Research practice
Research Methodology
How Indexora designs experiments, reads the results and reports the limits, so that a test result means something you can act on.
Why methodology comes first
Most SEO advice repeats what others have said. Indexora tries to observe search behaviour directly, and a result is only useful if the way it was produced is transparent.
Every Lab report follows the same shape: question, hypothesis, protocol, results and limitations. That structure lets a reader separate a genuine signal from noise and decide whether the finding applies to their own site.
Last reviewed .
The experiment protocol
- 01
Start from a narrow question
A study answers one question at a time. A broad claim such as "does X help rankings" is reduced to something observable on a defined set of pages within a defined window.
- 02
Write the hypothesis first
We record what we expect to see and why before touching anything. Writing the prediction in advance keeps us from reading a random fluctuation as a result.
- 03
Control the change
Where the environment allows, only one variable moves while related signals are held steady. When full control is impossible, that limit is stated instead of hidden.
- 04
Observe over a stated window
Metrics, tools and the observation period are fixed up front. We report the raw direction of the signal, not a smoothed story built after the fact.
- 05
Separate observation from conclusion
What we saw and what it means are kept apart. An observation on one site is not presented as a universal ranking rule.
- 06
Publish limitations
Sample size, confounding factors and anything we could not control are published alongside the result so readers can judge how far it transfers.
Standards every study meets
These are the tests a study has to pass before it goes on the site.
- Reproducible
- Method and inputs are described in enough detail that another practitioner could repeat the test.
- Falsifiable
- Each test could show the opposite of what we expected, and we report it when it does.
- Dated
- Search systems change, so every result carries the date it was observed.
- Non-precise where honest
- We avoid fake-precise numbers when the data cannot support them.