Market Research Journal: 5 Helpful Steps and a Template

CrowdSent Blog Team · Research skills

A market research journal records what you asked, which evidence you checked and how you reached a conclusion. Keep the original entry, then add dated reviews so you can distinguish what you knew at the time from what you learned later.

This guide gives you a copyable template and a fictional completed entry. You can use it in a document or spreadsheet beside your market research workspace. The fields are designed for questions about news, discussion and sentiment; they do not require an order, position or trading decision.

If you arrived looking for a trading journal template, decide which record you need. An execution log records orders, fills, fees and positions. A research entry records the evidence behind an interpretation. Keeping those purposes clear makes it easier to review your reasoning even when you took no action.

1. Define a question your market research journal can answer

Start with a question narrow enough to investigate. “What is happening in crypto?” could produce an endless collection of unrelated links. “Did the language in this selected Bitcoin discussion become more positive after a particular announcement?” tells you what material and comparison you need.

Write the market or pair, the subject and the comparison window before collecting examples. A question about language is different from a question about factual accuracy. A question about the people in one discussion is different from a question about the wider market. An entry may lead to another question, but keep each conclusion attached to its original scope.

Choose the comparison before seeing the result

For a before-and-after comparison, record both start and end times, their timezone and the rule for including an item at a boundary. Keep the source selection and counting unit consistent where possible. If the selection changes, explain the change rather than treating the two windows as interchangeable.

Suppose your first window includes all visible comments, while your second includes only comments with external links. You could learn something about linked comments, but a change in the summary might reflect the filter. The journal should make that difference easy to spot before you attribute the change to an announcement.

Our guide to reading historical sentiment covers window, timestamp and sample-composition checks in more detail. In the journal, record the choices you actually used. Do not copy a preferred method into the entry if your evidence was collected differently.

Give the question a stopping condition

Name the evidence that would make the entry ready for review: perhaps the original release, a defined collection of comments and a count of repeated material. This helps prevent an open-ended search for one more supporting headline. A useful stopping condition can also be a clearly documented access limit.

2. Capture the sources and the observations

In your market research journal, use a source record for each item that matters to the conclusion. Keep the original URL, publisher, document title or identifier, relevant publication time and your access time. For a longer document, add the page, paragraph, table or recording timestamp supporting your note.

Record what the source shows before explaining it. Mark quotations clearly and keep them short; distinguish them from paraphrases and your own calculations. If you calculated a percentage, retain the underlying counts, the denominator and the formula in ordinary words. This lets another reader check the calculation without reconstructing your entire browsing session.

Keep different times in different fields

• Observation period: the interval your evidence describes.
• Publication time: when the source says the information was released.
• Access time: when you viewed the particular version.
• Entry time: when you wrote your research note.

Those fields can legitimately differ. A monthly statistic might be released weeks after the month ends. A discussion screenshot might be collected later than the messages it shows. If a precise publication time is unavailable, say so; your own access time does not establish the first moment the information was public.

Preserve the version you actually used

Dynamic pages, filters and revised statistics can make a saved link incomplete. Where permitted, retain a screenshot or export with the visible range and filters. Keep it beside the source URL and the access date. Avoid storing other people’s private details when a short, anonymised description answers your research question.

The St. Louis Fed’s ALFRED service illustrates why versions matter: it preserves vintages of economic data reported on past dates. For a historical question, a later estimate can differ from the one available then. Check whether your own source offers an appropriate earlier version rather than assuming it does.

For an important headline, use the steps in our market news verification guide before labelling its underlying claim as established. You can still save an unchecked story as a lead. Put its verification status next to the observation so it cannot quietly become a confirmed fact in a later summary.

3. Separate your explanation from the evidence

Give your market research journal a separate field for your working explanation. Use wording specific enough to challenge: “The selected later comments contain more favorable language” is easier to examine than “Sentiment is strong.” Add the limits of the collection before extending the claim to a broader population or a cause.

Then write a plausible alternative explanation. Perhaps the later window contains a different group of contributors, repeated copies of one headline or a different topic. The alternative should fit the observed evidence; it need not be a claim that the alternative actually happened.

Name evidence that could change your view

Choose a concrete check that would weaken your current interpretation. If you think the apparent shift reflects several distinct contributors, count contributors where you can do so appropriately. If you think a claim came from an official announcement, read that announcement. “Stay open-minded” is an intention; a named check turns it into a reviewable action.

Keep your confidence description attached to one claim. You might have high confidence that a source displayed a particular value when you checked it, but low confidence that the source selection represents a larger discussion. One overall confidence label can hide that difference.

You do not need a numerical probability for every entry. A label such as “limited: original source checked, sample coverage unknown” can be more informative than an unsupported percentage. Our explanation of sentiment score meaning also separates classification confidence from the quality of the underlying evidence.

