Buyer tools

Is that supply path paying for its QPS?

Two tools for the same question. One works out the number for a single path; the other ranks every path you buy through and says which to stop listening to. Both run in your browser. Nothing is uploaded.

01rCPM calculator 02Supply path scorecard
Tool 01 · rCPM calculator

What is one path worth per million?

rCPM — also called RevCPM — is the QPS efficiency metric. Revenue per million bid requests. Not per impression. Per request, including every one that never bid.

$
Total revenue earned from the path
Total requests sent over the same period
rCPM
$27.78
Revenue per 1,000,000 bid requests
Per thousand requests
$0.0278
The same figure on a per-mille basis, if you need to compare
Requests per $1
36,000
How much throughput it takes to earn a dollar
rCPM = ( Revenue ÷ Bid Requests ) × 1,000,000
Why it matters

Every request costs QPS.
Only some of them pay for it.

QPS is finite

Throughput is the constraint, not inventory.
  • Both sides pay to process a request — the seller to send it, the buyer to evaluate it
  • Below a certain rCPM a path costs more to listen to than it returns

Impression CPM hides waste

It only counts the requests that won.
  • A path can post a healthy CPM while burning millions of requests that never bid
  • rCPM counts the misses, so two paths with identical CPMs can look very different

It exposes duplication

Resold inventory shows up in the maths.
  • The same impression offered through several hops multiplies requests without adding revenue
  • Falling rCPM at flat revenue is a supply path problem, not a demand problem

There is no universal “good” rCPM — it moves with channel, format, geo and season. What matters is the trend on a given path, and how it compares against the others competing for the same throughput.

Tool 02 · Supply path scorecard

Which paths are earning their QPS?

Put every path you buy through into one table. It ranks them on what they cost you in throughput against what they actually return, and tells you which ones to stop listening to.

Saved in this browser

Copy the rows straight out of your DSP report and paste them here — path, bid requests, impressions, spend, tab or comma separated, one path per line. A header row is ignored.

Path Bid requests Impressions Spend QPS share Requests per imp eCPM rCPM Verdict
Keep

Takes no more throughput than its share of your impressions. These are the paths worth protecting.

Watch

Costs up to three times its share. Usually fine, worth a look at what the path is actually sending.

Trim

Costs more than three times its share of impressions. Ask the SSP to filter, or throttle it.

Cut

Requests, no impressions at all. Pure throughput cost with nothing coming back.

How the verdict works

Measured against your own mix.

No industry benchmark

Your paths, compared to each other.
  • Share of requests ÷ share of impressions is the whole rule
  • A path with 20% of your requests and 20% of your impressions scores 1.0
  • The same path at 2% of impressions scores 10, and that is the one to call

Why the ratio, not the rate

Win rate flatters small paths.
  • A path sending a thousand requests can post a fine win rate and still be irrelevant
  • Throughput is the constraint, so the question is what each path costs you to listen to
  • Pair it with the rCPM calculator for the seller side of the same question

It stays with you

No upload, no account.
  • Everything runs in the page — nothing is sent anywhere
  • Saved in this browser so the table is still here when you come back
  • Export CSV to take the scored table into your own deck
Where HBX fits

We shape traffic so your rCPM goes up.

Traffic shaping
Requests with no realistic chance to win are never sent
rCPM controls
Thresholds set per bidder, so you decide the floor
Atomic QPS
Counted exactly — no overages, no silent throttling
Supply path targeting
Filter on s-chain node count and direct vs resold
Setup Deal Build a list in Forage