What is pre-bid filtering in programmatic advertising?

Pre-bid filtering is the practice of evaluating a bid request for fraud or invalid-traffic risk before the DSP decides whether to bid on the impression, rather than discovering the traffic was invalid after spend has already occurred. IP is one of the few signals reliably present in the bid request itself, which makes it usable at this stage without extra data collection.

Does pre-bid IP filtering add latency to the bid process?

Not if it's implemented as a file-based lookup rather than a live network call. IP Raccoon's datasets can be loaded directly into the DSP's own infrastructure and queried in-process, so bid requests are scored in single-digit milliseconds without a round trip to an external API during the bid itself.

Should a datacenter IP always be excluded from bidding?

Not automatically. A datacenter IP in a mobile in-app bid request is a strong invalid-traffic signal, since real mobile users don't connect from datacenter infrastructure. The same datacenter signal on a desktop web request is far weaker, since legitimate corporate VPNs and cloud-hosted browsers also originate from datacenters. IP Raccoon returns the category so buyers can set their own thresholds and rules per inventory type instead of applying one blanket rule.

How is this different from post-bid or post-campaign fraud detection?

Post-bid and post-campaign detection identify invalid traffic after the impression has already been bought and, often, after the click or conversion has been billed. Pre-bid filtering intervenes before the bid is placed at all, which is the only point in the pipeline where you avoid spending on the impression rather than trying to recover or dispute the spend afterwards.

Use case · Advertising, marketing & DSP's

Pre-Bid IP Filtering for DSPs & Programmatic Buying

A DSP has single-digit milliseconds to decide whether a bid request is worth bidding on at all. IP is one of the few signals reliably present before any bid is placed — which makes it the natural filter to run before spend, not after.

Start for free See how it works

Why bid-time is the moment that matters

Quick answer

Pre-bid IP filtering scores a bid request's IP for invalid-traffic risk before the DSP decides whether to bid, using category signals (datacenter, residential proxy, bot) rather than a network call, so the check adds no latency to the bid loop.

Every other stage of fraud detection — post-bid, post-click, post-conversion — is about identifying spend that's already happened and trying to get it back or exclude the source going forward. Pre-bid is the only point where you can simply not spend it in the first place.

The constraint is speed. A DSP's bidding logic runs in single-digit milliseconds, so anything added to that path has to be effectively free. IP is one of the very few signals present in the bid request itself before a network call, cookie sync, or device fingerprint is possible — which is exactly why it's the workable filter at this stage.

How IP Raccoon helps

IP Raccoon ships as downloadable files, so DSPs get an IP risk reading without a single network call inside the bid process. The dataset loads directly into the buyer's own infrastructure and gets queried in-process, and the permalink it comes from is stable, so every new release lands without re-integration.

The category breakdown is what makes it usable across inventory types, rather than one blanket rule that's either too aggressive or too permissive depending on context:

Signal	Mobile app inventory	Desktop web inventory
Datacenter IP	Invalid traffic — real phones don't connect from datacenters	Weaker signal — corporate VPNs and cloud browsers overlap here
Residential proxy	Suspicious, worth a discount on the bid	Suspicious but not disqualifying on its own
Known bot	Exclude	Exclude

Buyers set their own thresholds and rules per inventory type instead of inheriting someone else's blanket policy.

What to look for in a pre-bid IP filter

File-based delivery, so the check runs in-process with zero added latency to the bid loop
Category breakdown (datacenter, residential proxy, known bot, anonymiser), not a single opaque score
Refresh frequency fast enough to catch newly-spun-up datacenter ranges and proxy pools
Pricing that separates the per-query credits from the dataset subscription, so each can be sized to the test or the campaign

$ curl -o anonymizers.mmdb \
  "https://files.ipraccoon.com/download/anonymizers?authorization=$IP_RACCOON_TOKEN&format=mmdb"

slugs:   reputation · anonymizers · detected_bots · known_bots
formats: csv (subnet) · csv_plus (subnet + subcategory) · mmdb

Filter before you spend, not after

Load IP Raccoon's datasets into your bidding infrastructure and see the invalid-traffic share of your inventory before the next auction.

Start for free See pricing

FAQ

Frequently asked questions

More use cases

All use cases Affiliate fraud detection Credential stuffing detection Edge rules on CDNs & WAFs Glossary Read the docs