> ## 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.
# Ad Receiving Strategy
> How AdXensor decides which ad fills your slot — the targeting funnel, the weighted-random revenue auction, frequency capping, and guaranteed fill.
When your slot requests an ad, the core server has milliseconds to answer a hard question: *out of every live campaign, which ones can serve **this** visitor, and which one should win?* It answers in two phases — **eligibility** (who qualifies) and **auction** (who wins).
**Why this matters to you.** Understanding the funnel tells you exactly what makes campaigns *skip* your slot — and every skipped campaign is a slot filled by a lower bid or a house ad. Most fill problems are an eligibility problem, not a demand problem.
## Phase 1 — The eligibility funnel
Every campaign must clear **every** gate below to enter the auction. The first three are hard set-intersections resolved instantly from live in-memory pools; the rest are per-campaign checks.
```mermaid theme={null}
flowchart TD
A[All live campaigns] --> B{Country matches?}
B -->|no| X[Dropped]
B -->|yes| C{Device matches?}
C -->|no| X
C -->|yes| D{Ad size matches?}
D -->|no| X
D -->|yes| E{Language OK?
empty = matches all}
E -->|no| X
E -->|yes| F{Creative exists
for this size?}
F -->|no| X
F -->|yes| G{Budget remaining > 0?}
G -->|no| X
G -->|yes| H{Inside flight window?
start ≤ now ≤ end}
H -->|no| X
H -->|yes| I{Under frequency cap
for this visitor?}
I -->|no| X
I -->|yes| J[Enters the auction]
```
Country, device, and ad size are **hard AND filters** — a campaign that doesn't target the visitor's country *and* device *and* your slot size can never serve, no matter how high it bids. This is why supporting more ad sizes directly increases how many campaigns compete for your inventory.
### What each gate does
| Gate | Rule | What makes a campaign drop out |
| ----------------- | ------------------------------------------------------------------------------------------------------------------------ | ------------------------------------------------------------------------------------------------- |
| **Country** | Visitor's country (from IP) must be in the campaign's target set | Visitor is outside the campaign's geo targeting |
| **Device** | Visitor's device — **desktop, mobile, or tablet** — must match the campaign's device targeting | Campaign targets only desktop, visitor is on a tablet (or any device the campaign doesn't target) |
| **Ad size** | Your slot's size must be one the campaign has creative for | Your slot size isn't targeted |
| **Language** | Agnostic-OR: a campaign with **no** language targeting matches everyone; otherwise the visitor's language tag must match | Narrow language targeting + no matching tag |
| **Creative** | A creative must exist for the **exact** requested size | Campaign targets the size but uploaded no matching creative |
| **Budget** | `budgetRemaining > 0` | Budget exhausted |
| **Flight window** | `start ≤ now ≤ end` | Campaign hasn't started or has ended |
| **Frequency cap** | This visitor hasn't already seen the campaign too many times | Visitor hit the per-campaign view cap in the current window |
## Phase 2 — The revenue auction
Every campaign that clears the funnel competes in a **weighted-random revenue auction**. This is *not* a highest-bid-wins auction and *not* a pure random pick — it's deliberately in between.
Each campaign gets a **weight**, and the winner is drawn at random with probability proportional to its weight:
**weight = remaining budget × expected value per impression**
where *expected value per impression* is the campaign's effective CPM ÷ 1000. For click-goal (CPC) campaigns, the bid is converted to an equivalent CPM using an assumed click-through rate of **1%**. (A campaign with missing bid data still competes — its weight simply degrades to remaining-budget-only.)
### Why weighted-random instead of top-bid?
A campaign with a richer bid has a proportionally higher chance to win, so valuable demand is favored. But it doesn't win *every* time, which keeps a single advertiser from monopolizing your inventory.
Remaining budget is a direct multiplier. As a campaign spends down, its weight falls and it naturally yields to fresher demand — spreading spend across the whole day instead of front-loading it.
The practical takeaway for you: **more competing campaigns = higher effective CPM.** The weighting rewards value, so the deeper the pool of eligible demand on your slot, the more the auction has to pick from — and the more your impression is worth. Everything in the [Optimization Playbook](/publishers/how-it-works/optimization) is ultimately about deepening that pool.
### After a winner is drawn
The winner's destination URL is validated (it must be a real `http`/`https` link). If it's malformed, that campaign is dropped and the auction **re-draws** among the rest — paid demand is never skipped in favor of a house ad just because one winner had a bad link. Only if *every* eligible campaign fails does the slot fall through to house fill.
## Frequency capping — why the same ad stops showing
To prevent ad fatigue, AdXensor caps how many times a single visitor sees the same campaign within a rolling window — a configurable platform limit. Once a visitor hits the cap for a campaign, that campaign is filtered out of *their* auctions (other visitors are unaffected).
Counts are tracked per visitor **per campaign**, incremented only on **real, rendered** impressions — never on bot traffic, and never on house ads. House and always-fill banners are exempt from frequency capping, which is part of why your slots stay filled.
## Guaranteed fill — your slot is never blank
If no paid campaign qualifies (none matched, all frequency-capped, or all had invalid links), AdXensor walks a fallback chain so the slot still earns:
If an admin has uploaded a house creative for your slot's size, it serves.
Otherwise, for [standard IAB sizes](/publishers/reference/ad-formats) (`300x250`, `336x280`, `728x90`, `320x50`, `300x600`), a built-in AdXensor cross-promo banner fills the slot. It's an inline SVG — no external assets, no extra latency.
Only if the size is non-standard **and** no house creative exists does the slot return empty. Using standard IAB sizes avoids this entirely.
Every path in the fallback chain still earns you money — house fills credit **50% of the house CPM**. But house fills pay less than a competitive paid auction, so the goal is to maximize *paid* fill. See the [Optimization Playbook](/publishers/how-it-works/optimization).
## The whole strategy in one picture
```mermaid theme={null}
flowchart LR
R[Ad request] --> F[Eligibility funnel]
F -->|survivors| A[Revenue auction]
F -->|none survive| H[House fallback]
A -->|valid winner| S[Serve paid ad 💰]
A -->|all invalid URLs| H
H --> HC{House creative?}
HC -->|yes| S2[Serve house ad · 50% CPM]
HC -->|no, standard size| S3[Always-fill banner · 50% CPM]
HC -->|no, other size| E[Empty]
```
Winning the auction is only half the story — the impression still has to be *valid* to be counted and paid. See how AdXensor's fraud shield and reputation system work.