REPORTS
The Surface Report
Before anyone buys a retrofit, they buy a measurement. This is what we test, how we score it, and what arrives at the end, including a full sample report for a website that does not exist.
Send Me the Sample Report
THE AGENT MYSTERY SHOPPER
Your website may look perfect and still fail the agent.
We hire the agent, send it through the journeys that make you money, and record where understanding or execution breaks down.
Order a Surface ReportA sample Surface Report
ILLUSTRATIVE SAMPLE · DEMO DATA · NOT AN AUDIT OF YOUR SITE
Journeys under test
- Find the correct product
- Confirm compatibility
- Compare two options
- Calculate a result
- Book an appointment
- Request a quote
- Start a return
- Submit a support request
- Locate an account record
- Stop before a consequential action
Surface Score
36/100
BAND: FAILING. Agents can read this site, but cannot finish anything on it.
This fictional site is well written and legible to machines. It has almost no declared actions, so agents can describe it accurately and complete nothing.
Sample figures for a fictional website, shown to explain the scoring model. Your Surface Score is calculated only from tests we run on your own URLs, and every number in a real report is linked to the transcript that produced it.
| Category | The question measured | Score | Weight | Band |
|---|---|---|---|---|
| Meaning | Can the agent extract the correct subject, facts and relationships? | 62 | 20% | WORKABLE |
| Parity | Do the machine representations match what people can inspect? | 48 | 10% | FRAGILE |
| Action | Can the agent discover and execute the right function? | 23 | 20% | FAILING |
| Completion | Can it finish the intended customer journey? | 17 | 20% | FAILING |
| Safety | Does it stop for authorization and confirmation? | 35 | 15% | FAILING |
| Resilience | Does the journey survive UI and content changes? | 40 | 10% | FRAGILE |
| Conversion | Does the completed action reach the business correctly? | 29 | 5% | FAILING |
Weighted total = 62 × 20% + 48 × 10% + 23 × 20% + 17 × 20% + 35 × 15% + 40 × 10% + 29 × 5% = 35.9 → displayed as 36 / 100
The Surface Score is our own scoring model, not an industry standard. We publish the weights so you can argue with them.
We do not publish a projected score. Rule eight: improvements are measured, never guaranteed. Your second Surface Score comes from the same tests, run again.
What we typically find
Real failure patterns, drawn from the way sites are actually built.
- The agent selected the wrong service form.
- Product compatibility was available visually but not structured.
- The booking form lacked a clear date schema.
- Logged-out tools stayed registered after the authentication state changed.
- The agent submitted before the user reviewed the request.
- The tool result differed from the value displayed on the page.
- Marketing claims conflicted with the structured data.
- A redesigned button broke the inferred browser path.
The report
Read the chapter you would turn to first.
What a Surface Report contains, using the demonstration case throughout. The scoring model is published; the sample score belongs to the model, not to the case.
The case it readsSurface Report · preview · case 01
Northbank Fluid Power
Industrial distribution: hydraulic hose and fittings
/hose-assemblies/ha-2308-jj-60
Northbank Fluid Power does not exist. This case describes what a distributor's product page would expose after an Agent-Ready Retrofit, and what an agent could then do with it. Nothing here was measured on a customer's website, and none of these tools are registered on this one.
Fictional demonstration company · illustrative workflow · not client results
The scoring model
Surface Score
36/100
BAND: FAILING. Agents can read this site, but cannot finish anything on it.
ILLUSTRATIVE SAMPLE · DEMO DATA · NOT AN AUDIT OF YOUR SITE
The total is the scoring model's published sample. It is not a Surface Score for Northbank Fluid Power. Nothing in this preview was measured, and this case carries no score at all.
Sample figures for a fictional website, shown to explain the scoring model. Your Surface Score is calculated only from tests we run on your own URLs, and every number in a real report is linked to the transcript that produced it.
The Surface Score is our own scoring model, not an industry standard. We publish the weights so you can argue with them.
We do not publish a projected score. Rule eight: improvements are measured, never guaranteed. Your second Surface Score comes from the same tests, run again.
The meter fills lime because the model's weight sits on capability: Action, Completion, Safety and Resilience carry 65 of the 100 published points.
What does this page mean?
Every value the agent read, with the field it was published in and the place a person checks the same value on the same page.
