
A closed-ended question gives the respondent a fixed set of response options. Common forms include yes-no questions, multiple choice, Likert scales (1 to 5 or 1 to 7), and ranked options. The data is structured: every response slots into a known bucket, counts roll up immediately, and trends across cohorts compare cleanly.
Closed-ended questions are sometimes called fixed-alternative questions in older research literature. Both names mean the same thing. The "fixed" part is the response options; the "alternative" part is the menu the respondent picks from.
A closed-ended question gives the respondent a fixed set of options to pick from. An open-ended question lets the respondent answer in their own words. The trade is structure for depth. Closed produces data that counts immediately; open produces data that has to be read or coded but captures reasoning and unanticipated themes.
The honest survey uses both, paired so the closed answer says what and the open answer says why.
Use an open-ended question when you need the response in the respondent's own words, when you want to surface themes you did not anticipate, or when a closed-ended option list would force a real answer into the wrong bucket.
Open-ended questions belong right after a closed-ended score, asking for the reason behind the score. They also belong as a single "anything else we should know" prompt at the end of a survey. They do not belong as the bulk of a survey: response rates drop sharply when respondents face more than two open prompts on one form.
Use a closed-ended question when you need to measure prevalence, compare across cohorts, run a quick mobile survey, or report counts to a funder, regulator, or executive team.
Closed-ended questions are the right tool when you already know the response options that matter (yes-no, a defined Likert scale, a known list of barriers). They produce data that ships fast, segments cleanly, and benchmarks against external standards. Most surveys are 80 percent closed-ended for these reasons. The open-ended questions are reserved for the spots where reasoning matters.
Same thing, two names. "Fixed-alternative" is the older academic term used in research methods textbooks; "closed-ended" is more common in market research and program evaluation. If a textbook uses one and a survey tool uses the other, treat them as identical.
Both are subtypes of closed-ended. Multiple choice picks one (or several) from a list of distinct options. Likert picks a point on a scale (typically 1 to 5 or 1 to 7) measuring intensity or agreement. Different shapes, both structured.
Yes-no is the most constrained closed-ended form: two options, no middle. Scaled (Likert, 0-10, NPS) gives gradation. Yes-no answers prevalence questions cleanly; scaled answers intensity questions cleanly.
A partially structured question gives the respondent a list plus an "Other: please specify" field. Hybrid: closed for the common cases, open as a catchall. Useful when the team has a strong but incomplete sense of the response options.
The bulk of the survey is closed-ended. The open-ended questions are placed where reasoning earns its keep. These six rules are how teams keep response rates high and the why available at once.
Closed-ended question, then the open follow-up. Same screen.
The closed question delivers the score. The open question right after asks for the reason behind it. Two fields, side by side or stacked. The respondent already has the score in mind from the previous field, so the why is fresh and arrives in one or two sentences instead of a paragraph.
Why it matters: placing the open right after the closed is what gets the why typed at all.
"How many" and "how often" are closed-ended jobs.
Anything that needs a percentage, a count, or a distribution belongs in closed-ended form. Open-ended responses can be coded into counts later, but the work that turns a paragraph into a percentage is expensive. If the survey already knows the response options that matter, closed is the right form.
Why it matters: counts that ship the same day come from closed questions, not coded open ones.
Closed-ended forces every response into a known bucket.
Closed-ended questions can only return options the team thought of in advance. The barrier the team forgot to put on the list does not show up in closed data. Open questions are how the survey hears the things that did not occur to the survey designer. Use them where unanticipated themes would actually change a decision.
Why it matters: only open questions can return data the closed list did not anticipate.
Each one drops completion by 5 to 15 percent.
Open-ended prompts are expensive on response rate. The first one drops completion by a few percent. The third drops it past where the trend line stays comparable across cohorts. Keep one open after the most important closed score, and one "anything else" at the end.
Why it matters: response rate decides whether the data represents the cohort or only the self-selected slice of it.
A trend line needs the same scale every wave.
Cross-cohort comparison rests on the response options being identical wave to wave. Closed-ended questions deliver that for free; the Likert scale or yes-no is the same in cohort one and cohort six. Open-ended responses drift between coders and across waves. Use closed for the trend; use the cohort's open responses to explain a movement in the trend line.
Why it matters: longitudinal comparison stays honest only on closed data; open data is for explaining what the trend means.
"What got in the way?" beats "Tell us about your experience."
A narrow open prompt gets a usable answer in one or two sentences. A vague open prompt gets either a wall of text from the respondent who really wants to be heard, or "good" from everyone else. The narrower the prompt, the more comparable the open responses become across respondents. "What got in the way" is sharp; "tell us about your experience" is not.
