Hiring manager view
Start with the 60-second case study →Interactive tool · Synthetic data
Revenue Orchestration Lab
The same evidence should not always produce the same action.
VantageGrid sells cloud-operations software. Its revenue process still reads a content download as commercial priority and a production deployment as noise. This lab runs twelve situations through a model that keeps evidence, entity state, readiness and permission to act as separate things — and shows what changes when they stop being the same field.
People engage. Accounts buy.
A person's behaviour, an account's position, a product relationship and an opportunity are four different states. Atlas Logistics can be an Observe customer, a Control trialist and a closed-lost Control prospect at once, and collapsing those into one lifecycle field is how a system starts giving confident wrong answers.
Evidence is not inference.
A title string suggesting someone is an economic buyer is not the same fact as a human confirming it. Inferences carry confidence and can be withdrawn. Repeating an inference — even with a higher vendor score — produces the same evidence again, not a verification.
Readiness is not permission.
An account can be entirely ready to buy and entirely off limits, at the same moment, for good reasons. A hard block removes permission to contact. It does not remove the commercial case, and it does not delete the evidence that produced it.
VantageGrid, its two products, every account, every person and every event in this lab are invented. Nothing here is a real customer, a real outcome or a real performance claim.
Scenario controls
The failure it exposes
The case the legacy rule misses entirely: a healthy Observe customer validating Control in production, with a confirmed budget owner engaged on the business case.
Watch for
Both validation axes reaching meaningful, and the owner set becoming a coordinated group rather than a handoff.
Timeline
Step 1 of 7 · Day 1 · Atlas Logistics
Stepping forward replays the scenario from the beginning to the chosen event and re-derives every value on the page. Nothing is remembered between steps, so any step can be reached from any other and produce the same result.
S03 Expansion-ready customer. Step 1 of 7, day 1. Control trial started from inside Observe. Readiness: product interest. Permission to act: allowed. Next action owned by Marketing / Growth.
What changed
Day 1. Every dimension below is compared against the previous step.
Control relationship
Prospect changed to Trialing
Technical validation
None changed to Emerging
Buying-group coverage
Absent changed to Emerging
Trial momentum
No product usage changed to Active
Engagement breadth
0 known people changed to 1 known person
Commercial readiness
Not warranted changed to Product interest
Permission to act
No action warranted changed to Allowed
Orchestration
Evidence exists. No meaningful validation on either axis.
No constraint applies. The action can execute under existing rules.
No hard block, no competing owner, and no warning that raises the bar.
Next action
Hold in Control nurture and keep building account context. Do not occupy a seller with this yet.
Nurture
Evidence exists, but neither technical nor commercial validation is meaningful yet. Missing: Economic or business authority.
Recipients
- Jonah Reyes
Owner
Marketing / Growth
Anonymous-to-known conversion, nurture, air cover, account engagement.
Participating
- Product / PLG
Evidence timeline
Every event the system has seen, in order, with its source and whether it was observed or inferred. Inferences carry a confidence and can be withdrawn; observations cannot be upgraded by repetition.
- 01Day 1Product telemetryObservedCurrent step
Control trial started from inside Observe
Jonah Reyes launches a Control workspace from the Observe console. Existing customer, existing tenant, no new contract.
Associated entities: Jonah Reyes · Control
Entity and assessment state
Six entity axes and seven assessments, each held separately. Any of them can move while the others stand still, which is the whole reason they are not one field.
Account
- Name
- Atlas Logistics
- Relationship
- Customer
- Fit
- Icp strongRecorded context. No orchestration rule reads it.
- Engagement breadth
- 1 known person with recorded activityAccount-wide. Activity on any product counts here.
- In the Control motion
- 1 personOnly these can hold a seat, or be targeted, for this motion.
Product relationships
Observe
CustomerMonitoring and observability. The installed base.
Control
TrialingAutomation and remediation. Sold as an expansion onto Observe.
- Workspace createdDay 1 · Jonah Reyes
People
Jonah Reyes
Director, Platform Engineering
Technical leaderInferredParticipating in: Control
Sofia Marchetti
VP Infrastructure
Unknown roleInferred
Buying group
- Technical seat
- Jonah Reyes
- Authority seat
- Empty
- Missing
- Economic or business authority
Opportunity and customer context
- Opportunity — Control motion
- No Control opportunity on record
- Customer health
- Healthy
- Support escalation
- None open
- CSM commercial hold
- None
- Routing
- Not routed
Derived assessments
- Technical validation
- Emerging
Setup milestones only. Nothing production-grade has been connected or executed.
- Commercial validation
- None
Nobody at the account has engaged Control pricing, business-case, legal or procurement material.
- Signal confidence
- High
Nothing is degrading the reading: first-party product evidence and no unresolved data-quality conditions. Confidence measures the absence of known distortions, not the strength of the evidence.