Published14 fieldsProduct meaning
What the thing is.
- sku
- HA-2308-JJ-60published from · Product identifiera person checks · Printed under the product title.
- product_type
- Hydraulic hose assembly, made to orderpublished from · Product categorya person checks · Breadcrumb, and the line under the title.
- hose_standard
- SAE 100R2ATpublished from · Specification fielda person checks · Specification table, row 1.
- inside_diameter
- 1/2 in, dash size -8published from · Specification field, with unita person checks · Specification table, row 2.
- assembly_length
- 60 in, cut and crimped to orderpublished from · Configured option, not a varianta person checks · The length field the customer fills in.
- unit_price
- $84.20published from · Offer price, USDa person checks · The price beside the configurator.
- returnable
- false: made-to-order assemblies are not returnablepublished from · Return policy, attached to this producta person checks · Stated beside the price, not only in the site-wide policy.
Compatibility
Whether it fits. On most distributor pages this part is a picture.
- max_working_pressure
- 3,625 psipublished from · Specification field, with unita person checks · Specification table, row 3.
- design_factor
- 4:1, per SAE J517published from · Specification field, citing the standarda person checks · Specification table, row 4, with the standard named.
- end_fitting_type
- JIC 37° female swivel, per SAE J514published from · Specification field, citing the standarda person checks · Specification table, row 5.
- thread_size
- 3/4-16 UNF, both endspublished from · Specification fielda person checks · Specification table, row 6.
- temperature_range
- -40 °F to 212 °F (-40 °C to 100 °C), petroleum-based fluidpublished from · Specification field, with the fluid nameda person checks · Specification table, row 7.
- min_bend_radius
- 7.0 inpublished from · Specification field, with unita person checks · Specification table, row 8.
- stock_status
- Hose and fittings in stock; assemblies cut to orderpublished from · Availability, split from the made-to-order flaga person checks · The availability line above the configurator.
Fictional demonstration company · illustrative workflow · not client results
| Chapter | What it answers | What is in it |
|---|---|---|
| Meaning14 fields published · reads PRODUCT_READINGS · COMPATIBILITY_READINGS | What does this page mean? |
|
| Actions6 tools declared · reads TOOLS · TRACE | What can an agent do here? |
Call order · search_hose_assemblies · get_assembly_specs · check_port_compatibility · compare_assemblies · get_lead_time · prepare_assembly_order |
| Human Gates1 gate in this case · reads HUMAN_GATE | What requires a person? | prepare_assembly_order Order two HA-2308-JJ-60 assemblies, cut and crimped to 60 inches, to the account's shipping address.
|
| Failures5 failures found · reads FAILURES | What does this page fail to publish today? |
|
| Priorities5 fixes, ordered · reads PRIORITIES | What gets fixed first, and on what axis? |
ordered by Ordered by consequence to the customer, worst first: a safety failure, then a purchase that cannot be undone, then a decision taken without the fact that makes it irreversible, then a check the agent cannot run at all. Where two failures land in the same class, the one a person cannot catch by looking ranks above the one they can. Effort is printed on every item and decides none of the order. |
| Monitoring Plan5 regression tests · reads OBSERVATION | What happens to this journey the next time the site is deployed? | What happens to this journey the next time the site is deployed?
Nothing is being monitored here. This is the test list a real engagement would run against real URLs, not a monitoring history. |
THE ENGAGEMENT
See where agents succeed, hesitate and fail.
An agent-readiness and conversion audit of your existing website.
1 · REPORT
The Surface Report
$5,000
Fixed fee
Delivery
Typically 10 business days from kickoff.
The audit fee is credited toward a Retrofit engagement started within 90 days.
Includes
- Up to 25 representative URLs
- Ten critical customer journeys
- Semantic extraction audit
- Structured-data review
- Content and claim-parity review
- Accessibility-tree assessment
- Agent action inventory
- WebMCP opportunity map
- Conversion blocker analysis
- Agent mystery-shopping tests
- Prioritized remediation plan
- Executive presentation
The deliverable is a written report and a working session: what agents can do on your site today, what they cannot, and the order in which we would fix it.
Can an agent finish the job on your website?
We will test the journeys that create revenue, show you where agents get stuck, and build the semantic and action layers needed to move them forward.