Why it matters: a sharp prompt produces responses that are short, comparable, and actually answer the program question.
Six design decisions a survey writer faces. The broken pattern is treating these as either-or. The working pattern is putting both types where each one earns its place on the survey.
| The choice | Broken way | Working way | What this decides |
|---|---|---|---|
|
Measuring satisfaction
A program rating, an NPS score, a Likert on instructor quality. |
Broken
Closed-ended only. The team gets a score that ticks up or down between cohorts. When leadership asks why the score moved, the survey has nothing to point to. Or the opposite: an open prompt asking for a vague reaction without a score, leaving the team with paragraphs and no number. |
Working
5-point Likert plus one open prompt right after. "How was the program for you?" then "What was the main reason for your rating?" Score for the dashboard, reason for the brief. |
Whether you can explain a score change when the funder or program team asks why. |
|
Catching unexpected feedback
The thing the survey designer did not think to ask. |
Broken
All closed-ended with predefined option lists. Whatever did not appear on the list disappears from the data. The barrier nobody anticipated cannot show up because there is no field to express it. |
Working
Closed for the predefined list, plus one open prompt for "anything else." Place at the end of the survey. Two-thirds of respondents will skip it; the third who fills it in surfaces the unanticipated themes. |
Whether new themes surface or get forced into old codes. |
|
Comparing cohorts over time
A six-month program running quarterly cohorts. |
Broken
All open-ended. Themes drift between coders and across waves. The same code might mean different things in cohort one versus cohort five. Trend lines built on coded open data look comparable but are not. |
Working
Closed-ended for the trend, open-ended for the why behind a movement. The closed Likert stays identical wave to wave, so the trend line is real. Open responses explain what the trend means in cohort-specific terms. |
Whether trend lines stay valid across cohorts and waves. |
|
Building the executive brief
A one-page summary that lands on a leadership desk. |
Broken
All closed-ended. The brief is numbers and bar charts. Leadership reads it, asks "what's behind this," and sends it back. Or all open-ended: the brief is anecdotes without any prevalence claim, which leadership reads and asks "is this representative." |
Working
Closed counts plus three to five quoted open responses. The number sets the scale; the quotes give it texture. The brief reads as a whole instead of leaving the reader to assemble both halves. |
Whether the brief tells a story or only ships data. |
|
Mobile-first short survey
A 90-second survey on a phone, single page. |
Broken
Three or four open prompts on mobile. Each one requires typing on a small keyboard. Drop-off compounds: 80 percent start the survey, 50 percent reach the second open, 30 percent finish. The data represents whoever was patient enough to finish. |
Working
Closed-heavy with one strategic open at the end. Tap-to-answer Likert and yes-no for the bulk; one open prompt last so respondents who drop off after the closed questions still leave a usable response set. |
Whether response rate stays above 60 percent on mobile. |
|
Reporting to a funder or regulator
An accountability report with a tight format expectation. |
Broken
All open-ended responses summarized as a narrative. The report reads as anecdote and the funder asks for prevalence claims the data cannot back. Or all closed: the report has counts but cannot quote a participant in the funder's exact language. |
Working
Closed counts as the headline numbers, open responses as quoted evidence. Standard funder reports want both. The closed answers what was measured; the open quotes show participants' own framing. |
Whether the report holds up to a thirty-second scan and a thirty-minute review. |
The first row controls the rest. Pairing closed and open at the survey-design stage is what gives the trend, the brief, the funder report, and the next-cycle program decision all the same source of truth. Skip the pairing on the survey, and downstream every team patches together two half-surveys that never agree on what the cohort said.
Three cohorts of trial and error at a workforce training program. First an all-open survey. Then an all-closed survey. Then the paired version, which kept the response rate of the closed and the why of the open.
We run six-month workforce training cohorts. Forty-five participants per cohort, three cohorts a year. The post-program survey covers four areas: skills gained, instructor quality, support services, what to change. Year one we ran all open-ended questions. Response rate was 41 percent. Reading 180 responses across four prompts each cycle took our evaluator two weeks. Year two we switched to all closed Likert scales. Response rate jumped to 76 percent. We had numbers in 90 seconds. Then the funder asked why the support-services rating dropped in cohort two. We had a number that went down. We had no answer. Year three we paired them: a five-point Likert plus one targeted open prompt per area. Response rate held at 73 percent. The number told us where to look. The open response told us why. We answered the funder.
Workforce-training program evaluator, after three cycles of trial and error.
