Almost everyone who starts tracking rankings hits this within a month: Search Console reports one position, a rank tracking tool reports another, and a manual search shows something different again. The instinct is to assume one of them is broken.
Usually none of them are. They are measuring different things.
Search Console Reports an Average, Not a Position
The Position column in the Performance report is the average position across every impression recorded for that query over the selected period. If your page appeared at 3 for some searches and 28 for others, the reported figure lands somewhere in between — a number that may match no result page anyone actually saw.
Two consequences follow. Averages are excellent for spotting a trend across weeks and misleading when quoted as a current ranking. And because impressions are only recorded when you actually appeared, Search Console gives you no information about queries where you ranked nowhere.
There is also a lag of roughly two to three days. A change you made yesterday is not in today's data.
Rank Trackers Report a Sample, Taken Under Controlled Conditions
A tracker requests the keyword from a specific location, on a specific device, in a specific language, at a specific time. That is a real observation of a real result page — but it is one observation.
Change any parameter and you can get a different answer. A keyword checked from a data centre in one city and a user on a phone in another are genuinely different searches. This is why two reputable trackers can report different positions for the same keyword on the same day and both be accurate.
Manual Search Reports One Personalized Result
Your own search is shaped by your history, your device, your language settings, and the location Google infers from your IP. Incognito removes some of that and not all of it. A manual check is the least reliable of the three and the one people trust most, because they saw it with their own eyes.
The Other Reasons Numbers Diverge
Different keyword sets. Trackers check an exact phrase. Search Console aggregates close variants, plurals, and misspellings into one query. Slightly different inputs produce different results that are easy to mistake for a discrepancy.
SERP features. Whether a featured snippet, an AI overview, or a local pack occupies the top of the page changes where organic results begin. Two tools that count position differently — one counting the snippet as position zero, one not — will disagree by one on the same page.
Volatility. Google runs continuous ranking experiments. Two checks an hour apart on the same keyword can legitimately return different orders.
Time zone and timing. A tracker that runs at 02:00 and one that runs at 14:00 are sampling at different points in that volatility.
Which One to Trust
Pick one source and use it consistently. Mixing sources is what creates the illusion of a problem.
Use Search Console for trends on your own site. It is Google's own data, it covers every query you actually appeared for, and its averages smooth out daily volatility in a way that makes month-over-month direction trustworthy. Accept that it cannot answer "where are we today."
Use a tracker for current position and competitor data. It gives you a controlled, repeatable measurement including competitors, which Search Console structurally cannot. Accept that it is a sample.
Use manual checks for verification, not measurement. Good for confirming a page is indexed or that a title change took effect. Not for reporting.
The practical rule: never present a position from one source next to a trend from another. If you report weekly deltas, take every number from the same tool with the same settings. Switching tools mid-project resets your baseline, and the apparent movement afterwards is an artefact of the switch.
When a Discrepancy Is Real
Sometimes the numbers disagree because something is genuinely wrong. Signals worth acting on: your tracker shows you ranking and Search Console shows zero impressions for that query over weeks (suggests the tracker is matching a variant nobody searches, or the page is not indexed); or Search Console shows steady impressions while your tracker shows you absent (suggests a location, device, or language mismatch in how the keyword is configured).
Both are configuration problems, and both are worth fixing before you draw conclusions from either number.
Where SerpCraft Fits
SerpCraft records rank history from a consistent, explicitly configured set of conditions — same location, same device, same language — so week-over-week movement reflects something real rather than a change in how the check was run. The SERP analysis also captures which features were present, which is often the explanation when a position holds steady and traffic does not. The free plan covers a handful of analyses per month, enough to compare its numbers against what you are already seeing.