- Product momentum
- Active
Decision trace
The same chain runs for every scenario. Reading it top to bottom is reading the reason for the action.
Event
Day 1 — Control trial started from inside Observe
Evidence
product telemetry, observed
Entity state
Control: trialing · buying group emerging
Assessment
Technical emerging · commercial none · readiness product interest
Constraint
None
Permission to act
Allowed
Action
Marketing / Growth — Hold in Control nurture and keep building account context. Do not occupy a seller with this yet.
The chain, in general
The trace on the left is this account at this step. The chain below is the shape every account runs through.
- 01
Event
Something happens. A form, a login, a policy execution, a support ticket, an opt-out.
- 02
Evidence
The event is recorded with a source and a basis: observed, or inferred with a confidence.
- 03
Entity state
Person, account, product relationship, buying group, opportunity and customer context each move on their own axis.
- 04
Assessment
Technical validation, commercial validation, buying-group coverage, signal confidence and readiness are derived separately.
- 05
Constraint
Hard blocks, existing-owner precedence and warnings are evaluated against what the system wants to do.
- 06
Eligibility
Permission to act is resolved. It is a different question from readiness and can move the opposite way.
- 07
Orchestration
An action, an owner, participating teams, and the reason the engine chose them.
VantageGrid’s legacy qualification rule
VantageGrid scores marketing engagements and routes at 60 points. This is that company’s own rule, run on the same events — not a claim that marketing engagement or lead scoring is invalid in general.
Legacy rule
0 / 60
Below threshold — hold in nurture
Scored at zero by this rule
- Product Telemetry × 1
Orchestration model
Product interestAllowed
Hold in Control nurture and keep building account context. Do not occupy a seller with this yet.
Divergence
Neither model acts, for different reasons. The legacy rule is below threshold at 0 points; the orchestration model reads product interest.
Where automation stops
This lab makes no model call. Every value on this page is produced by deterministic rules over a fixed fixture, which is why the same scenario at the same step is identical every time. The boundary below is what the architecture would delegate if a model were wired in.
No model call is made by this page, and no value on it was generated by one.
Deterministic automation
Rules, not intelligence. Running today.
- Routing from explicit rules
- SLA timers and breach detection
- Suppression and hard-block enforcement
- State transitions with defined preconditions
Assistive, low consequence
A model could do these. A wrong answer is cheap and visible.
- Summarise account activity for an owner
- Normalise a job title into a tentative role category
- Identify which buying-group seats are empty
- Explain why a deterministic decision was reached
Human approval required
A wrong answer here costs an account, a contract or a compliance position.
- Personalised outbound execution
- Material lifecycle or status override under ambiguity
- Overriding a governance veto
- High-consequence exception handling
What this model does not infer
- Purchase intent from third-party intent alone. Intent aims a motion; it does not start one.
- A verified role from a repeated inference, however high the vendor's own confidence climbs.
- Technical validation from marketing behaviour, or commercial validation from telemetry. The inputs are structurally separate, not filtered.
- A resolved identity from name similarity. Reconciliation is an action someone takes, and the lab shows the state before and after it.
- Anything about a real company. There is no data source behind this page.
Design tradeoffs
Deterministic fixtures over a live model
Every reviewer sees identical output, and every rule is inspectable. The cost is that the lab cannot demonstrate handling genuinely novel input — it demonstrates the decision architecture, not a model's judgement.
Coarse levels over continuous scores
Validation is none, emerging or meaningful rather than 0-100. Coarse levels force the rule that separates them to be stated. A score would hide the same rule behind a number that looks more precise than it is.
Scoped blocks over a global kill switch
A hard block forbids a class of action, not the account. It costs more to model, and it is the difference between suppressing outreach and destroying the intelligence that would have justified it later.
One account-level motion, not a portfolio view
This is an inspection lab for a single account's decision chain. The executive question — where is commercial velocity breaking down across the business — needs aggregation this version deliberately does not build.
Known limits of this version
- Twelve fixed scenarios. There is no free-form input, no way to author an account, and no persistence.
- The legacy comparison is VantageGrid's own scoring rule, invented for this lab. It is not a claim about marketing qualification in general.
- Momentum, confidence and coverage use small integer thresholds chosen to make the mechanics legible, not calibrated against any real dataset.
- The clock advances only when an event fires. Trial momentum decays at the next recorded event rather than continuously, which is what makes a step reproducible; a production implementation would run it on a timer.
- Account fit is recorded and displayed but no rule reads it. It is context here, not an input to routing.
- Identity reconciliation is triggered by an event in the fixture. There is no matching service behind it.
VantageGrid, its two products, every account, every person and every event in this lab are invented. Nothing here is a real customer, a real outcome or a real performance claim.
Synthetic scenario dataVantageGrid is a fictional company. No real customer data is used anywhere on this page.