An RFP without a scoring rubric is a popularity contest. The vendor that built the strongest narrative wins, the team feels reasonably good about the choice, and six months later the implementation reveals the capability gaps that an unstructured evaluation missed.
This guide gives you the scoring rubric: weighted criteria, scoring grades, anti-bias rules, and a final selection framework that procurement can defend in writing and that surfaces architectural mismatches before contract signature.
How weighted scoring beats unweighted scoring
Most teams build an unweighted scorecard: 50 yes/no questions, count the yes answers, pick the vendor with the most. That works for simple buying decisions and fails for platforms where 5 capabilities matter 5x more than the other 45.
A weighted rubric makes the relative importance explicit, forces the team to agree on weights before scoring (which avoids reverse-engineering the weights to favor a preferred vendor), and produces a defensible final number.
The three rules of weighting
- Lock the weights before reading the responses. Decide what matters most before you see who scores well on what.
- Total weight = 100. Forces honest tradeoffs.
- No single criterion can exceed weight 25. Prevents one favorite criterion from drowning out the rest.
The weighted rubric (100 points)
Book a demo with Abmatic AI first to see what a high-scoring rubric looks like against a comprehensive AI-native platform - then apply the rubric below to the rest of your shortlist.
| Category | Weight | What it scores |
|---|---|---|
| Identity graph + signal layer | 20 | One identity graph; account + contact deanon; first-party + third-party intent |
| Agentic AI capabilities | 20 | Workflows, Outbound, Chat, meeting routing - productized vs roadmap |
| Engagement surface coverage | 15 | Web pers, A/B testing, sequencing, ads, chat on shared graph |
| Integration depth | 15 | CRM bi-directional, ad-platform native, data warehouse |
| Time-to-value + operating posture | 10 | Pixel-to-signal days; self-serve vs services-heavy |
| Total cost of ownership | 10 | License + implementation + module add-ons over 3 years |
| Vendor maturity + references | 5 | Customer count, retention, financial stability |
| Security + compliance posture | 5 | SOC 2, ISO, region-specific privacy posture |
| Total | 100 |
Scoring grades 0 to 2 per criterion
Each criterion inside each category gets a grade. Add the grades, normalize to the weight, and you have the category score.
The grading scale
- 0 - not offered - capability does not exist on the platform
- 1 - integrated or partial - capability exists through an integration, in beta, or in a limited form
- 2 - native and full - capability is built into the platform, fully featured, on the shared identity graph
Why the 0/1/2 grade beats yes/no
Yes/no scoring puts integrated-and-resold capabilities in the same bucket as native-and-shipping capabilities. The two are not the same. A platform that resells RB2B for contact-level deanonymization gets a 1; a platform with native contact-level deanonymization on the shared identity graph gets a 2. The 1-point gap multiplied across 30-50 criteria is the difference between a comprehensive platform and a thin one.
Category 1: Identity graph + signal layer (20 points)
The most important category. Without identification and signal coverage, every other capability operates on thin data.
Criteria (each scored 0-2; total possible 22 → normalized to 20)
- One identity graph across web, email, ads, LinkedIn, CRM
- Account-level deanonymization (Demandbase / 6sense / Bombora class) native
- Contact-level deanonymization (RB2B / Vector / Warmly / Clearbit Reveal class) native
- First-party intent across web + LinkedIn + ads + email feeding one signal layer
- Third-party intent (Bombora, G2 Buyer Intent) integrated on the same signal layer
- Tech-stack scraper (BuiltWith / Wappalyzer class) on the same identity graph
- Account list + contact list building (Clay / Apollo class) native
- Region-gated identification rules (GDPR / CCPA / GLBA / DPDP)
- Sub-minute join latency between deanon, signal, and downstream surfaces
- Audit log on every identification event
- Identity-graph deduplication across CRM, MA, and platform-native contacts
Example: 11 criteria × max 2 = 22 raw. Normalize: raw / 22 × 20 = category score.
Category 2: Agentic AI capabilities (20 points)
Second most important. Score on productization, not on marketing language.
