> ## Documentation Index
> Fetch the complete documentation index at: https://docs.adxensor.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Glossary & Key Metrics
> Every publisher term in plain language, plus a one-page cheat-sheet of every threshold, formula, and window AdXensor uses — auction weights, revenue share, reputation bands, the IVT penalty, payout tiers, and fraud-shield windows.
This is the reference you keep open in a second tab. **Part A** defines every term you'll meet across the [How AdXensor Works](/publishers/how-it-works/overview) pages in plain language. **Part B** is the cheat-sheet — every number, band, and formula the platform actually runs, consolidated so you can scan them in one place instead of hunting across pages.
**One source of truth.** Every value below is the value the live system uses. Where a page like [Traffic Quality & Scoring](/publishers/how-it-works/traffic-quality) or [Earnings & Payouts](/publishers/how-it-works/earnings) explains the *why*, this page is the *what* — the exact thresholds, side by side.
***
# Part A — Glossary
## Serving & the auction
Terms that decide **which** ad fills your slot. Deep dive: [Ad Receiving Strategy](/publishers/how-it-works/ad-selection).
| Term | What it means |
| --------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Eligibility funnel** | The ordered set of gates a campaign must clear to compete for your slot: country → device → ad size → language → creative-exists → budget remaining → flight window → frequency cap. Country, device, and size are hard AND-filters; the rest are per-campaign checks. |
| **Weighted-random auction** | How the winner is picked among eligible campaigns. Not highest-bid-wins and not pure random — each campaign gets a *weight*, and the winner is drawn with probability proportional to its weight. Favors valuable demand while pacing budget and preventing one advertiser from monopolizing your inventory. |
| **Flight window** | A campaign's live dates (`start ≤ now ≤ end`). Outside this window it can't serve. Enforced at serve time as defense-in-depth so a stale campaign can never leak an impression before its start or after its end. |
| **Frequency cap** | The limit on how many times a single visitor sees the *same* campaign inside a rolling window. Once a visitor hits it, that campaign is filtered out of **their** auctions only. Counted on real rendered impressions — never on bots, never on house ads. Configurable per platform; disabled when set to `0`. |
| **Fill / fill rate** | Whether an ad request returns an ad, and the share of requests that do. AdXensor targets **\~100% fill** via the fallback chain, so a fill problem is almost always an *eligibility* problem, not a demand problem. |
| **House ad** | The cross-promo ad served when no paid campaign qualifies, so your slot still earns. It credits a **flat 50% of the house CPM**. |
| **Always-fill** | The last-resort built-in AdXensor banner (an inline SVG, no external assets) that fills a standard-size slot when no paid campaign *and* no admin house creative is available — so the slot is never blank. Standard sizes only: `300x250`, `336x280`, `728x90`, `320x50`, `300x600`. |
## Money & metrics
The signals and rates that turn a served ad into revenue. Deep dive: [Earnings & Payouts](/publishers/how-it-works/earnings).
| Term | What it means |
| ----------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Impression** | The ad was rendered and its beacon fired. This is the billable event advertisers pay for on CPM campaigns — counted **exactly once** per rendered ad, even if multiple beacons fire. |
| **View** | A *viewability* signal — the ad met the IAB standard (≥ 50% on screen for ≥ 1 second), measured client-side by the SDK. **Not billable**; a placement-quality signal only. |
| **CPM** | *Cost per mille* — the price for **1,000 impressions**. The default billing model for every campaign goal except traffic. The advertiser is charged `CPM bid ÷ 1,000` per valid impression. |
| **CPC** | *Cost per click* — the price for one click. Used **only** for traffic-goal campaigns. The advertiser is charged the full `CPC bid` per valid click. |
| **eCPM** | *Effective CPM* — the blended revenue a slot earns per 1,000 impressions, regardless of underlying model. Internally it's the auction's value signal: `CPM ÷ 1,000` per impression for CPM demand, or `CPC bid × assumed CTR` for CPC demand. Higher eCPM = more valuable inventory. |
| **Revenue share** | Your cut of a paid campaign's gross spend. Platform default **60% to the publisher** (per-account configurable, 0–100%). House-filled slots instead credit a flat **50% of the house CPM**, with the share already baked in. |
## Quality & reputation
How AdXensor separates real humans from fraud, and grades your inventory. Deep dive: [Traffic Quality & Scoring](/publishers/how-it-works/traffic-quality).
| Term | What it means |
| ------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **IVT / invalid traffic** | *Invalid traffic* — events flagged by the fraud shield as bots or automated. Detected by three layered signals: known **bot user-agents**, a **known-bot / data-center IP** blocklist, and **velocity** flood protection. Invalid traffic is recorded for your reports but **never counted, never charged, never paid**. |
| **Invalid rate** | The share of your impressions the shield filtered as invalid, over 30 days: `invalid ÷ (valid + invalid)`. Drives the reputation penalty and gates your payout tier (which uses your **worst** site's rate). |
| **Reputation score** | A **0–100** grade recomputed daily per site from the last 30 days: volume band + a healthy-CTR bonus, minus the invalid-traffic penalty. Higher reputation raises how much your inventory is valued when advertisers plan campaigns, and lifts your payout tier. |
| **Verified site** | A site that has served **at least one valid impression** in the last 30 days (`isVerified = true`). Unserved sites sit at reputation `0`, unverified. At least one verified site is required for the Standard and Trusted payout tiers. |
## Wallet & payouts
Where your money lives and how it moves. Deep dive: [Earnings & Payouts](/publishers/how-it-works/earnings).
