12 ms
Analyze
The tag reads up to 14 signals: country, device, OS, browser, format, placement, time of day, currency and session depth. No personal data is stored.
The engine
One decision per impression. Made in 83 ms.
BGrowth reads the context of every ad request, predicts the net yield of each connected partner and routes the impression to the best one, before the page finishes loading.
Decision trace
Brazil · Android · Banner 320×50
- Analyze14 signals read12 ms
- Predict15 partners scored41 ms
- DecideNetwork 07 selected18 ms
- LearnOutcome logged12 ms
$2.84eCPM+42% vs. waterfall
83 msmedian decision time
1.2Bdecisions per day
15partners scored per request
99.99%uptime over 12 months
How a decision is made
Four steps, inside the time
it takes an ad slot to load.
No waterfall, no fixed priorities. Each request is scored on its own context, then the outcome teaches the model for the next one.
41 ms
Predict
For each connected partner, the model predicts eCPM, fill probability and response time for this exact segment, net of fees.
18 ms
Decide
The partner with the highest expected net yield wins, within your rules: price floors, blocklists, latency budget and frequency caps.
12 ms
Learn
The real outcome is logged and fed back. A small share of traffic keeps exploring alternatives so the model never goes stale.
Architecture
One layer between your traffic
and every partner you trust.
Your networks keep their tags, contracts and payouts. BGrowth sits in front of them and decides which one serves each impression.
Your ad partnersand 9 more
ExoClick
Clickadu
Adsterra
HilltopAds
TrafficStars
TwinRed
- Geography
- Currency
- Device
- User context
- Performance
in real time
AI orchestration engine
Analyze·Predict·Decide·Optimize
Continuous learning
Optimized delivery
- Banners
- Pop-under
- Interstitials
- Video
- Native ads
Diagram: your ad partners feed the AI orchestration engine, which analyzes geography, currency, device, user context and real-time performance, then routes impressions to banners, pop-unders, interstitials, video and native ads, learning continuously.
Signals
What the model looks at.
Contextual signals only. Enough to price an impression precisely, nothing that identifies a person.
Geography
Country and region of the visit, the first driver of eCPM.
Device and OS
Desktop, tablet, Android or iOS, and the browser family.
Format and placement
Banner size, popunder, native slot or interstitial, and its position.
Time of day
Hour and weekday in the visitor's time zone, where demand peaks.
Currency
Payout currency and exchange rate, so yields compare net to net.
Traffic source
Organic, social, direct, referral or paid, each with its own buyers.
Session context
Page depth and ads already seen, to respect frequency and quality.
Partner performance
Live eCPM, fill rate and latency of every partner, minute by minute.
Why it earns more
A waterfall guesses.
The engine measures.
A waterfall calls partners in a fixed order and takes the first answer. BGrowth compares all of them on every request and keeps your rules in charge.
| Waterfall | BGrowth | |
|---|---|---|
| Partner order | Fixed, reviewed by hand | Recomputed for every request |
| Decision unit | Ad unit or country | Single impression |
| Latency | Adds up with each fallback | 83 ms median, capped by you |
| Experimentation | Manual A/B tests | Continuous, automatic |
| Explainability | Spreadsheets | Every decision logged |
Get started
Let the engine route
your next impression.
Connect your networks today and see your first optimized impressions within the hour.
Free 30-day trial · No setup fee · Cancel anytime
Estimate your uplift
Live estimateWith BGrowth, per month
$57,500 – $65,500
+15% to +31% estimated uplift