Decision 01
Sharpen positioning
See which buyer segments and priorities already align with the product, and where the answer becomes ambiguous.
Independent AI recommendation benchmarks
See which products AI recommends.
Measure which products enter AI shortlists, which one becomes the first choice, how the answer changes by buyer, and what sources shape the result. Starting with CRM.
Current public evidence: 600 accepted answers across 200 US CRM buyer prompts and three waves · observed September 8, 2026–September 14, 2026
Exhibit A · US CRMPublished replication
HubSpot led recommendation reach. Salesforce narrowly led first choice.
| CRM product | Recommended | First choice |
|---|---|---|
| HubSpot CRM | 86.3% | 23.5% |
| Salesforce Sales Cloud | 59.8% | 24.9% |
Recommendation uses 600 accepted answers; first choice uses 591 determinate answers. Explicit co-leaders split first-choice credit. Bars share a 0–100% scale.
September 8, 2026–September 14, 2026 · OpenAI Responses API · gpt-5.6-luna
200 prompts repeated in three waves; exploratory findings. Method crm-replication-2026-09-v1.
Read the findings and limitations →The CRM market answer
One overall leader. Several contextual winners.
The three-wave CRM replication found a broad leader, but recommendation reach and first choice diverged, while buying priorities still changed the winner. Benchmark Bureau maps those conditions instead of compressing them into one universal rank.
By company scale
Each slice contains 120 accepted answers from 40 prompts repeated in three waves. Estimates describe answer behavior, not product quality or buyer conversion.
What we measure
A mention is not a recommendation.
AI visibility becomes more useful when each narrowing step is measured separately. The hierarchy shows whether a brand was merely named, seriously considered, actively recommended, or selected first.
Signal 1 of 4
Mention
Did the product appear in the answer at all?
Signal 2 of 4
Shortlist
Did it enter the set the buyer was asked to consider?
Signal 3 of 4
Recommendation
Did the answer actively recommend it?
Signal 4 of 4
First choice
Was it the explicit leader after co-leaders were split?
Rates count eligible answers. Several products can appear in one answer, so product recommendation rates may sum above 100%. Every release states its denominator, configuration, and observation window.
Why recurrence matters
A snapshot is evidence. A series becomes intelligence.
A credible market signal needs to show what persisted, what changed, and what remains uncertain. The September release repeats the same 200 prompts across three timed waves and reports matched changes from August.
Current release · September 2026
Three-wave CRM replication
Recommendation and first-choice uncertainty, buyer context, sources, historical change, and wave stability.
Open the dated release →Baseline · August 2026
Initial CRM market map
The matched starting point for measuring what changed and what persisted.
Open the August release →Gated next step
Provider comparison
Added only after matched configuration and customer-decision value are established.
For CRM teams
Turn the market answer into a better business question.
Benchmark Bureau is designed for product marketing, competitive intelligence, brand, growth, and leadership teams that need evidence, not a promise to manipulate a ranking.
Decision 02
Track the real shortlist
Measure which competitors appear in the same consideration sets instead of watching isolated brand mentions.
Decision 03
Explain first choice
Separate broad recommendation reach from the prompts where another product becomes the explicit leader.
Decision 04
Find evidence gaps
Inspect the source domains and evidence patterns present when a market answer is assembled.
Decision 05
Brief leadership
Bring a dated, scoped, downloadable market signal into planning without presenting anecdotes as market share.
Why trust it
Independent by design. Inspectable by default.
The method travels with the result
Prompts, model configuration, dates, denominators, scoring rules, and limitations remain beside every finding.
Public rank is never sold
Commercial work cannot change inclusion, prompts, scoring, findings, or rank.
The underlying evidence stays inspectable
Public reports link to HTML tables, exact scope, and downloadable data rather than unsupported scorecards.
Uncertainty remains visible
A single provider, small segment, unstable result, or missing comparison is labeled instead of generalized away.
Public record
Start with the evidence, then follow the history.
Interactive decision view
CRM AI Recommendation Explorer
Select a product, competitor, company scale, and buying priority to build a shareable evidence snapshot.
Build a decision view →Dataset
Released CRM evidence
Six public files with product, segment, prompt, source, and event data.
Open the dataset record →Research archive
Utah personal injury
The founding study demonstrates the method in a location-sensitive market.
Open the archived study →Follow the evidence
Know when the market answer changes.
Follow the public release feed for new benchmark records, or request a simple email notice when a material CRM release is published. No newsletter list is implied.
CRM AI recommendation intelligence
See where your product enters the market answer, where it loses first choice, and what evidence surrounds the decision.
Private work never changes public inclusion, scoring, or rank.