How do I set up an RTB marketplace for incremental reach expansion in prospecting campaigns?

To set up an RTB marketplace for incremental reach expansion in prospecting campaigns, you want a system that can identify new users who are likely to add reach beyond your current audience, then serve bids only when the expected incremental value exceeds the cost.

Here’s a practical setup framework.

---

1) Define the objective clearly

“Incremental reach” means:

• Finding users not already exposed to your campaigns or brand
• Prioritizing users who are incrementally reachable via RTB inventory
• Avoiding overbidding on audiences you would have reached anyway through cheaper channels

Your marketplace should optimize for:

• Unique reach
• New-to-brand conversions
• Incremental lift
• Cost per incremental reached user

---

2) Build the audience and measurement foundation

A. Seed audience

Start with:

• Converters
• High-intent site visitors
• CRM lists
• Engaged app users

Use these to model lookalikes or propensity scores.

B. Control groups

You need a way to estimate incrementality:

• Holdout audience not exposed to RTB
• Geo-level holdouts
• Time-based holdouts
• Ghost bids / PSA control where possible

C. Identity and reach mapping

To maximize incremental reach:

• Use identity graphs or household/device resolution where compliant
• Deduplicate across DSPs, SSPs, and channels
• Track frequency at person/household level if possible

---

3) Create an incrementality scoring model

For each impression opportunity, estimate:

[ Incremental\ Value = P(convert | impression) \times Incrementality\ Factor - Expected\ Cost ]

Where the incrementality factor captures whether the user is:

• New to the campaign
• Not heavily saturated elsewhere
• Outside current audience overlap
• Unlikely to convert via existing owned/paid channels

Useful signals

• Prior ad exposure count
• CRM match status
• Site/app visitation recency
• Cross-channel exposure history
• Geo/device/household uniqueness
• Past campaign overlap
• Contextual signals indicating new audience potential

Output

A score like:

• 0.0–1.0: incremental reach likelihood
• expected_new_user_value
• recommended_max_bid

---

4) Design the marketplace logic

Your RTB marketplace should function as a decision engine:

Inputs

• Bid request from SSP/exchange
• User and context features
• Audience membership / suppression rules
• Incrementality score
• Budget and pacing status
• Bid landscape / clearing price estimates

Decision rule

Bid only when:

[ Expected\ Incremental\ Value > Clearing\ Price + Margin ]

Example:

• User has high propensity
• Low prior exposure
• Outside retargeting pools
• High likelihood of being net-new
• Then bid aggressively

If the user is already saturated or likely reachable elsewhere:

• Bid low or skip

---

5) Set up audience suppression and exclusions

To ensure true incremental reach:

• Exclude recent converters
• Exclude heavy exposed users
• Exclude users reached in owned channels where possible
• Exclude internal employees, bots, fraud-prone inventory
• Exclude audiences already in your remarketing pools

This helps prevent waste and forces prospecting toward genuinely new users.

---

6) Inventory strategy

Focus on inventory that tends to deliver incremental users:

• Open exchange / RTB supply with broad reach
• Premium programmatic publishers with high unique users
• CTV, audio, native, and display for prospecting
• Contextual placements outside your typical remarketing environments

Use supply-path optimization to:

• Favor unique, high-quality supply
• Reduce duplicate reach across intermediaries
• Improve cost efficiency

---

7) Bid strategy for incremental reach

Use differentiated bidding:

High-incrementality segment

• Higher bids
• Looser frequency caps
• Broader inventory

Medium segment

• Moderate bids
• Standard frequency caps

Low-incrementality segment

• Low bids or no bid
• Strict suppression

A common strategy is to use:

• Base bid = expected conversion value
• Incrementality multiplier = based on novelty/uniqueness
• Supply quality adjustment = viewability, fraud risk, attention

Formula example: [ Bid = Base\ Value \times Incrementality\ Score \times Quality\ Score ]

---

8) Build feedback loops

Your marketplace should learn from outcomes:

• Impressions served
• Unique users reached
• Incremental conversions
• Frequency distribution
• Overlap with other campaigns
• Post-exposure lift

Then retrain:

• Propensity model
• Incrementality model
• Bid response model
• Supply quality model

Use this to automatically shift budget toward the sources of true incremental reach.

---

9) Measurement and KPIs

Track:

• Unique reach
• Incremental unique reach
• Cost per incremental user reached
• Incremental conversion rate
• Lift vs control
• Frequency to first conversion
• Audience overlap percentage
• Waste rate on saturated users

Good marketplace performance is not just cheap CPMs; it’s net-new reach at acceptable cost.

---

10) Operational architecture

A typical setup:

Data layer

• Event collection: bid requests, impressions, clicks, conversions
• Identity resolution
• Audience store
• Control/holdout assignment

Modeling layer

• Propensity model
• Incrementality model
• Supply quality model
• Bid optimization model

Decisioning layer

• Real-time bid scoring
• Suppression logic
• Budget pacing
• Fraud filters

Reporting layer

• Incremental reach dashboard
• Lift reporting
• Supply/path analysis
• Attribution views

---

11) Practical rollout plan

Phase 1: Baseline

• Set up tracking and audience suppression
• Launch with standard prospecting
• Measure current unique reach and overlap

Phase 2: Incrementality scoring

• Add novelty and exposure-history signals
• Create high/medium/low incrementality segments

Phase 3: Marketplace optimization

• Dynamic bids by incremental value
• Suppress low-value opportunities
• Shift spend to net-new inventory

Phase 4: Automated learning

• Retrain models weekly or daily
• Optimize by lift and cost per incremental reach

---

12) Important considerations

• Privacy/compliance: ensure consent, data minimization, and lawful processing
• Identity limitations: perfect person-level deduplication is rarely possible
• Attribution bias: use holdouts to avoid over-crediting RTB
• Fraud/viewability: incrementality is meaningless if inventory is low quality
• Frequency management: too much frequency reduces incremental reach

---

Simple operating rule

If the user is:

1. New or lightly exposed
2. Likely to convert
3. In quality inventory
4. Not reachable more efficiently elsewhere

…then bid.

Otherwise, suppress or bid conservatively.

---

If you want, I can also provide:

1. a reference architecture diagram,
2. a sample bidding formula, or
3. a data schema for the marketplace.