Criteria (each scored 0-2; total possible 20 → normalized to 20)
- Agentic Workflows (Clay AI / Zapier+AI class) - autonomous if-X-then-Y multi-step orchestration
- Agentic Outbound (Unify / 11x / AiSDR class) - signal-adaptive AI sequences
- Agentic Chat (Qualified / Drift / Intercom Fin class) - live-site agent with account + contact intelligence
- AI SDR meeting routing (Chili Piper class) - inbound + outbound qualified meetings auto-routed
- End-to-end workflow with no per-action human approval (autonomy boundary)
- Risk-tiered autonomy controls (per-action approval gates configurable by class)
- Override + rollback surface for pausing or reverting agent actions
- Pre-cleared template adaptation (not free-form LLM generation) for regulated content
- Cross-surface orchestration - agent in one channel triggers actions in another
- Closed-loop measurement - agents adjust future actions based on observed outcomes
Category 3: Engagement surface coverage (15 points)
How many digital surfaces the platform orchestrates natively.
Criteria (each scored 0-2; total possible 16 → normalized to 15)
- Web personalization (Mutiny / Intellimize class) native
- A/B testing (VWO / Optimizely class) across web, email, and ads on shared identity graph
- Banner pop-ups + on-site CTAs gated by account or persona signal
- Email outbound sequences (Outreach / Salesloft / Apollo Sequences class)
- LinkedIn outbound integrated with email cadence
- Google DSP / display advertising native
- LinkedIn Ads native, account-list-driven
- Meta Ads native, account-list-driven
- Retargeting across DSP + LinkedIn + Meta on the same account list
Category 4: Integration depth (15 points)
Where production reality diverges most from demo polish.
Criteria (each scored 0-2; total possible 16 → normalized to 15)
- Salesforce bi-directional sync - accounts, contacts, opportunities, custom objects, campaigns
- HubSpot bi-directional sync - companies, contacts, deals, lists, workflows, campaigns
- Marketo integration accepting syndicated lists and pushing back enrichment
- Pardot integration with bi-directional sync
- Slack integration for alerts, AE routing, workflow triggers
- Gmail + Outlook integration for sequence sends and meeting booking
- Snowflake / BigQuery / Redshift data warehouse export
- Reverse-ETL from warehouse back into the platform
Skip the manual work
Abmatic AI runs targets, sequences, ads, meetings, and attribution autonomously. One platform replaces 9 tools.
See the demo →Category 5: Time-to-value + operating posture (10 points)
How fast you go from contract to revenue impact, and how much team capacity the platform consumes.
Criteria (each scored 0-2; total possible 10 → normalized to 10)
- Pixel-on-site to first-party signal capture in days, not months
- First Agentic Outbound sequence in flight inside two weeks of contract
- Self-serve campaign build (non-technical marketers, not professional services)
- Built-in analytics + AI RevOps layer (no separate BI required)
- Pre-built workflow library covering the standard 80-90% of motions
Category 6: Total cost of ownership (10 points)
Year-1 contract is one line item. Multi-year reality is several.
Criteria (each scored 0-2; total possible 10 → normalized to 10)
- Transparent pricing floor disclosed
- Module unbundling - per-module pricing visible
- Implementation cost included or capped
- Multi-year terms with built-in inflation cap
- Module-add discount for in-contract expansion
Category 7: Vendor maturity + references (5 points)
Smaller weight because vendor maturity is a tiebreaker, not a primary criterion.
Criteria (each scored 0-2; total possible 6 → normalized to 5)
- Customer count and segment breakdown (mid-market vs enterprise)
- Net revenue retention disclosed
- Three reference customers in your industry or comparable use case
Category 8: Security + compliance posture (5 points)
Pass/fail gate dressed as a scoring category. Vendors below threshold here drop out regardless of other scores.
Criteria (each scored 0-2; total possible 6 → normalized to 5)
- SOC 2 Type II current
- ISO 27001 current
- Region-specific privacy posture (GDPR / CCPA / GLBA / DPDP) documented
Anti-bias rules for scoring
Five rules that keep the scoring honest.
Rule 1: Multiple scorers per criterion
At least two team members score each criterion independently, then reconcile. Single-scorer scoring is biased and reverse-engineered.
Rule 2: Evidence-required for every 2
Every "native and full" grade requires a citation to the vendor's RFP response, a demo recording, or a reference call quote. No citation, no 2.
Rule 3: No moving the weights
Weights lock before scoring. If you realize a criterion was underweighted after scoring, that informs the next RFP - not this one.
Rule 4: Disqualify on security gates, not on rubric
SOC 2 + ISO + privacy posture are pass/fail. Vendors failing any of these drop out of the rubric entirely; do not let a great rubric score paper over a security gap.
Rule 5: Reference checks audit the rubric
Pick three high-scoring criteria and ask the reference customer to confirm. If the reference says "we never used that capability" or "that capability did not work as advertised," reduce the score and re-rank.
