How this score is calculated

We publish the weights. That makes the score gameable, and it is meant to: moving these five means claiming your portal profiles, collecting real reviews and actually posting, which is what the report advises anyway.

Model version be-v3

The five signals

They sum to 100. The weights are derived from what Belgium actually exposes publicly, not inherited from a market whose portal publishes more.

  • 35FindabilityYour own website and your profiles on Immoweb, Zimmo, Immovlan, Immoscoop and Realo.
  • 30ReputationHow many public reviews you have, weighted by your average rating.
  • 20ReachYour public followers across every channel.
  • 10Posting rhythmHow many weeks you posted in, on your most active channel.
  • 5VideoThe share of your recent posts that are video.

Three states, not two

The difference between "there is nothing there" and "we could not look" is the whole difference between a finding and an outage of ours.

  • measured measured, and counted.
  • absent we looked and found nothing. Scores zero and counts: a buyer being unable to find you is the finding, not missing data.
  • unavailable we could not check. Excluded, and its weight redistributed. Penalising you for our outage would make the number unreproducible.

What we refuse to do

  1. Below 50% coverage we publish no number at all. A score assembled from a third of the model is not a worse score, it is a different quantity.
  2. No rank below five peers. "#1 of 2" reads as a market position and is not one. The peer set is who we have graded, not who exists.
  3. AI visibility is a separate number, not a sixth signal. We never name you in the question we put to a model, we ask cold and check afterwards whether you were named.
  4. The same evidence always scores the same number. No randomness, no clock, no model judging your content.
  5. You can always ask to be removed, even if you never requested a report.
Methodology: ImmoSocials Agent Grader | ImmoSocials