How our audits work

Measured data and missing values

Connected professional accounts may supply authorized Graph API insights. Public audits use the public data available from our provider. Samples are limited, may be delayed, and are not necessarily representative of every post. We analyze the latest fetched feed sample, up to 30 posts, and up to 30 deduplicated Reels. These are sample limits, not a claim that the sample covers the last 30 days.

Missing, hidden or unsupported counters are unavailable, not zero. A measured zero remains zero. Averages exclude unavailable samples. Reach, views, plays, likes, shares and saves are distinct metrics; we do not substitute one for another.

Engagement and Reels

Visible engagement rate = (likes + comments) ÷ current account followers × 100. Both interaction counters and a positive follower count are required. Post audits reuse the fetched account follower count when the post response omits it. We never use one as a replacement denominator. This is not engagement divided by reach, and current followers may differ from followers when the post was published.

Reels view rate uses actual views ÷ followers × 100; play rate uses actual plays ÷ followers × 100 and is labeled separately. Plays can include repeat plays. Likes plus comments are reported only as visible engagement, never as views. Without usable counters and followers, no Reels performance score is assigned. Neither view nor play counts establish unique reach, retention, or hook quality.

The directional Reels score is the average views-per-follower percentage multiplied by two, capped at 100. If only plays are available, that ratio is used and identified as the basis. This scaling is a product heuristic, not a calibrated benchmark.

Profile scores and overall score

Profile strength, bio, consistency, engagement, content mix and hashtag judgments use AI analysis of supplied observations. Scores can vary between runs and model versions. They are labeled estimates. Missing-data categories are excluded. We do not score audience authenticity or quality from follower totals, biography keywords or a follower/following ratio.

Overall score = sum of available category scores × their weights, divided by the sum of the available weights, rounded to the nearest integer.

• Basic: profile strength 60%, bio 40%.
• Full: profile strength 20%, visible engagement 25%, content mix 20%, consistency 20%, hashtags 15%. Bio is shown separately.
• Advanced: profile strength 15%, bio 10%, consistency 10%, engagement 15%, content mix 10%, hashtags 10%, Reels 15%, audience 15%. Audience currently remains unavailable and its weight is excluded.

Recommendations and sentiment

Content mix means the frequency of each format in the sample, not its performance. Posting frequency is estimated from valid timestamp intervals. Advice and seven-day plans are hypotheses to test against your own results. No fixed reach multiplier, shadowban diagnosis, optimal schedule, or growth guarantee is supported by these observations.

Comment sentiment is a simple English keyword estimate. Each sampled comment belongs to exactly one category: positive, negative, neutral or unknown. Mixed positive/negative wording, negation and unsupported text are unknown. With no usable comments, percentages are unavailable. Percentages include unknown comments and use rounding that totals 100%. This method cannot reliably interpret sarcasm, context or every language.

Post caption scores are heuristics: hook scoring starts at 50 and adds points for first-line length, punctuation and opening patterns; caption structure starts at 50 and adds points for length, a detected call to action and line breaks. Hashtag structure scores count and variety, not discovery reach or semantic fit. These are capped at 100 and are not measured outcomes.

Limits and audit credits

One completed profile or post audit uses one credit. Comparisons use one credit per successfully analyzed profile, including the baseline. Credits are reserved before processing so simultaneous requests cannot spend the same allowance. Failed requests release unused reservations where safely confirmed. An interrupted server request may require support review before its credit can be restored.

Older saved audits are snapshots and may use an earlier methodology. Run a new audit for the corrected calculations. Reports display unavailable data explicitly; provider access and permissions determine which observations are possible.