Metric meaning across platforms
Compare quantities only when units, denominators and source meaning match. A shared envelope does not make every platform metric the same measurement.
Fields, sources and denominators
| Endpoint / field | Meaning | Comparison rule |
|---|---|---|
| reddit/post · score | Net score, which can be negative. | Not likes or exact up/down vote totals. |
| github/repo · watchers | Repository subscribers_count. | List rows can lack it; null is not zero. |
| facebook/profile · followers / likes_count | Separate page metrics; followers can be rounded. | Do not substitute page likes for followers. |
| google_play/app-info · installs / min_installs | Display range and its lower bound. | The lower bound is not an exact install count. |
| google_play/app-info · score | Average rating on a five-star scale. | Ratings count differs from written reviews count. |
| tiktok/user-audience · returned_count_share_percentage | Share of returned country counts only. | Not a verified whole-audience share; methodology and sampled_at can be unknown. |
| tiktok/post · region / author.region | Source-reported content/account region. | Not residence or audience geography. |
| instagram/engagement · engagement_rate | Percentage over the documented sample and follower denominator. | Do not compare directly to a views-based ratio without checking units and denominators. |
| Lists · count / total / cursor | Returned count, source total or continuation, depending on operation. | A missing cursor does not prove complete history. |
Absent, null, zero and estimated values
Absent or null means unknown, not zero, false or an empty string. Published schemas list optional shapes. Missing data can reflect visibility, list/detail differences, unsupported metrics, failed components or source changes. Only call a null structural when the endpoint documents that rule.
Preserve currency with money, rating scales with ratings, and source timestamps separately from retrieval time. Keep exact IDs in the documented representation. Location text, account flags and model labels are not verified personal attributes.
Make comparisons reproducible
Select the exact endpoint and field for each source. Save source IDs/URLs, request_id, retrieval time, source time when supplied, value, unit and missing-value reason. Check sample bounds, rounding, partial flags and denominator before computing comparisons.
If a list lacks a required field, inspect the detail operation and its price before enrichment. Do not assume an include flag or free enrichment exists everywhere. Report incompatible metrics separately and retain unknown/unavailable rows instead of substituting zero.