| Term | What it means |
| --------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| **Pending balance** | Freshly credited earnings. Yours, but not yet withdrawable — serving out the **hold period** for your tier. |
| **Available balance** | Funds past their hold. This is the balance you can request a payout against, subject to the minimum. |
| **Locked balance** | The amount moved out of *available* when you request a payout, so it can't be double-spent while an admin reviews. Returned to *available* if the payout is rejected or fails; drained when it's paid. |
| **Hold period** | The delay before *pending* becomes *available*. Set by your payout tier: **30 days** (New / Standard) or **15 days** (Trusted). |
| **Payout tier** | Your account-wide standing — `new`, `standard`, or `trusted` — computed daily from your sites' reputation and invalid rates. Controls your hold period, minimum payout, and admin-queue priority. |
***
# Part B — Key Metrics cheat-sheet
Every threshold the platform actually runs, grouped by system. Selection-only heuristics are marked — they influence *which ad wins*, never *what you're paid*.
## Auction & serving
**Auction weight.** Each eligible campaign's win probability is proportional to:
```
weight = max(0, budgetRemaining) × expectedValuePerImpression
expectedValuePerImpression =
goal is "traffic" → CPC bid × assumed CTR
otherwise → CPM bid ÷ 1,000
(floored to a tiny positive so a campaign with missing bid data still competes)
```
| Lever | Value | Notes |
| -------------------------------- | ------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Assumed CTR** (CPC → eCPM) | **1%** (0.01) | Selection weighting **only** — never billing. Converts a CPC bid to a comparable per-impression value in the auction. |
| **CPM charge** (advertiser) | `CPM bid ÷ 1,000` per valid impression | Charged once per rendered impression, at beacon time. |
| **CPC charge** (advertiser) | `CPC bid` per valid click | Traffic-goal campaigns only. |
| **Invalid-URL re-draw** | Drop winner, re-draw | If a drawn winner's destination isn't a valid `http`/`https` URL it's dropped and the auction re-draws — paid demand is never skipped to house over one bad link. |
| **Fallback (always-fill) sizes** | `300x250` · `336x280` · `728x90` · `320x50` · `300x600` | Standard IAB sizes that get a built-in banner. Any other size with no house creative returns empty. |
CPM and CPC are charged at **impression / click time**, not serve time — there's no serve-time budget decrement. The auction weight uses *remaining budget* purely for pacing, so a campaign naturally yields as it spends down.
## Revenue & earnings
| Model | You're paid per | Gross formula |
| ------------------------- | ----------------- | --------------------------------- |
| **Traffic (CPC)** | Click | `clicks × CPC bid` |
| **Everything else (CPM)** | 1,000 impressions | `(impressions ÷ 1,000) × CPM bid` |
| Rate | Value | Notes |
| --------------------------- | -------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| **Publisher revenue share** | **60%** (default) | Per-account configurable, 0–100%. Applied by client-api to paid-campaign gross. |
| **House ad share** | **50% of house CPM** | Publisher net computed as `round((impressions ÷ 1,000) × house CPM × 0.5)`; credited directly (share already baked in). |
| **Credit cadence** | Nightly | Valid, non-bot events aggregated per site + campaign each night. |
| **Earnings lookback** | **3 calendar days** | Every run re-scans the last 3 days. A missed/failed day self-heals; an already-credited day is a silent no-op (never double-paid). |
| **Settlement currency** | XOF | Paid earnings settle in XOF; mixed-currency wallets can't be paid out in one request. |
**Worked example.** A 500 XOF CPM campaign serves 20,000 valid impressions on your site in a day → gross `= (20,000 ÷ 1,000) × 500 = 10,000 XOF` → your 60% share `= 6,000 XOF` credited.
## Reputation ladder
Base score by 30-day **valid** impression volume, with a bonus for keeping CTR in the healthy band.
| 30-day valid impressions | Healthy CTR (0.5%–6%) | Unhealthy CTR |
| ------------------------ | :------------------------------------: | :------------: |
| 0 | 0 (unverified) | 0 (unverified) |
| 1 – 999 | 10 – 39 | 10 – 39 |
| 1,000 – 9,999 | **55** | 42 |
| 10,000 – 49,999 | **70** | 58 |
| 50,000+ | **82**, or **90** if avg CPM ≥ 300 XOF | 65 |
The **1 – 999** band scales linearly with volume: `10 + round((impressions ÷ 1,000) × 29)`. **Healthy CTR** is `0.5% ≤ CTR ≤ 6%` (`clicks ÷ impressions`). The top **90** score also requires a 30-day average CPM of **≥ 300 XOF**.