What axis
Closed-ended score
The five-point Likert per area. Counts roll up immediately. Trend across cohorts is the dashboard's first chart. The number every funder report leads with.
Why axis
Open-ended reason
One targeted prompt per area, asked right after the score. The respondent already has the rating in mind, so the reason arrives in one or two sentences. The thing that explains a number going down.
Skills gained, instructor quality, support services, what to change. Each area scored on a 1-5 scale the respondent taps once. The trend line across cohorts holds because the scale is identical wave to wave.
"What was the main reason for that score?" placed directly under each Likert. Sharp prompt, fresh reasoning. Response is one to three sentences in 80% of cases. AI codes the responses against a six-code scheme.
Mobile-friendly. Average completion time: 4 minutes. The closed Likerts carry the bulk of the cognitive cost; the open prompts attach to a fresh thought instead of asking the respondent to start fresh.
"Why did support-services drop in cohort two?" Within an hour: a count of how many cohort-two respondents flagged the support-services area, plus the three most-mentioned themes from their open responses. The brief writes itself.
Four open prompts on the survey. Mobile drop-off compounded: 80 percent started, 60 percent reached prompt three, 41 percent finished. The 41 percent who finished were the most-engaged participants. The data over-represented their experience.
Reading 180 free-text responses across four prompts (about 720 paragraphs total) took the evaluator two weeks per cycle. The themes were rich, but they landed two weeks after the cohort closed. Programmatic adjustments stayed at quarterly cadence.
Likert scales replaced every prompt. Response rate jumped. Numbers shipped in 90 seconds. Then the funder asked the why behind a drop. The team had nothing to point to. They sent a follow-up survey to a sample of cohort-two participants asking the question that should have been on the original.
The follow-up landed three months after the cohort closed. Response rate on a follow-up survey: 18 percent. The themes the team eventually surfaced were the same themes that would have arrived in year three's paired version, but at six weeks of additional researcher time and worse data quality.
The team's year-three improvement is not a better instructor, a better cohort, or a better dashboard. It is the same data team running the same program with a survey form that pairs each closed score with one targeted open prompt. The fix lives in the survey design, not in the analysis pipeline. The pairing has to be there before the first respondent fills out the form, because the why field cannot be added back in after the survey has closed.
Three different organizational shapes and three different data flows. Same architecture: closed for the score, open for the why, paired on the same survey. The output of each survey changes per context. The structure does not.
Cohort-based programs running 30 to 60 participants per cycle, three to four cycles a year, with end-of-program evaluation surveys.
Workforce training programs run six- to twelve-month cohorts and survey participants at program close. Four to six survey areas are typical: skills gained, instructor quality, support services, career-readiness, what to change. Volume per cohort is small (30 to 60 responses), which makes the survey-design choice consequential: a 5 percent drop in response rate is one or two participants going unheard.
What breaks. All-open surveys drop completion to 40-50 percent because mobile typing on five prompts is cumulative work. All-closed surveys hold response rate above 70 percent but cannot answer "why did the score drop." Both outcomes hurt the program team's ability to act between cohorts.
What works. A 5-point Likert plus one open prompt per area, with the open prompt placed directly under the Likert. Response rate stays above 70 percent. AI codes the open responses against a fixed six-code scheme that matches the program's theory of change. Counts and themes ship to the program team within 24 hours of the cohort closing.
A workforce-training program running 45 participants per cohort, four cohorts a year. Time-to-themes dropped from two weeks to one day. Program adjustments now happen between cohorts instead of at annual review.
Clinics, federally qualified health centers, or community-health programs surveying patients after a visit or care episode.
Patient feedback surveys are usually short by design (60 to 90 seconds), arrive by SMS or post-visit email, and need to balance closed-ended Press Ganey or CAHPS-style standardized items with the open feedback that catches what those standardized items miss. Volume is large: a mid-size clinic runs 200 to 500 surveys a week.
What breaks. Standardized closed instruments (CAHPS, NPS) carry the dashboard but cannot capture site-specific issues. Open feedback piled at the end of the survey gets ignored: most patients answer the closed items and skip the free-text field. The data team has standardized scores plus a half-empty open column.
What works. Standardized closed items for the dashboard plus one targeted open prompt placed after the lowest-scored area for each respondent. Adaptive: a patient who scored "wait time" lowest sees the open prompt about wait time; a patient who scored "communication" lowest sees that one. Open response rates rise sharply because the prompt is specific to the score the patient gave a moment earlier.
A federally qualified health center running 1,200 patient surveys per month with adaptive open follow-up. Open response rate climbed from 18 percent to 42 percent, and site-specific issues now surface alongside the standardized scores.