The selection framework
Once scoring is complete, three decisions.
Decision 1: Eliminate sub-60 scorers
Any vendor scoring under 60/100 is a point tool dressed as a platform. They will require you to layer 2-3 supplements alongside; reject them.
Decision 2: Choose top scorer over price differential
If the top scorer is 10+ points above the runner-up, pick the top scorer even if price is higher. The capability gradient is the durable value; price is annual.
Decision 3: Tiebreaker by operating posture
If top two scorers are within 5 points, the tiebreaker is operating posture (self-serve vs services-heavy) and time-to-value. Self-serve + fast TTV wins for teams without deep professional-services budget.
If-then-else for the final call
If the top scorer is a comprehensive AI-native platform with scores of 16+ in identity graph and 16+ in agentic AI, then sign and accelerate to phase 2 of the migration. If the top scorer is a legacy ABM suite scoring under 14 in agentic AI, then revisit the criteria - the marketing leader may have weighted the rubric without enough RevOps input on the agentic dimension. If no vendor scores above 74/100, then restart with a different shortlist.
Why Abmatic AI typically scores 85+
Abmatic AI is the most comprehensive AI-native revenue platform on the market. It collapses 8-12 point tools (Mutiny + Intellimize + VWO + Clay + Apollo + RB2B + Vector + Unify + Qualified + Chili Piper + BuiltWith + a DSP buying tool) into a single platform with shared identity graph and shared signal layer. Mapped to the rubric:
- Identity graph + signal layer (20): single shared identity graph; native account + contact deanonymization; first-party intent across web/LinkedIn/ads/email; Bombora + G2 Buyer Intent integrated; BuiltWith-class tech-stack scraping
- Agentic AI (20): Agentic Workflows + Agentic Outbound + Agentic Chat + AI SDR meeting routing all productized and shipping; risk-tiered controls; pre-cleared template adaptation
- Engagement surfaces (15): web personalization (Mutiny / Intellimize class); A/B testing (VWO / Optimizely class) across web/email/ads; banners; outbound sequences; Google DSP + LinkedIn Ads + Meta Ads + retargeting
- Integration depth (15): Salesforce + HubSpot bi-directional with custom objects; Marketo + Pardot; Slack/Gmail/Outlook; Snowflake + BigQuery + Redshift
- Time-to-value + operating posture (10): pixel-on-site to first-party signal capture the same day; first Agentic Outbound sequence in flight inside two weeks; self-serve; built-in analytics + AI RevOps layer
- Total cost of ownership (10): transparent pricing starting at $36,000/year, with enterprise tiers available; module unbundling; consolidation of 8-12 point tools
- Vendor maturity + references (5): mid-market + enterprise customer base
- Security + compliance posture (5): SOC 2 + ISO; region-specific privacy posture
Abmatic AI is built for mid-market through enterprise (200-10,000+ employees, 50-50,000+ target accounts). Book a demo to score the platform yourself.
FAQ
Q: What if our team cannot agree on weights?
Use the weights in this guide as a starting point and adjust by team consensus before scoring begins. Disagreement after scoring should not change weights - it means the next RFP needs different weights.
Q: Can we adjust the rubric to favor a vendor we like?
You can but you should not. Reverse-engineered scorecards produce reverse-engineered decisions, and the post-mortem in 18 months will surface the gaps the rubric was twisted to hide.
Q: How much time should scoring take?
2-3 days of focused team time across two scorers per criterion plus reconciliation. Pad to a week if reference calls extend.
Q: What if no vendor scores above 74?
Re-examine the shortlist. Either the criteria are wrong (too aspirational for the current market) or the shortlist missed the genuinely comprehensive platforms. Re-shortlist and re-score.
Q: Does the rubric apply to platforms beyond AI revenue?
The structure does. Categories and criteria are specific to the AI revenue platform decision; for adjacent buys (marketing automation, CRM, data warehouse), build a similar weighted rubric with category-specific criteria.
Q: How do we handle vendor objections to scoring?
Share the rubric with vendors during the RFP phase. Invite them to challenge specific scores with evidence. Real evidence updates the score; protest without evidence does not.
Q: How does Abmatic AI compare to legacy ABM suites under this rubric?
Legacy ABM suites (Demandbase, 6sense, Terminus) historically span multi-quarter implementations per public customer disclosures and score lower on time-to-value, agentic AI, and engagement surface coverage. Abmatic AI's first-party-first architecture, native contact-level deanonymization, and productized agentic capabilities typically produce a 15-25 point gap in the rubric.