4. Preserve a dated review instead of rewriting the first entry

Preserve the initial market research journal entry and append a review when material new evidence arrives. State what changed, which source supports the change and whether it affects the observation, the explanation or both. If you made an error, identify it plainly and preserve enough context to understand the correction.

A source revision and your own mistake are different events. The Bureau of Labor Statistics explains that statistical revisions can result from additional data, seasonal adjustment and benchmarking. Your first note may accurately describe the initial release even when the producer later revises it. Record both versions when they matter to the question.

Review process and result separately

Ask whether your sources supported the conclusion at the time you wrote it. Then ask what later evidence changed. A subsequent price movement alone cannot tell you whether the explanation in a research note was well supported. Many different explanations can be compatible with the same movement.

For a sentiment observation, review whether you understood the measure, matched the comparison windows and retained the sample limits. For a news claim, review the original source and the version you used. This creates a more useful learning record than marking every entry simply “right” or “wrong.”

Record an unresolved ending honestly

Sometimes the needed source is unavailable or the comparison cannot be reconstructed. Mark the entry unresolved and name the missing evidence. You can close the current investigation without pretending the question has been answered. If new information later appears, append a new dated review.

5. Build a repeatable routine with this journal template

Use one market research journal entry per main question. Start with the short version below, then expand only the fields that affect your conclusion. A lightweight research journal you can maintain is more useful than a large form filled with vague or copied text.

Copyable market research journal template

• Entry ID and recorded at: [Unique label; date, time and timezone.]
• Market and question: [Asset or pair; one specific question.]
• Observation window: [Start, end, comparison period and boundary rule.]
• Selection: [Sources, filters, unit counted and known exclusions.]
• Source record: [Original URL, publisher, title or identifier, publication and access times.]
• Observed information: [What the evidence shows; short attributed quote or clearly marked paraphrase.]
• Calculations: [Inputs, denominator and method, or “none.”]
• Coverage and limitations: [Missing, delayed, repeated or unavailable material.]
• Working explanation: [A tentative interpretation within the observed scope.]
• Alternative explanation: [Another plausible account of the same observation.]
• Evidence that would change my view: [A specific contradiction or missing check.]
• Confidence and reason: [Claim-specific assessment and its basis.]
• Next check: [What to inspect and what would trigger another review.]
• Dated review: [New evidence, corrections and current status; preserve the initial entry.]

A fictional completed entry

Every detail in this example is invented. The counts and times illustrate the journal method; they are not real EUR/USD observations or CrowdSent measurements.

• ID: DEMO-01, recorded 14 September 2025 at 14:40 UTC.
• Question: Did language in a selected EUR/USD discussion become more favorable after a fictional 14:00 UTC announcement?
• Windows: 13:30 up to 14:00, then 14:00 up to 14:30 UTC. A comment exactly at 14:00 belongs to the second window.
• Selection: The same fictional discussion view and language filter; one comment is one observation.
• Observation: The first window contains 12 comments, 4 manually described as favorable. The later window contains 20 comments, 12 described as favorable.
• Counting note: Favorable comments represent 4 of 12 and 12 of 20 observations. The categories are the fictional researcher’s descriptions, not a model output.
• Source status: The announcement itself remains unchecked. Six favorable later comments repeat the same headline. Contributor independence has not been established.
• Working explanation: Favorable language occupies a larger share of the selected later comments.
• Alternative: Repeated coverage or a change in participating contributors could account for part of the difference.
• Confidence: Limited. Counts are recorded, but the classification, source chain and sample composition require review.
• Next check: Read the original announcement and identify exact repetitions without treating them as independent supporting evidence.

The entry supports a narrow description of the selected comments. It does not establish that the announcement caused the difference, that all participants changed their views or that a price movement will follow.

An appended fictional review

At 15:10 UTC, the researcher adds a review: “The original release describes a proposal rather than a completed decision. Six repeated headlines omitted that qualification. The initial counts remain a record of the comments collected, but their favorable tone does not establish the claim in the repeated headline. A broader sentiment conclusion remains unresolved.”

Notice what stayed intact: the original question, counts and access limits. The update adds source verification and narrows the interpretation. It does not rewrite the earlier entry to make the researcher appear to have known the qualification from the start.

Keep the routine proportionate

For an ordinary observation, a few precise sentences may cover the essential fields. Spend more attention on a claim that anchors a substantial conclusion or will be shared with others. Reuse source identifiers and link related entries so the same announcement does not become several apparently independent pieces of evidence.

When reviewing your market research journal, look for recurring process gaps: missing timezones, unknown denominators, unverified sources or conclusions broader than the sample. Choose one specific improvement for the next entries. A journal becomes useful through these corrections, rather than through the number of pages it contains.

Source documentation checked 6 October 2026. The template and fictional example are educational material from the CrowdSent Blog Team, not investment recommendations or performance evidence.