Open your platform's native analytics next to a third-party dashboard and you'll often find they disagree — sometimes by a little, sometimes embarrassingly. The instinct is to ask which tool is broken. Usually neither is. They're measuring different things, at different times, through different plumbing, and understanding that plumbing is the difference between reading your numbers and misreading them.
The three ways a tool gets your data
Every social analytics product sits on one of three foundations, and the foundation determines what it can know.
- Official APIs. The platform grants the tool structured access to your account's data, with your permission. This is the most reliable path, but the platform decides what the API exposes, how often you can call it, and how far back it reaches. Plenty of metrics you see in the native app simply aren't in the API at all, so the tool can't show them — not because it's inferior, but because it was never given them.
- Scraping. The tool loads public pages the way a browser does and reads numbers off them. This reaches data the API withholds, but it's brittle: page layouts change, platforms actively resist it, and scraped counts are whatever the public page happened to display when the scraper visited — rounded figures, cached values, or numbers a test bucket was shown. It also typically violates platform terms, which is why scraper-based tools sometimes break overnight.
- Your own exports. CSVs downloaded from the native analytics screen. Most accurate snapshot, most manual, and frozen the moment you download it.
Where the disagreements actually come from
Even two tools using the same official API can differ, and the reasons are mundane once you see them:
- Timing. One tool synced at 6am, the other at noon, and your post accumulated views in between. Numbers for recent posts are always moving targets.
- Definitions. "Impression," "view," and "reach" are not standardized across platforms, and tools sometimes relabel them. One dashboard's "views" may be another's "impressions" measured under a different counting rule — does a repeat viewer count twice? Does two seconds of visibility count at all? The platform defines this; the tools inherit it unevenly.
- Time zones and day boundaries. A "Tuesday" that runs midnight-to-midnight UTC and a "Tuesday" in your local time are different days, and weekly rollups quietly diverge because of it.
- Retroactive correction. Platforms remove spam accounts and re-audit counts, so a number can legitimately go down. A tool that stored last week's figure now disagrees with a tool that refetched it.
- Sampling and estimates. Some figures — especially audience demographics — are modeled estimates, not counts. Two estimates of the same quantity will differ; that's what estimates do.
How to work sanely with imperfect numbers
The practical response isn't to find the one true tool. It's to change what you ask of the tools you have.
- Pick one source per metric and stay with it. Consistency matters more than accuracy for decision-making. A dashboard that's uniformly 10 percent off still shows you every trend correctly; switching sources mid-stream shows you trends that don't exist.
- Treat native analytics as the reference and third-party tools as the workspace. When numbers conflict, the platform's own screen is closest to the source of truth. The third-party tool earns its keep on comparison, history, and export — not on being more correct.
- Compare ratios, not raw counts. Engagement per impression, follows per post. Ratios cancel out much of the measurement noise that plagues absolute figures.
- Let numbers settle before judging them. A post's first-day figures are provisional in every system. Score posts after a fixed window — say, seven days — so you're always comparing settled values to settled values.
Analytics discrepancies feel like a scandal the first time you notice them. They're actually a lesson in what these numbers are: not readings from a precision instrument, but reports from a messy pipeline, useful in aggregate and over time. Trust the direction, hold the third decimal place loosely, and never let two dashboards' disagreement about a number distract you from what both agree on — whether it's going up.