Put two analytics tools on one site and they will disagree, usually by ten to thirty per cent. That is not a fault in either. They are measuring with different definitions.
They define a visit differently
Most tools end a session after a period of inactivity, and the period is a choice. A different timeout is a different number of sessions from identical traffic, and every per-session metric moves with it.
Some also restart a session at midnight, or when the campaign changes. Two of those rules produce two session counts from one afternoon.
They filter bots differently
Every tool removes automated traffic, and none of them removes the same set. One publishes its list, another treats it as proprietary, a third labels rather than drops.
This is usually the largest single cause, and it is invisible: neither dashboard shows you what it decided not to show you.
One of them is being blocked
Blockers do not treat all analytics equally. A tool on a widely blocked domain loses visits a less well-known one keeps, in a share that varies enormously by audience. A developer-facing site and a retail site are not comparable here.
They are in different time zones
A "day" is a boundary somebody chose. If one tool reports in UTC and the other in your local zone, every daily figure is a different slice, and the gap is largest in the hours around midnight.
They count page views at different moments
On load, on render, or after a delay to filter bounces. On a slow connection those are meaningfully different populations.
One of them is sampling
Some tools estimate above a volume threshold, and some do it only in certain reports. See when analytics data is sampled. It is worth knowing which of your numbers are counted and which are inferred.
They attribute differently
Last click, first click, a lookback window of thirty days or ninety. Two tools with different attribution give one conversion to two different sources and both are internally consistent.
Telling a normal gap from a broken install
A gap that is stable in proportion, one tool consistently 15% below the other, across pages and days, is definitional. Live with it and pick one as your number of record.
A gap that changes shape is a fault: it appears suddenly, or affects one page, one browser or one country. That is worth investigating; see why your traffic dropped overnight.
Pick one to be right
The mistake is trying to reconcile them. Choose the tool your decisions come from, use the other for cross-checking direction only, and never put both numbers in the same report.
If you want the definitions Skomi counts by, they are written down rather than implied: visit, visitor and page view each say what they include and exclude. The Analytics page applies them consistently, and the comparison pages put the tools next to each other row by row.
Related: when analytics data is sampled, why Search Console and analytics disagree, and switching analytics without losing your history.