Membership organizations, credit unions, B2B SaaS teams, or community-facing programs running quarterly NPS or satisfaction surveys.
Quarterly NPS-style surveys with one closed score (0-10) and one open follow-up ("Why did you give that score?") are the most common shape of paired closed-and-open in the wild. Volume scales: a regional credit union runs 5,000 to 15,000 surveys per quarter, a B2B service team 500 to 5,000.
What breaks. The closed score gets a dashboard. The open responses go to a spreadsheet that nobody reads. Sentiment-only AI scoring gets bolted on and produces a positive-negative-neutral count that obscures more than it reveals. The leadership team debates the score; the actual themes that drove the score never reach the team that owns the issue.
What works. The closed NPS or satisfaction score plus the open follow-up coded against a scheme that names what people actually mention (price, response time, product capability, support quality, branch experience). Themes route to the team that owns each issue. Sentiment becomes one code among several, not the only signal.
A regional credit union running quarterly NPS with 8,000 responses per wave. Theme counts now route to branch managers within 48 hours of survey close, alongside the headline NPS number rather than buried in a spreadsheet attachment.
Every general-purpose survey tool handles both closed-ended and open-ended question forms in collection. The architectural gap is downstream: closed responses go to a dashboard, open responses go to a spreadsheet that someone is supposed to read later. The design intent (closed for the score, open for the why) gets lost between the survey form and the reporting pipeline. Two halves of the same survey live in two different places, and the team that needs both has to assemble them by hand.
Sopact Sense closes the gap by keeping closed counts and open codes in the same workflow. Closed responses count and segment automatically. Open responses get coded against a scheme as they arrive, with AI handling the volume and human review on borderline cases. Both feed the same dashboard, so the closed score and the themes behind it sit next to each other from the moment the survey closes. The fix is not a better text-coding tool bolted onto a survey platform. It is making the two question types' downstream lives identical.
A closed-ended question gives the respondent a fixed set of options to pick from. An open-ended question lets the respondent answer in their own words. The trade is structure for depth. Closed produces data that counts immediately; open produces data that has to be read or coded but captures reasoning and unanticipated themes. The honest survey uses both, paired so the closed answer says what and the open answer says why.
An open-ended question is a survey question that lets the respondent answer in their own words rather than picking from a list. The response is free text, anywhere from one sentence to several paragraphs. Open-ended questions capture reasoning, context, and themes the survey designer did not anticipate. They are slower to analyze than closed-ended questions but they are the only way to hear the response in the respondent's own words.
A closed-ended question gives the respondent a fixed set of response options. Common forms include yes-no questions, multiple choice, Likert scales (1 to 5 or 1 to 7), and ranked options. The data is structured: every response slots into a known bucket, counts roll up immediately, and trends across cohorts compare cleanly. Closed-ended questions are sometimes called fixed-alternative questions in older research literature. Both names mean the same thing.
Closed-ended examples: "How likely are you to recommend this program? 0 to 10." "Did the program help you find a job? Yes / No / Unsure." "Which of the following barriers did you face? Childcare / Transportation / Housing / Food / Healthcare." Open-ended examples: "What was the most useful part of the program?" "What would you change about the program?" "Tell us about a moment in the program that mattered to you." The two types pair well: a closed question delivers the score, the open question that follows it delivers the reason behind the score.
Use an open-ended question when you need the response in the respondent's own words, when you want to surface themes you did not anticipate, or when a closed-ended option list would force a real answer into the wrong bucket. Open-ended questions belong right after a closed-ended score, asking for the reason behind the score. They also belong as a single "anything else we should know" prompt at the end of a survey. They do not belong as the bulk of a survey: response rates drop sharply when respondents face more than two open prompts.
Use a closed-ended question when you need to measure prevalence, compare across cohorts, run a quick mobile survey, or report counts to a funder, regulator, or executive team. Closed-ended questions are the right tool when you already know the response options that matter (yes-no, a defined Likert scale, a known list of barriers). They produce data that ships fast, segments cleanly, and benchmarks against external standards. Most surveys are 80 percent closed-ended for these reasons. The open-ended questions are reserved for the spots where reasoning matters.
Closed-ended questions are fast to answer (one tap on mobile), fast to analyze (counts roll up automatically), comparable across cohorts and time, and benchmarkable against external standards like NPS or industry Likert scales. Response rates are higher because the cognitive cost per question is lower. The data slots into dashboards directly, segmenting by demographic, region, or cohort without any coding step. For surveys that report to funders, regulators, or executive teams on a tight timeline, closed-ended questions are doing most of the work.
