Which research sources should I trust most?

Tier 1 primary research firms — Gartner, IDC, Forrester, McKinsey, Bain, BCG, WARC, IAB — for category sizes and forecasts, because they collect primary data at scale with documented methodology. Trust them for reliability, but still check that their market definition matches your claim.

How do I weight analyst versus aggregator data?

Analyst (Tier 1) figures can be cited directly; aggregator (Tier 3) figures are leads to be traced, not citations. On category definitions Tier 1 wins; on observed behaviour inside a niche, a Tier 2 source close to the raw event — deal data, traffic panels, review platforms — can outrank a Tier 1 estimate.

What is a source credibility tier?

A ranking of how close a source is to the raw evidence and therefore how much verification a figure needs before use: Tier 1 conducts primary research, Tier 2 holds specialist or transactional data from its market position, Tier 3 compiles and republishes others' numbers.

How do I trace a Statista or aggregator statistic to its original source?

Find the attribution line on the chart or page, search the exact figure with the attributed firm and year, then locate the original publication in that firm's own newsroom. Cite the primary with its original date. If the trail only loops through other aggregators, treat the figure as unverified.

How to tier your research sources

Guide · Frameworks · 5 min read · last verified 2026-07-27

Reviewed before publication Editorial board Independent commercial review

In shortA three-tier system for research sources — Tier 1 primary analysts, Tier 2 specialist data, Tier 3 aggregators — plus the method for tracing any statistic to its primary.

A source tier is a credibility ranking that tells you how much weight a statistic can carry before you need to verify it further. Tier 1 sources conduct primary research and can be cited directly; Tier 2 sources hold specialist or transactional data that is strong within its niche; Tier 3 aggregators repackage other people's numbers and are useful for discovery but should never be your final citation. Tiering matters because every claim you build inherits the credibility of its weakest source.

Why tier at all

Research without a tiering discipline treats all statistics as equal, which means your market study is only as strong as the flimsiest blog post that slipped into it. A tier system does two jobs. It sets a verification budget — Tier 1 figures you can cite and move on, Tier 3 figures you must trace before use. And it resolves conflicts — when two sources disagree, the tiers tell you which one wins by default and when the default should be overridden.

The tiers below reflect how the source produces its numbers, not how famous it is. The question is always: how close is this organisation to the raw evidence?

Tier 1: primary analyst and consulting research

Tier 1 is research organisations that collect primary data at scale and stake their reputation on the results: Gartner, IDC, and Forrester on the analyst side; McKinsey, Bain, and BCG on the consulting side; bodies like WARC and the IAB for advertising and media measurement.

What earns the tier is the machinery behind the number — documented methodology, large proprietary datasets or survey panels, analyst review, and a commercial incentive to be right over many years rather than to be dramatic once. When a Tier 1 firm publishes a category forecast, you can cite it directly and your reader's remaining question is about relevance, not reliability.

Two caveats keep Tier 1 honest. First, the detail is usually paywalled, so what circulates publicly is the press-release headline — a real figure stripped of its definition, which is how Tier 1 numbers get misused. Second, Tier 1 firms disagree with each other, because they define markets differently. The tier tells you the number is professionally produced, not that it is the only defensible number.

Tier 2: specialist and transactional data

Tier 2 is organisations whose data comes from a position they occupy rather than a survey they ran: venture firms like Menlo Ventures publishing enterprise AI adoption research from their vantage point in the market; CB Insights and PitchBook aggregating deal and funding data; Similarweb observing web traffic; G2 sitting on review and category data generated by real buyers.

The strength of Tier 2 is proximity to the event. PitchBook's deal data is not an estimate of funding activity — it is a record of it, with known gaps. Similarweb's traffic panels observe behaviour rather than asking about it. Within their niche, these sources can outrank Tier 1 estimates precisely because they are closer to the raw signal.

The limit of Tier 2 is scope. A venture firm's adoption survey reflects its network; review-site category data reflects who bothers to review. Use Tier 2 as strong evidence inside its lane, and be explicit about where the lane ends.

Tier 3: aggregators — discovery, not citation

Tier 3 is the aggregator layer: Statista, Grand View Research, Precedence Research, and similar houses that compile, model, and republish figures across thousands of markets. They are genuinely useful — for discovering that estimates exist, mapping the range of published sizes, and getting a category taxonomy quickly. Magrios uses this layer the same way in its own research pipeline: as a directory of leads, with authority weighted towards wherever the trail ends.

The risk is that aggregator figures often arrive without visible methodology, and sometimes without a visible original source. Aggregators cite each other, which creates circular sourcing: a number bounces between compilation sites until it looks independently confirmed. Citing an aggregator as your final source tells a careful reader that your trail ends one click in.

How to trace an aggregator statistic to its primary source

This is the core practical skill, and it takes minutes once it is a habit.

Look for the source line. Aggregator pages and charts usually carry a small attribution — a firm name, a survey, a year. That name is your lead.
Search the exact figure in quotation marks together with the attributed firm and year. Primary publications, press releases, or coverage of the original report usually surface quickly.
Go to the named firm's own newsroom or publications page and find the original release. Read the definition and date, which the aggregator will have stripped.
If the trail leads to another aggregator, keep following. If it ends in a loop of compilation sites with no primary publication anywhere, treat the figure as unverified.
Cite the primary, not the place you found it — with the original date, since aggregators frequently republish old numbers under new dates.
If the figure is genuinely untraceable but you still need it, label it honestly: an unverified aggregator estimate. Sometimes that is acceptable; presenting it as confirmed never is.

Weighting rules when sources conflict

A few defaults cover most collisions. On market definitions and category sizes, Tier 1 beats Tier 2, and both beat Tier 3. On observed behaviour inside a niche — traffic, deals, reviews — a Tier 2 source close to the event can outrank a Tier 1 estimate of the same thing. Recency matters within a tier: a two-year-old Tier 1 forecast of a fast-moving category can be weaker than a current Tier 2 reading. And independent agreement compounds: two genuinely independent Tier 2 sources pointing the same way often carry more weight than one stale Tier 1 figure — provided you checked that they are independent and not both citing the same origin.

Tiers and AI answers

One more reason to care: AI assistants synthesise across all three tiers without labelling them, so an untraceable aggregator figure and a Gartner forecast can arrive in the same generated answer with equal confidence. You cannot control that mix, but you can control your side of it — content built on traceable primary sources is the kind of evidence answer engines tend to quote, and it survives the spot-check when a buyer follows the citation. Tiering your inputs is quietly becoming part of how your outputs get discovered.