Mobile and Desktop Are Two Rank Sets

One URL, one keyword, two positions — both correct. Mobile and desktop result pages are assembled and laid out separately, so a page can be fourth on one and seventh on the other on the same afternoon. Tracking one device and calling the number “our position” is not a rounding error; it’s reporting the wrong series.

The fix is cheap. The cost of not fixing it is a class of unexplainable movement in your data.

Where the difference comes from

Layout, first and foremost. A narrow viewport shows fewer results without scrolling and gives feature blocks proportionally more room. A local pack that occupies a third of a desktop screen can occupy most of a phone screen. That doesn’t change the ordinal position, but it changes what the position is worth, and it interacts with which blocks get shown at all.

Feature composition. Some blocks are much more common on one device than the other. Query intent skews by device too — the same keyword can have a different mix of intents behind it on a phone than on a laptop, which is enough to change what kind of results get assembled.

Page experience factors. Whether a page is usable on a small screen has been a stated consideration for mobile results for a long time. A page that’s slow or awkward on a phone can plausibly sit lower there. Beyond that general point, anyone who tells you the weighting is speculating.

Result count per page. Not always identical across devices, which shifts where “page two” starts and therefore how a tracker with limited search depth reports deep positions.

Which one to track

Track both if your budget allows, since it’s usually a per-keyword-per-device charge and doubling a small tracked set is cheaper than doubling a large one.

If you have to pick one, pick the device your audience actually uses, which you can read from Search Console’s device breakdown rather than guessing. Some rules of thumb that hold up:

Local and consumer queries skew mobile heavily. Track mobile.
Professional, B2B and software queries often skew desktop. Track desktop.
Anything with an obvious on-the-move use is mobile regardless of what your current traffic says, because your current traffic reflects where you already rank.

The important part is that you chose, and that the choice is recorded next to the report. A tracked set assembled by whatever the tool defaulted to is the common case and the worst one.

The reporting trap

Here’s the failure this causes. Your tracked set is desktop. Your traffic is mostly mobile. Mobile positions slip; desktop positions hold. Clicks fall, your tracker shows nothing, and you spend a week looking for a technical cause in the wrong place.

Nothing in the data is wrong. The data just isn’t about your traffic. This is the same structural problem as sampling one point in a city and calling it the city — see a city is not a location.

The reverse trap is subtler: tracking both devices and then averaging them into one number. Now you’ve built a series that moves when either device moves and tells you nothing about which, which is strictly less useful than picking one.

Comparing the two series usefully

When you do track both, the gap between them is worth a glance each month.

A stable gap is normal and uninteresting. A gap that widens is a signal, and usually a specific one: something about the page’s mobile rendering or a change in mobile layout for that query. It’s one of the few genuinely diagnostic patterns in rank data, because a device-specific divergence rules out most sitewide and most content-level causes at once. That narrowing logic is the core of how to diagnose a ranking drop.

A gap that inverts — mobile better than desktop, having been worse — usually means the desktop page composition changed rather than anything about you.

What to actually do

Look up your real device split in Search Console before choosing what to track.
Track both devices for your top terms even if the long tail is single-device.
Never average devices into one number. Two charts, or one chart with two lines.
Record the device on every report, in the same place you record location.
Watch the gap, not just the levels. A widening device gap is one of the few ranking signals specific enough to act on.