# Survey tips from 15 years of fieldwork

Short survey tips posted from 2012 to 2014. Balcostics Research was incorporated in 2011.

## Open or closed questions

An open question leaves the wording to the person answering. They do not pick "yes" or "no", and they do not pick a score. A closed question gives a list, and the answer has to come from that list. The February 2014 note compared a closed question to a multiple-choice item on a school paper, and an open question to a short written answer.

Use a closed list when you already know the answers you need to count. Use an open question when you do not yet know the words people will use. If the list may be missing something, add a write-in line. The BlackBerry download under that tip is a dead link. BlackBerry World shut in January 2022.

First posted 11 February 2014.

## You can field your own survey

You can run your own survey if you set aside time to prepare. Write the purpose, the questions the study must answer, and how you will collect the data. The February 2014 note listed a web page, email, face-to-face, post and telephone. The choice depends on the budget, who you need to reach, the sample size and the deadline. "Vary base on" in the original means the approach varies based on the sample size.

Face-to-face work in Jamaica is often a tablet. STATIN's director of field services said in September 2026 that interviewers had not used paper for about ten years. They use tablets and Survey Solutions ([JIS, 21 September 2026](https://jis.gov.jm/statin-assures-strict-confidentiality-for-2026-jslc/)). The 2022 census was the first Jamaican census taken on tablets ([JIS, 17 October 2025](https://jis.gov.jm/statin-begins-planning-for-jamaicas-16th-national-census-in-2031/)).

Collect the file first, then analyse it.

First posted 15 February 2014.

## Customers you already have

A survey of people who already buy from you can cover who they are, whether the product satisfies them, and what they want changed. The February 2014 note named email addresses you hold, a feedback link on the site, and a face-to-face conversation.

Read the result as news about those customers, or about the people who chose to write in. A figure for the whole market needs a sample of that market. An open link does not give each person in the market a known chance of being asked. [AAPOR](https://aapor.org/journalist-guide-to-understanding-polls-surveys/) calls that a non-probability sample, and a probability sample the case where that chance is known.

First posted 17 February 2014.

## Design the survey before fieldwork

The February 2013 note said a survey works when the design is thought through, in the same way a product does, and it pointed readers to a slideshow. The slide text (SlideShare embed 11788937) was not on the page we recovered.

Write down the decision the survey has to serve before you write the items. Name who should answer. Drop any question that will not change that decision. Try the questionnaire on a few people who resemble the real respondents, and fix what they stumble on, before the main round starts.

First posted 27 February 2013.

## What a questionnaire has to do

The November 2012 note opened with a line it credited to "Pubius Syrus": "It is not every question that deserves an answer." [Bartlett's Familiar Quotations](https://www.bartleby.com/lit-hub/familiar-quotations/8654-publius-syrus-42-bc-john-bartlett/) lists that sentence as Maxim 581 of Publius Syrus. The usual scholarly name is Publilius Syrus. The spelling in the note was wrong.

The note said its slides would cover which questions to ask, how to phrase them, how to order the questionnaire, and a pretest. Those slides (embed 1305056) did not come back. On those four jobs, ask what the study needs, put one idea in each question, place easier items before sensitive ones, and pretest the form before fieldwork.

First posted 19 November 2012.

## Choose the approach, then the sample

The March 2013 note called research a systematic way to collect, enter and analyse data. It then said collection starts by picking a sampling method, and it offered "qualitative or quantitative" as those methods. Qualitative and quantitative are ways of studying a question. The sample is the rule for who is included.

[AAPOR](https://aapor.org/journalist-guide-to-understanding-polls-surveys/) separates a probability sample, where each person in the target population has a known chance of selection, from a non-probability sample, where that chance is not known. A percentage that is meant to stand for a parish, or for Jamaica, needs the first kind. Otherwise label it as the people who took part. The sampling slideshow (embed 703965) was not recovered.

First posted 7 March 2013.

## Research starts with a question

The November 2012 note asked four questions and left the answers on a slideshow (embed 885335). The slide text was not recovered. The questions were what research is, what the process is, why do it, and why review the literature.

Research is the plan for answering a stated question. The process runs from that question to a design, then to collection, analysis and a decision. You do it when a decision needs evidence you do not already hold. A literature review shows what is already known, so the new fieldwork aims at the gap.

First posted 12 November 2012.

## Save the code that made the table

The June 2013 note said a colleague should be able to read your code and follow each step that produced the result. SPSS, RATS, Stata and R all let an analyst save syntax and insert comments. That advice still holds.

The syntax file is the start. [The Turing Way](https://www.turing.ac.uk/research/research-projects/turing-way), the Alan Turing Institute handbook, describes reproducible work as sharing the data and the code that generated the published result. Keep the data file the syntax ran on, and write down the software and the version. A comment that says why a recode was done will stop the next analyst from guessing.

First posted 8 June 2013.

## Start online research from the question

The November 2012 note said most people start online, looking for literature on the problem, and then posted a third-party chart, `how-to-research.jpg` (936 by 2160). The chart is titled "How to research" and is marked "Curated by @JulianaPayson" and "Brought to you by InMotion Hosting". It is a map of search, traffic and keyword tools, including Google+, Alexa, Yahoo Clues and Compete.

Google brought the close of consumer Google+ forward to April 2019 ([Google, 10 December 2018](https://blog.google/innovation-and-ai/technology/safety-security/expediting-changes-google-plus/)). Amazon discontinued Alexa Internet, the traffic-ranking service, on 1 May 2022 ([Alexa Internet](https://en.wikipedia.org/wiki/Alexa_Internet)). Use the research question to choose the literature. A poster of 2012 ranking sites is not a survey method.

First posted 15 November 2012.

## Updated October 2026

- Nine short notes are one post. SlideShare text for embeds 885335, 1305056, 11788937 and 703965 was not in the recovered pages, so those sections follow the prose and the headings on the page.
- "Pubius Syrus" is corrected to Publilius Syrus (Bartlett prints Publius Syrus), Maxim 581.
- The March 2013 line that treated qualitative and quantitative as sampling methods is rewritten. The distinction used is AAPOR's probability and non-probability samples.
- The BlackBerry App World link is removed. BlackBerry World shut in January 2022.
- Face-to-face collection is updated with STATIN's tablet practice (JIS, 21 September 2026) and the 2022 census on tablets (JIS, 17 October 2025).
- Saved syntax is kept, and tied to sharing the data and the code (The Turing Way).
- The 2012 tool chart is kept as a description of that image. Google+ and Alexa Internet are cited as closed. Yahoo Clues and Compete are named because they are printed on the chart. Their closure dates were not checked for this post.


Source: https://balcostics.com/blog/survey-tips-from-our-archive/