Placement Score: one number, with the evidence attached
Buyers ask for one number. The danger of one number is that it hides its parts and invites gaming, so ours is versioned, shown with its components, computed only from measurements a seller cannot influence, and deliberately excludes the things that belong to the buyer’s brief rather than to the publisher.
What goes into it?
- Authority (weight 25): claimed DR and our measured rank, averaged; a subdomain’s parent DR is ignored.
- Visibility (25): measured traffic and ranking keywords on log scales.
- Quality (30): 100 minus a penalty per quality test that fired, spam over 30, outbound-to-inbound ratio, restricted content.
- Graph (10): inbound domains, PageRank in the web graph, share of a topic’s leaders linked.
- Survival (10): live share of our own placements on the publisher.
What is deliberately left out?
Relevance to your brief, supplier reliability and observed rank movement. Each is reported beside the score so you can see which is which; folding them in would make the same publisher score differently in every quote, and observed effectiveness would be weighted on too few outcomes to mean anything. The weights are stated, not fitted; they will change against outcomes, and the version will change with them.
How should you read it?
- Read the components, not just the number; the API and the agent both return them.
- A suspect publisher is capped at 40 whatever its authority reads.
- Confidence is how much of the weight had a measurement behind it. Low confidence means “we do not know much”, not “bad”.
- Two publishers a few points apart are not ranked by it. There is no sort-by-score, on purpose, until outcomes back the weights.
It is on every row and in every why:
“Plan 5 verified finance placements and show each one's Placement Score with its components.”
Methods: research methodology.