Open-ended questions capture the response in the respondent's own words, surface themes the survey designer did not anticipate, and explain the reason behind a closed-ended score. They are how a respondent tells you something you forgot to ask. The data is rich: a single open-ended response often carries more usable signal than ten closed-ended ones. The trade-off is analysis time, which AI scoring against a coding scheme largely closes. Open-ended questions are how surveys produce stories worth quoting alongside numbers worth reporting.
Open-ended questions take more time to answer, which lowers response rate by 5 to 15 percent per open prompt added. They take longer to analyze: a researcher reading 200 free-text responses takes hours; the same volume coded against a scheme by AI takes minutes, but a coding scheme has to be drafted first. Themes from open responses do not benchmark against external data. And response quality varies: some respondents write a paragraph, others write "good." Open-ended questions are useful but expensive, which is why surveys use them sparingly.
Closed-ended questions force every response into one of the predetermined options. Themes the survey designer did not anticipate disappear. The respondent cannot tell the team something they did not think to ask. A score that drops between cohorts is a number going down without a closed-ended question alone telling you why. Closed-ended questions are also vulnerable to response set bias: respondents who read fast can select the same option for every question. The data looks clean and means less than it appears.
A fixed-alternative question is the older academic name for a closed-ended question. The two terms mean the same thing: a survey question with a predetermined set of response options the respondent picks from. The phrase fixed-alternative is more common in research methods textbooks; closed-ended is more common in market research and program evaluation. Examples include yes-no questions, multiple choice, Likert scales, and ranked options. If a research methods text refers to fixed-alternative questions, treat it as identical to a closed-ended question.
Yes, in most cases. The closed-ended question gives you the score, the count, the trend across cohorts. The open-ended question that follows it gives you the reason behind the score. The two together is the survey. The mix matters: roughly 80 percent closed-ended and 20 percent open-ended is a working ratio for most surveys, with one open prompt right after each scored area rather than a wall of opens at the end. Surveys that go all-closed lose the why. Surveys that go all-open lose response rate and the comparable trend across cohorts.
One to three open-ended questions per survey is the working range. One open prompt right after a closed score asks for the reason. A second open prompt at the end captures anything the survey did not ask. A third can deep-dive on one specific theme. Beyond three, response rates drop sharply: respondents see a wall of free-text fields and abandon the survey. Surveys that need more open data should split across waves, not pile opens onto one survey.
Open-ended questions cost more time per question than closed-ended ones. The cost compounds: each additional open prompt drops completion rate by 5 to 15 percent. Mobile respondents are most affected, since typing on a phone keyboard is slower than tapping a Likert scale. The fix is to use open-ended questions sparingly (one to three total) and to place them right after a closed-ended question rather than at the end of the survey. A respondent who has scored a program in the previous field is more likely to explain the score in the next field than to start fresh on a "tell us anything" open prompt at the survey's end.
Six pages that pick up where this one ends. Two sister guides on writing one type well and analyzing it at scale, two pages on the broader survey-typology and analysis questions, and two pages on related survey-design choices that pair with the type decision.
Once the survey calls for an open-ended prompt, the next decision is what makes a usable one. A six-point rubric and worked examples for getting a sentence-or-two response instead of "good" or a paragraph.
Once the open responses arrive, what to do with them. AI scoring against a coding scheme, sampling for human review, and how to keep open-response themes flowing into the same dashboard as the closed counts.
The broader map. Yes-no, multiple choice, Likert, ranked, matrix, semantic differential, open-ended, and the partially-structured forms in between. What each type measures and where it fits.
What happens after the survey closes. Closed counts, open coding, segmentation, trend analysis, and what to ship to whom. The analysis side of the question-type decision plays out here.
The timing companion to the type decision. When two surveys bracket a program, the type choices on one have to match the other for change to be measurable. Same closed scale before and after; same open prompt design.
The wider methodology view. Sample design, question wording, sequencing, scale construction, and how the type decision interacts with the rest of the survey-design system. Useful for teams designing a survey for the first time.
A 60-minute working session. We review the survey draft together, identify the spots where a closed score should pair with an open prompt, and draft one open prompt per area that pulls its weight. You leave with a survey form that scores and explains itself in the same dashboard.
60 minutes, working format. Survey draft on screen. We mark up the closed-and-open pairs together, in real time. No deck.
A survey draft you are about to send or one you ran last cycle and want to fix. Even a list of question topics is enough to start.
Each scored area paired with a sharpened open prompt. Plus a coding scheme draft for the open responses, ready to apply at survey close.
