Buyer readiness
Media buying RFP evidence checklist
A media RFP should make weak evidence harder to hide. Before a proposal is scored, ask each seller to show the context being bought, the inventory boundaries, the reporting fields, and the measurement language the campaign can actually support.
A common problem appears late in the buy: three proposals look polished, each promises a smart audience, and the cheapest package seems easiest to approve. Then the buyer realizes that one seller cannot name the eligible pages, another cannot export the fields used in the pitch, and the strongest performance claim has no comparison design behind it. The RFP has to catch that gap before price becomes the only visible difference.
This checklist is for marketers, agencies, analysts, and publisher teams that need cleaner proposal comparisons. It works for private marketplace, sponsorship, newsletter, native, display, and contextual packages where the buying decision depends on reader intent and credible campaign evidence.
Start with the decision
The RFP should name the decision the campaign result will inform. A proposal for a renewal decision needs different proof than a proposal for a brand study, lead-quality test, or measurement-method comparison.
Use plain language: what will the buyer do differently if this campaign works? That answer determines whether the proposal needs descriptive delivery proof, directional response evidence, or a designed test that can support causal language.
| Decision the RFP supports | Ask sellers to show | Do not let the proposal substitute |
|---|---|---|
| Choose a contextual package | Included desks, URL patterns, placement IDs, package ID, exclusions, and reader job. | Broad audience claims without page-level scope. |
| Compare qualified traffic | Landing-page fit, URL parameters, session-quality fields, invalid-traffic review, and device mix. | Click volume without traffic-quality definitions. |
| Test lead quality | Source trail, lead status fields, duplicate rules, disqualification reasons, and follow-up timing. | Form-fill count as the only success measure. |
| Evaluate brand movement | Survey population, exposed and control definitions, question wording, field dates, and uncertainty. | A recall or awareness number without sample context. |
| Estimate incrementality | Holdout, matched market, assignment rule, leakage check, outcome window, and readout threshold. | Attribution reporting presented as a causal test. |
Minimum evidence request
Every strong answer should move through the same evidence chain: the seller states a claim, defines the inventory boundary behind that claim, shows the fields that will survive launch and reporting, then limits the final readout language to what the design can prove.
1. Reader contextAsk for the reader job the inventory supports: choosing a method, auditing a claim, reviewing a vendor, planning a campaign, or studying a failure mode. Context should be tied to URLs or page groups, not only to a content category.
2. Inventory boundariesRequire package ID, placement IDs, accepted creative sizes, device rules, page-set inclusion, page-set exclusion, geography, frequency rules where available, and any run-of-site fallback language.
3. Creative and destination fitAsk sellers to identify the offer type, destination type, creative sizes, message angle, landing-page requirement, and any mismatch that could weaken response or lead quality.
4. Measurement fieldsRequire delivery, exposure quality, traffic quality, creative, destination, context, lead, matchback, and test-cell fields before launch. If the seller cannot export a field, the final report should not depend on it.
5. Evidence levelAsk the seller to label the proposed readout as descriptive, directional, or causal. The label should match the comparison design, not the strongest metric in the dashboard.
Proposal questions
These questions make proposals easier to compare because each answer maps to a reporting row, a trafficking requirement, or a claim boundary.
| RFP question | Useful answer | Weak answer |
|---|---|---|
| Which exact contexts are included? | Named desk, URL pattern, article group, package ID, and excluded inventory. | General editorial category with no page-set boundary. |
| How will delivery be reported? | Impressions, spend, viewability, device, creative, placement, and page group. | Total impressions and clicks only. |
| How will traffic quality be separated? | Engaged sessions, bounce, return visits, destination action, UTM structure, and invalid-traffic review. | Click-through rate as the only traffic signal. |
| How will lead quality be joined? | Campaign source trail, lead status, duplicate rule, qualification status, and disqualification reason. | Matched lead count without quality fields. |
| What comparison is available? | Holdout, matched context, prior flight, matched market, or explicit no-comparison status. | Before-and-after trend with no concurrent context. |
| What claim language will the report use? | Pre-agreed wording for delivery, observed response, directional lift, or measured lift. | A promise to report the strongest result after the flight. |
Scoring rubric
Score proposals on evidence readiness before negotiating price. A cheaper package can be more expensive if the final report cannot separate delivery quality, context fit, and outcome quality.