### Invalid-traffic (IVT) penalty
Applied after the base score, then clamped to ≥ 0.
```
final score = max(0, round(baseScore − invalidRate × 40))
```
| Invalid rate | Penalty |
| ------------ | ------------- |
| 10% | −4 |
| 25% | −10 |
| 50% | −20 |
| 100% | −40 (maximum) |
**Invalid rate** `= invalid impressions ÷ (valid + invalid impressions)`, measured over 30 days. A clean site (\< 2% invalid) is barely touched; a dirty site can shed nearly half its score.
## Dashboard labels
How raw numbers read in your dashboard.
| Score | Label |
| ------- | ------------- |
| ≥ 80 | **Excellent** |
| 65 – 79 | **Good** |
| 45 – 64 | **Average** |
| \< 45 | **Low** |
| Invalid rate | Label |
| ------------ | --------- |
| ≤ 2% | **Clean** |
| ≤ 10% | **Good** |
| ≤ 25% | **Watch** |
| > 25% | **High** |
## Payout tiers
Computed daily from your sites' reputation. Gating uses your **worst** site's invalid rate and requires at least one **verified** site.
| Tier | Hold | Minimum payout | Qualification |
| ------------ | :---------: | :--------------------------------: | ---------------------------------------------------------------------------------------------------------------------- |
| **New** | 30 days | Full platform minimum (×1.0) | Default for every new account |
| **Standard** | 30 days | Full platform minimum (×1.0) | ≥ 1 verified site · avg reputation ≥ 45 · worst invalid rate ≤ 20% |
| **Trusted** | **15 days** | **40%** of platform minimum (×0.4) | ≥ 1 verified site · avg reputation ≥ 65 · worst invalid rate ≤ 10% · (≥ 1 completed payout **or** avg reputation ≥ 80) |
**Minimum payout** `= round(platform minimum × tier factor)`, where the factor is **1.0** (New / Standard) or **0.4** (Trusted). Trusted also **fast-tracks** your requests to the top of the admin queue.
Because tiers key on the **worst** invalid rate across *all* your sites, one neglected, dirty property can hold your entire account at a lower tier — keep every site clean.
## Fraud shield, guards & windows
The IVT shield's signals and the key time-windows the platform runs.
| Setting | Value | What it governs |
| -------------------------------- | ---------------------------------- | -------------------------------------------------------------------------------------------------------------------- |
| **Bot signals** | `ua_bot` · `cidr_bot` · `velocity` | Bot user-agent, known-bot / data-center IP, and flood detection — layered, precise-first. |
| **Shield modes** | `off` · `log` · `block` | In `log` and `block`, flagged traffic is **never counted or paid**; `block` additionally serves no ad / no redirect. |
| **Velocity flood-guard** | Rolling per-IP + user-agent window | Catches abnormal request bursts from a single source; tuned so ordinary multi-slot page views never trip it. |
| **Reputation / audience window** | 30 days | Rolling window for reputation scoring and audience stats. |
| **Stats aggregation** | Every hour | Valid vs. invalid impressions/clicks rolled up per site, country, device. |
### TTLs & dedup windows
| Key / event | TTL | Purpose |
| -------------------------------------------- | ----------------- | ------------------------------------------------------------------------------------- |
| **Serve-token metadata** (`event:`) | 2 hours (7,200 s) | Links an impression/click beacon back to its campaign + destination. |
| **Click / impression dedup** | 2 hours (7,200 s) | One-shot guard so a rendered ad counts and charges **once** across every beacon path. |
| **Attribution key** (`attr:`) | 30 days | Long-lived so a late conversion still resolves to the right campaign. |
| **Conversion dedup (with click / order id)** | 30 days | Two real conversions behind one NAT both count; exact replays don't. |
| **Conversion dedup (no id)** | 1 day (86,400 s) | Coarse per-campaign + IP flood guard fallback. |
| **House config / creative cache** | 60 s | In-memory cache of house CPM + creatives. |
| **Stale draft purge** | 7 days | Unsubmitted draft sites are cleaned up. |
Advertiser-reported **conversion revenue** is analytics-only — it never affects budget, billing, or your earnings — and is clamped to a 1e9 XOF ceiling so a forged value can't distort reporting.
## Wallet lifecycle at a glance
```mermaid theme={null}
flowchart LR
E[Valid events] -->|nightly credit| P[Pending]
P -->|hold elapses
30 or 15 days| A[Available]
A -->|request payout ≥ minimum| L[Locked]
L -->|admin: approved → processing → paid| PAID[Paid 🎉]
L -.->|rejected / failed| A
```
***
## Keep reading
The full lifecycle of one impression, request to wallet.
The eligibility funnel and the weighted-random auction in depth.
The IVT shield and the 0–100 reputation score.
Revenue formulas, the 60% share, and the wallet lifecycle.
Turn every number above into a concrete checklist to earn more.