How we turn publishing data into useful research

The aim is to help readers make a better decision. A clear definition and an honest limit are part of every finding.

Who this publication is for

Postly Research is free public education for marketers, businesses, creators, researchers and anyone learning how social media works. Readers do not need a Postly account. Postly funds and publishes the work; its customer sample is not a random sample of all social media users.

Three kinds of evidence

Measured studies use anonymized publishing and compatible performance records. Platform explainers cite official documentation for platform mechanics. Playbooks contain clearly identified editorial recommendations and experiments. A recommendation is not presented as a measured finding, and correlations do not establish algorithm rules or causation.

Who the research represents

Studies describe eligible observations available to Postly. They do not represent every account on a social network. Analytics samples depend on provider permissions, supported metrics, and which accounts have refreshed their reporting. Each published table identifies the number of sampled accounts and observations. A retained-data census excludes deleted and expired records. Historical totals are distinct from a study sample; we never claim a million-post sample without verifying it.

Data collection and source statements

Historical research reads use bounded batches and field projections. The initial census reads a database secondary and makes no provider calls or publishing changes. Source statements distinguish inspected documents, eligible posts and successful destination deliveries. A cross-post is not several independent creative assets. Categories require sufficient independent accounts before publication; product or industry classifications require validation before supporting sector rankings.

Publishing activity

A delivery is a verified successful publication to a destination. A cross-post sent to several destinations can contribute several deliveries. Retries with the same destination receipt are deduplicated. Failed attempts, draft content, and unconfirmed uploads do not count. Format labels describe attached media where recorded; they do not infer a native format such as Reels from a video file alone.

Performance and engagement

We keep platforms and metric definitions separate. Interaction benchmarks use recorded interaction counters, not views renamed as engagement. Click-through rate requires clicks and impressions for the same destination. Missing counters remain unavailable. Recorded lifetime counters describe accumulated activity at the observation time, not necessarily activity generated within the report month. Mixed post ages limit comparisons.

Growth and matched observations

Audience growth requires compatible snapshots of the same account. Content momentum requires compatible counters for the same post destination. Reports state the comparison basis; a missing baseline is not zero. A decline may reflect audience loss or provider corrections. Changes in content performance do not establish follower growth or causation.

Sample thresholds and privacy

Published rows require at least 30 distinct accounts and 100 observations. Smaller groups are withheld. Public reports and PDF downloads contain aggregate results, never customer names, account identifiers, captions, private messages, or access credentials. Internal matching keys are excluded from the publication schema.

Monthly editions

The reporting period is a completed calendar month in UTC. The monthly job becomes due on the first day of the following month and retries if preparation fails. A new edition appears only after its source checks complete; changing the date alone never refreshes a finding. Archived editions retain their reporting month and generation timestamp. Browse the edition archive to read past findings and download either the complete edition or an individual report. The original retained-data census is identified separately from monthly samples. If a refresh fails, the last valid edition stays available.

Reading unavailable results

“Awaiting eligible data” means the report has not met its data requirements. It does not mean activity was zero. A report may be available for some platforms while others lack compatible metrics, sufficient samples, or a previous snapshot.

Citing a report

Use the report title, Postly Research, reporting month, and the edition URL. Include the sample and metric definition when quoting a finding. Keep your interpretation separate from the measured result.

Use the research citation template or read the performance measurement guide.