Rank Tracking on Surfaces That Aren't Google

A position is only meaningful once you know what is being ordered, and how deeply you can see. That is true of Google and it is easy to forget when a stakeholder asks you to start tracking rankings inside a video platform, an app store, a marketplace, or another web search engine. The tracking mechanics look familiar. The data object underneath is not the same one, and the habits you built on organic web positions transfer unevenly.

Four questions decide whether a number from a new surface is worth putting on a chart.

One: what is the unit being ordered?

On web search it is a URL. Elsewhere it may be a video, an app, a product listing, a seller offer against a shared product page, or an account. The distinction matters because your identity in the ordering changes: on a marketplace where several sellers attach to one listing, “your position” may be the listing’s position, the offer’s position, or both, and those move for unrelated reasons.

Get this wrong and you will build a series that silently changes what it measures the first time the surface reorganises.

Two: is the ordering stable per retrieval?

Web positions already vary per retrieval — the reason a tracker’s number and your own screen disagree is that they are two separate retrievals under different conditions (incognito is not a neutral search). On surfaces where the user is nearly always signed in and the feed is built from their history, that variance is larger, and a position captured from a clean session may not resemble what any actual user saw.

That does not make it useless. It makes it a measurement of the default ordering, which should be labelled as such and never presented as “where our customers see us”.

Three: how deep can you see, and what does absence mean?

Every tracked surface has an observation depth, imposed by the interface or by whatever your tool pays for. Absence in the data means “not found within that depth”, which is not the same claim as “not present”. This is the same trap as a web keyword slipping past your tracker’s last page — the row goes blank and it reads as a catastrophe (when a keyword falls out of the top 100).

Surfaces with infinite-scroll interfaces make this worse, because there is no page boundary to define the depth. Write down the depth you are sampling and keep it constant, or your series is measuring your own configuration.

Four: does location apply, and how?

Web results have a location model with known coarseness — a city is not a point, and the tracker is choosing something on your behalf (local rank tracking: a city is not a location). Other surfaces vary: some are country-scoped only, some have storefronts per country with different catalogues, some order primarily by proximity, and some ignore location for most queries. The same query tracked “in the US” can mean four different things across four surfaces.

Language is a separate axis from country on some surfaces and fused with it on others. Record both explicitly.

Other web search engines are a traffic question first

A second general search engine has its own index, crawler, and ranking, so its positions are unrelated to Google’s and are not comparable — averaging them together produces a number that describes nothing.

The decision to track one is not primarily methodological, it is arithmetic: look at where your organic sessions actually come from before adding a tracked set. Tracking a surface that sends you almost nothing costs you the same attention as tracking one that matters, and it dilutes reporting. If the traffic is real, track it as an entirely separate set with its own baselines, not as extra rows in the Google set.

Rankings on platforms nobody outside can explain

In-platform search on video, marketplace, and app surfaces orders results using engagement and commercial signals that the platform does not publish. That has one honest consequence for a tracker: you can measure position and you cannot infer the mechanism. Anyone selling you a weighting for these surfaces has inferred it from a sample, and the same caution applies as to web ranking factors.

Practically, that means feature-style tracking rather than factor-style analysis. Record the position, record the attributes you can observe about the winners, and resist the write-up that explains why.

Don’t blend surfaces into one number

The temptation is a single visibility score across everything, because executives like one line. It hides the only useful information — which surface moved — and it weights surfaces by whatever depth each tracker happened to sample. Keep an index per surface, and if a combined figure is demanded, state the weighting explicitly and keep it fixed (visibility scores and share of voice).

What to actually do

1. Write down the unit being ordered for each surface, and check it again after any interface redesign.
2. Label default-session data as default-session data. It is not what a logged-in user sees.
3. Fix and document the observation depth, especially on infinite-scroll interfaces.
4. Record country and language as separate fields, with the surface’s own scoping rules noted.
5. Check your referrer data before tracking another engine, and keep every surface in its own set with its own baselines.