The short answer
Preserve both observations and first establish whether they measure the same thing. Match account and post, metric definition, population, time basis, observation timestamp and aggregation. Resolve a documented scope difference explicitly; keep an unexplained discrepancy open. Averaging two incompatible totals does not make either one correct.
What you’ll take away
- An identical label does not establish an identical definition.
- Capture time, publication time and a selected reporting period are separate fields.
- Keep unavailable values distinct from observed zeros and unresolved comparisons.
Freeze the question and both source observations
A native panel shows 1,250 impressions while a dashboard shows 1,100. Before calling this an error, write the decision you need: a current post total, performance during a calendar month or an observation at a fixed post age. Those questions can require different values. Save each source privately with enough surrounding labels to establish what was displayed.
Give the case a local identifier. Record the publishing identity and original post URL, the source panel or export, publication timestamp if known, capture timestamp with zone, metric label, unit, filters and value or unavailable state. Keep the unmodified source and work on a separate note. A cropped number without its panel and filter context is insufficient to reconcile the claim.
Match meaning before subtracting values
Check the same post and account first; an original post and a repost are different objects. Then distinguish impressions from unique members reached, a post-level total from a channel daily total, organic from paid or combined distribution, and a current lifetime observation from events attributed to a selected period. A matching calendar filter does not settle these differences.
Metricool’s current personal-profile guide explicitly distinguishes content viewed within the period from posts published within the period; the latter can show cumulative performance through yesterday. LinkedIn’s member-statistics API documentation also records UI consistency caveats. These explain why scope checks matter, but do not establish that an individual mismatch is harmless or quantify a universal correction factor.
Compare collection clocks without inventing history
If one source was captured yesterday and another today, retain that difference. A later cumulative count may legitimately differ, but the direction alone does not prove the cause. Check whether the vendor reports a collection timestamp, yesterday’s endpoint or a documented delay. A manual refresh request is not proof that fresh source data was actually returned.
Buffer’s current discrepancy guidance describes batch collection and possible recent-post lag. Its other current help gives different history windows, so an undocumented historical value remains unverified. Do not erase a disagreement by reconnecting an account: a connection change can alter available history. First preserve the evidence and inspect the specific field and post you need.
Work through three different outcomes
R1 — Constructed time-scope case: the same post has 800 current lifetime impressions in source A and 500 impressions attributed to October in source B. The source labels and saved date filters confirm those different bases. Outcome: no comparable 300-impression error has been established. Use 800 only for the captured lifetime statement and 500 only for the supported October statement; do not average them to 650.
R2 — Constructed missing-value case: A contains an observed zero, while B contains a blank and says collection unavailable. Outcome: preserve 0 and unavailable separately. There is no pair of numeric observations to compare. The next step is to check B’s required source access, not to fill its blank with zero or exclude the post without recording why.
R3 — Constructed unresolved case: the saved account, original post, metric definition, distribution, grain, time basis and capture time all match, yet A has 1,250 and B has 1,100. Outcome: preserve both values and mark the discrepancy unresolved. The arithmetic difference is 150; it does not identify which source is correct. Hold the precise comparative claim and retain the minimal case for an authorized support review if needed.
Close with a defensible statement and an open-case record
The private worksheet records the matching checks, supporting source locations, resolved differences and remaining unknowns. Its conclusion must name what the chosen number actually supports. If a statement requires seven-day impressions but neither source preserves that observation, narrow the statement or hold it; a reconciliation cannot reconstruct an unsaved past measurement.
The examples are original fictional exercises, not measurements of vendor accuracy or a private LinkedIn account. No tool connection, refresh, export or support message was performed for them. Keep real identifiers and source files private, and share only the smallest authorized excerpt when seeking help. A completed form is a decision record, not independent validation.
Sources & product references
Product details reflect the publication date. External features and interfaces may change.
