Listing Alignment Grader¶

Beta

Listing Alignment Grader is a Beta capability (is_beta = true). It's stable enough to act on, but dimensions, weights, and wording may change as the proof model matures.

Capability slug: listing_alignment_grader Plan Tier: Command Backend: One external website read + a single Bedrock (Claude Sonnet) judge call Estimated credits: 12

Overview¶

A buyer researching your product — or their AI assistant doing it for them — doesn't just read your AWS Marketplace listing. They cross-check it against your website, third-party review sites, and any content AWS has co-published. When those sources disagree with your listing, you look unproven, and assistants reach for a rival they can corroborate.

The Listing Alignment Grader runs that cross-check for you before you publish. It grades how well your external proof aligns with what your listing claims, against your stated GTM objectives, and turns every shortfall into a concrete next action.

It is the pre-publish counterpart to the Discovery Dashboard's Live Mirror:

Live Mirror — "Do AI assistants surface me?" (what buyers see now)
Alignment Grader — "Is there consistent, credible proof for them to surface?" (whether that proof exists)

What it grades¶

Four dimensions, scored 0–100 each and combined into a weighted overall score:

1. Website alignment¶

Does your company website clearly present this offering, and is its positioning consistent with the listing? A resilient external read pulls your site text, and a single Bedrock judge call scores alignment strictly from that text — it degrades honestly when there's no website on file or the site can't be read.

2. Third-party validation¶

Does credible third-party proof corroborate the offer — G2, Capterra, TrustRadius, Trustpilot, Gartner, named customer logos, quantified testimonials, awards, case studies? The same judge call lists only the signals actually present and scores by strength and specificity.

3. AWS co-authored content¶

Does first-party AWS proof exist — joint blog drafts or Partner Central case studies? This is read from internal records (no external call): 0 pieces = none, 1 = thin, 2+ = established.

4. Runnable proof¶

Can a buyer — or a buyer's AI agent — actually try the offering without talking to someone first? The judge looks for self-serve surfaces on your site: free trial, sandbox, interactive demo, public API docs, a playground. Proof that exists but sits behind "book a demo" is scored as gated and caps low by design.

The dimension resolves to one of four states:

State	Meaning
`runnable`	Self-serve proof a buyer can exercise on their own (score ≥ 70)
`gated`	Proof exists, but only after a meeting or a form
`thin`	Something is there, but it barely demonstrates the offering
`none`	No demo, trial, or sandbox found
`unknown`	No readable website to check — the website objective owns this gap, so no demoability nag is raised

Peer benchmark¶

A score on its own is hard to judge, so this dimension also reports how often companies like you expose self-serve proof at all:

Category	Peers exposing self-serve proof	Sample
AI Agents & Tools	72%	36
SaaS	56%	172
Professional Services	26%	31

AMI and Container have no cohort of their own and borrow the SaaS figure — the payload flags that with inherited: true.

These come from the AWS Registry of Open Data — a Common Crawl index read of whether a company's public site exposes a trial, sign-up, or get-started path. Two honest limits, which the product repeats wherever the number appears:

It's a peer adoption rate, not a percentile. There's no "you're in the 40th percentile of alignment scores" here — that statistic doesn't exist and we don't invent one.
The cohort is a curated set of well-known companies, not a random sample, so it skews toward larger and surviving firms. Read it as "peers", not "the market".

No benchmark line is shown when your site couldn't be read — there's nothing to compare against.

Category-aware weights¶

Each AWS Marketplace category proves itself differently, so the four dimensions carry different weight per category. Each profile sums to 100%.

Category	Website	Third-party validation	AWS co-authored content	Runnable proof
SaaS (baseline)	35%	25%	25%	15%
AI Agents & Tools	35%	15%	25%	25%
AMI	40%	20%	25%	15%
Container	40%	20%	25%	15%
Professional Services	35%	40%	20%	5%

The category keys mirror the Listing Optimizer's ListingScoreWeights, so a listing resolves its category once and both tools weight consistently. The rationale tracks the discovery physics: agent-native products lean on website + AWS technical proof (young domains, thin reviews), while professional services lean on third-party trust because humans procure.

Runnable proof carries its highest weight for AI Agents & Tools — an agent buyer transacts with what it can exercise, not with a sales narrative. It carries its lowest weight (5%) for Professional Services, which sits below the objective floor described next, so services partners are never told to add a demo for a relationship-led offering.

What you get back¶

Field	Meaning
`overall_alignment_score`	0–100, the weighted composite
`status`	`launch_ready` (≥85, no high-priority gaps) · `mostly_aligned` (≥70) · `needs_proof` (≥40) · `unproven` (<40)
`dimension_scores`	Per-dimension score, weight, and weighted contribution
`objectives`	Prioritized list of gaps, each with why it matters and a CTA
`summary`	One-line, publish-aware verdict

Objectives — every gap becomes a next action¶

Each dimension that comes up short (below the ready threshold of 70) becomes an objective, ranked by priority and expected score gain. Where a Vellocity capability closes the gap, the objective carries its slug and route so the dashboard renders a button:

Gap	Objective	Wired capability
Website weak / missing	Make your website tell the same story as your listing	`enrich_brand_voice`
No third-party proof	Earn third-party validation buyers trust (set up a G2 profile)	(external — no in-app CTA)
No AWS co-content	Co-author proof content with AWS	`aws_blog_co_author`
No runnable proof	Add runnable proof buyers and their agents can try	`marketplace_listing_optimizer`
Proof gated behind a meeting	Ungate your proof — give buyers something to try without a meeting	`marketplace_listing_optimizer`

The objective floor — no demo nag for services

A dimension only raises an objective when it carries at least 10% weight for that category. Professional Services weights runnable proof at 5%, so it is scored and shown but never turned into a "add a demo" objective — the grader respects that services are procured through relationships, not sandboxes.

Required & optional inputs¶

Required: brand. Optional: offering, website_url, category (and category_key to select the weight profile), user_id, listing_published.

No listing URL? It falls back to your Brand Voice

When the listing carries no website URL, the grader falls back to the company website on file in your Brand Voice profile (normalizing the scheme). Website and runnable proof used to score 0 in that case — a "no website on file" verdict for partners whose site was sitting in Brand Voice all along.

On-demand and cached — never on page load

Like the Live Mirror, a grade involves an external read plus a model call (~seconds). It is designed to run on demand and be cached, not on every page load.

Plan availability¶

Capability	Starter	Accelerate	Command
`listing_alignment_grader` (Beta)	—	—	✅