The method
belongs in the story.
A credible finding should come with enough information to understand what was measured, how it was measured and where its limits lie.
Design around the decision.
We start by defining the question and the population whose views matter to it. The method follows: recruitment, interview mode, questionnaire, sample plan and intended analysis should work together.
A large response count is useful only within a suitable design. Coverage, selection and question quality require attention alongside sample size. Weighting can address selected imbalances; it does not eliminate every source of bias.
What an original public release should disclose.
| Disclosure | What readers should be able to find |
|---|---|
| Commissioning | Sponsor or client, research provider and publication purpose. |
| Population & timing | Who the study represents and when interviews took place. |
| Recruitment & mode | How people were selected, contacted and interviewed. |
| Sample & weighting | Total and subgroup bases, weighting variables and relevant design effects. |
| Questionnaire | Exact wording, answer options and relevant preceding questions. |
| Uncertainty | A design-appropriate explanation, including limitations. |
| Results | Clearly labelled denominators, rounding and treatment of undecided answers. |
A note on margins of error.
Conventional sampling margins of error apply to probability sampling under specified assumptions. We do not assign a conventional sampling margin of error to an opt-in non-probability sample. Any model-based uncertainty estimate would need its own explanation and assumptions.
For a simple random sample, a 50% proportion and a 95% confidence level, ignoring finite-population correction and design effects:
Formula: 1.96 × √(0.25 / n). Sampling error is only one source of error. This illustration does not describe a Northbound study or validate a non-probability sample.
Method reference: Statistics Canada: Questions to ask about surveys ↗
Keep evidence types distinct.
Original research means interviews or data collection conducted or commissioned by Northbound. Northbound’s polling reports present supplied study figures and methods, with available disclosures and missing details identified on each page. The separate insights journal contains secondary analysis and guides.
Secondary analysis interprets an existing source, such as official election results, Statistics Canada data or a Bank of Canada publication. We identify the source and do not present its work as our own.
Perspectives and guides explain a research question, practice or interpretation. Hypothetical examples are identified and are not presented as measured findings.
Publication dates and the archive.
The journal’s initial articles were published on 25 October 2026. Historical period labels refer to the event or data covered. Retrospectives are labelled on the archive and article pages, and machine-readable publication dates reflect the actual publication date. A retrospective does not establish that Northbound operated or published at that earlier time.
Editorial standards.
Write descriptive headlines. Link the primary evidence. Distinguish correlation from causation, and intentions from behaviour. Avoid claims about a subgroup when the evidence does not support them.
Commissioned public work should identify its sponsor. Potential conflicts relevant to interpretation should be disclosed. Client work should not be presented as independently commissioned journalism.
Corrections and reuse.
Material corrections should be recorded with the change date and an explanation. No correction history is claimed for the initial journal. When citing a Northbound article, include its title, actual publication date and link. Third-party data remains attributed to its original provider.
To prepare a source query or correction note, use the research brief builder. It creates a local document; direct message delivery is not yet connected.
What do you need to know?
Start with the question. We’ll help shape the research.