Recommendations

Recommendations are the action items in your reports: concrete changes to make, drawn from what your numbers say. They are generated from your own measured results, the same per-metric, per-engine numbers your dashboard shows, plus where you and your tracked competitors stand relative to each other.

Each one is tailored to the engine it targets, drawn from a reviewed list of actions that can actually influence that engine. Claims about things only the vendor controls are kept off the table.

Every figure a recommendation states is checked before it reaches you. A draft may only state a number, percentage, or rank that traces back to a measured value it cites as evidence; a draft that states a figure its evidence does not support is rejected automatically.

Each recommendation carries a confidence score computed from the sample sizes behind its evidence: the fewer responses a number rests on, the lower the confidence. That score is computed by us, never self-rated by the model that drafted the text.

A person reviews every recommendation before it is published, with each cited number displayed beside the draft, so the claims can be checked by eye as well as by machine.

Recommendations never promise placement. The actions raise your odds; no lever guarantees a ranking, a source link, or a visibility outcome, and we do not pretend otherwise.