Standards & Methodology

Don’t just trust our numbers. Check how we got them.

A poll’s credibility is built on transparency, not claims. Every Markdata release explains how its numbers were obtained, so the public, the media, and other researchers can check for themselves.

Official PERSEPI member

Methodology disclosed in every release

Full reports free to download

Scopus-indexed academic publications

Our transparency commitment: six things we always disclose

In line with the PERSEPI code of ethics, every survey result we release to the public includes the following.

  1. 01

    Sampling method

    How respondents were selected, for example multistage random sampling, and why.

  2. 02

    Sample size & margin of error

    How many respondents, the margin of error, and at what confidence level.

  3. 03

    Fieldwork dates & coverage

    When interviews started and ended, and which areas were covered.

  4. 04

    Interview mode

    Face-to-face, phone, or online, because mode affects results.

  5. 05

    Question wording

    Key questions are shown verbatim, so numbers can be read in context.

  6. 06

    Funding source

    Who paid for the study, including when Markdata funded it itself.

Our public study reports can be downloaded in full, not just as summaries. Researchers and journalists can request raw data or additional tables.

Browse our reports

From question to report

The six stages we follow in every survey.

  1. 1

    Defining the question

    We start from the decision the client needs to make, then turn it into questions that can be measured.

  2. 2

    Sample design

    Probability samples based on official population data, sized to the precision required.

  3. 3

    Questionnaire & pilot

    Questions are written neutrally, piloted, and revised before fieldwork.

  4. 4

    Interviewer training

    Interviewers are trained on the questionnaire, interview ethics, and respondent selection.

  5. 5

    Quality control

    A random share of interviews is re-verified; data that fails the standard is removed.

  6. 6

    Analysis & reporting

    Results are weighted to population characteristics and reported together with their limitations.

One question, many ways to answer it

We choose the method to fit the question, and often combine several. Our studies with Yayasan Indonesia CERAH, for example, combined surveys, stakeholder interviews, and big-data analysis.

  • Face-to-face surveys

    For national and regional surveys that must reach every segment of the population.

  • Phone surveys

    Fast for issue tracking, with the coverage caveats disclosed.

  • Online surveys

    For specific groups; never presented as a picture of the general population.

  • Focus groups & in-depth interviews

    Uncovering the reasons behind the numbers, including stakeholder interviews.

  • Big data & social media

    Reading public conversation and public figures’ records from open sources.

  • Spatial analysis (GIS)

    Mapping results down to local areas for planning and field strategy.

  • Poll-of-polls aggregation

    Filtering and averaging many pollsters’ releases using the same criteria.

How to read a poll

A short guide for journalists and readers, including when reading other pollsters’ surveys.

A gap within the margin of error is not a gap

If two numbers differ by less than the margin of error, they are statistically tied.

Random samples differ from open online polls

Only probability samples can be generalised to the whole population.

Compare trends within one pollster

Every pollster has a house effect. Change between waves means more than a single number.

Read the question wording

“Satisfied with the President” and “satisfied with the government” can produce different numbers.

Example

In our Two Years of Prabowo-Gibran study, we screened 57 poll releases, used only random-sample surveys for the averages, and explained why pollsters’ results could differ by 44 points.

See that study’s methodology
PERSEPI

Official PERSEPI member

Markdata is a registered member of the Indonesian Public Opinion Survey Association and is bound by its code of ethics and the oversight of its Ethics Council.

persepi.org

Clients do not change findings

Clients set the research question, not the answer. Results commissioned for internal use are never released without consent, and results we publish are never altered to suit anyone.

Rooted in academic research

Markdata’s founder publishes in Scopus-indexed journals and academic books, including Routledge and Nova Science titles. We bring the same research standards to every survey.

View publications

Questions about our methods?

We are happy to explain our survey design, our data, or how we analyse it.