AdKernel DSP Bid Optimization Modes: Technical Overview (2026)

Modern programmatic isn’t just about “bidding higher” or “blocking more”—it’s about systematically learning what works, enforcing cost controls, and reallocating budget toward segments that prove value.

AdKernel’s Bid Optimization Mode is designed to do exactly that: it replaces initial bids with mathematically optimized values based on traffic segment performance and learning confidence.

Overview: exploration vs. exploitation

Bid Optimization Mode uses an exploitation vs. exploration strategy:

• In early learning, it can produce optimistic predictions to discover high-performing pockets of traffic.
• As segments mature, the optimizer transitions to realistic, data-driven predictions based on measured performance.

Core optimization logic (segment-based)

The optimizer operates at a traffic segment level. Each segment receives bid adjustments relative to its learning maturity:

• Well-learned segments get bids closer to the fully optimized value.
• New segments start with more optimistic predictions to enable discovery.
• Underperforming segments receive minimum viable bids (effectively removing them from budget allocation).
• Promising segments receive increased allocation as low-performers are filtered out.

The result is less waste during exploration and more efficient budget concentration as the campaign learns.

Bid prediction engine: what the optimizer uses

Optimized bid prices are computed from three primary inputs:

1. Event likelihood probability (statistical prediction a conversion event will occur)
2. Target cost parameters (your cost thresholds for the chosen conversion step)
3. Base eCPM (initial bid value used during insufficient-data periods)

Traffic segmentation parameters

By default, AdKernel analyzes performance across multiple dimensions to create granular segments.

Default parameters

• Campaign
• Creative
• Country / State
• OS
• Browser / Browser Version
• Exchange
• Domain/Bundle
• Site ID/App ID
• Tag ID
• Publisher

Customizable parameters

Bid models can be extended to include:

• Content categories (category-based optimization)
• Internal data segments (proprietary segment ID integration)
• Custom attributes (client-specific performance indicators)

Custom parameter work typically requires coordination with the technical implementation team.

Configuration options (what matters now)

Removed legacy settings

• CTR Max is no longer required (handled automatically).
• VCR Max is deprecated in favor of dynamic threshold management.

Current required settings

Target cost configuration

AdKernel supports a single target cost setup:

• Specify a target cost only for the chosen conversion step.
• No need to define intermediate-step target costs.

Base eCPM requirement (for optimizer-driven modes)

When using Target CPC, Target CPA, or Target CPCV, the Base eCPM field is required:

• It’s the starting bid during the exploration phase.
• It remains critical until the optimizer transitions to segment-driven bidding.

Budget management options

• Increase bids on underspend: improves budget utilization when delivery is below plan.
• Decrease bids on overspend: smart bid minimization for spend control.

Campaign setup: recommended procedure

1. Select the bid optimization type:
   • Target CPC (Cost Per Click)
   • Target CPA (Cost Per Acquisition)
   • Target CPCV (Cost Per Completed View)
2. Set Base eCPM:
   • Use a starting value that enables exploration.
   • Higher values can broaden initial testing.
3. Configure the Target Cost for your primary objective.
4. Enable budget optimization features:
   • Turn on “Increase bids on underspend”
   • Configure “Decrease bids on overspend” thresholds

Exploratory phase: what to expect

During initial learning:

• Base eCPM drives early bidding.
• The optimizer collects segment data across the full traffic space.
• Bid adjustments increase as confidence rises.
• Budget allocation shifts toward proven segments over time.

Reset optimization (when you make major changes)

AdKernel supports a Reset Optimization action to clear learning data and restart the learning process.

• Location: Campaign list page → Actions → Three dots menu
• Use case: major campaign changes that invalidate prior learning

Technical notes

To get the most from Bid Optimization Mode, ensure:

• Sufficient impression volume for segment analysis
• Conversion tracking is implemented correctly
• Event attribution is configured and validated

For monitoring, teams typically review:

• Real-time bid adjustment behavior
• Segment performance analytics
• Budget allocation reports

Best practices (fast start, clean learning)

• Start with a conservative Base eCPM for initial testing.
• Allow 48–72 hours for initial data collection before making major changes.
• Revisit target costs after early performance indicators appear.
• Use the reset function strategically for major pivots.
• Use underspend/overspend features to align delivery with objectives.

---

If you’d like help tuning Base eCPM and target costs for your vertical, request a demo or reach out to your account manager.