| Score | What it means | Buying implication |
|---|---|---|
| 3 - ready | Context, inventory, creative, source-trail fields, comparison rule, and readout language are specific before launch. | Ready for a normal proposal comparison. |
| 2 - usable | Most fields are specific, but the comparison or downstream outcome join is limited. | Buy for delivery or directional learning, not causal proof. |
| 1 - descriptive only | The package can report delivery and basic response, but context or outcome fields are incomplete. | Use only for low-risk learning or awareness delivery. |
| 0 - unclear | The proposal relies on broad audience claims, pooled reporting, or performance language without inventory and measurement detail. | Request a revised evidence pack before scoring price. |
Convert the score into an award rule
The scoring step should end with a buying action, not a vague confidence note. Treat the score as a guardrail for what the buyer is allowed to buy, test, revise, or reject.
| Evidence result | What the buyer can do | Readout language to allow | Next request |
|---|---|---|---|
| Ready across context, inventory, fields, and comparison. | Score price, negotiate terms, and prepare the launch brief. | Use the agreed evidence level: descriptive, directional, or causal. | Ask for the sample report shell and field owner list. |
| Strong package, weak downstream outcome join. | Buy only if the decision is about delivery, traffic quality, or learning. | Use descriptive or directional language, not revenue causality. | Ask for lead-quality or CRM join rules before renewal claims. |
| Clear context, unclear comparison. | Use as a pilot or reserve budget for a designed follow-up test. | Say observed response, not lift, incrementality, or caused. | Ask for holdout, matched context, or baseline options. |
| Attractive price, missing inventory or reporting fields. | Do not score price yet. | No performance claim should be accepted. | Request a revised evidence pack with exportable rows. |
| Broad audience promise with no page or placement boundary. | Reject or rewrite the RFP response before award. | No context-fit claim should be used. | Ask for eligible URLs, placement IDs, exclusions, and fallback rules. |
Red flags to resolve before award
| Red flag | Why it changes the decision | Ask for |
|---|---|---|
| The proposal sells a context but reports only total delivery. | The buyer cannot tell whether the purchased environment was actually delivered. | Reporting by desk, page group, placement ID, and device. |
| The proposal promises performance without a comparison. | Attribution can count demand that would have happened anyway. | A descriptive label, a matched baseline, or a designed test. |
| The proposal treats all leads as equal. | Low-quality form fills can make a weak campaign look efficient. | Qualification status, duplicates, disqualification reason, and follow-up timing. |
| The proposal mixes creative versions without a rotation plan. | The final readout may confuse message response with delivery imbalance. | Creative ID, active dates, delivery by creative, and minimum cell size. |
| The proposal cannot export the fields it uses to sell the package. | The campaign thesis cannot be checked after launch. | A sample report shell before the buy is awarded. |
RFP language to reuse
Use plain requirements that make the evidence boundary visible before the seller writes a success story.
- List all included and excluded inventory, including URL patterns, placement IDs, device classes, and fallback rules.
- Identify the reader job this package reaches and the creative or destination type that fits that job.
- Provide a sample report showing delivery, context, creative, traffic quality, lead quality, comparison rule, and recommended action.
- Label the proposed readout as descriptive, directional, or causal and state the comparison design behind that label.
- Define the decision threshold that would support renewal, creative change, budget shift, deeper testing, or a stop decision.
Pair with
Use this RFP checklist with the contextual package proof checklist before scoring a package, the contextual package proof sheet when proposal evidence needs to become a one-page handoff, the contextual campaign brief template once the buy is selected, the campaign data-layer spec for fields that must survive trafficking and analytics export, the private marketplace readout export sample for row-level package, placement, device, creative, traffic-quality, and outcome-status fields, the campaign reporting sample for the proposed report shell, the private marketplace campaign measurement checklist for readout boundaries, and the campaign readout QA checklist after results arrive.