A demo tenant that tells a story
A demo account full of uniform noise makes every chart flat and every story fall apart under the first question. "Why does this churned customer have a renewal date?" Declare the story the demo needs to tell and the data keeps it.
Who: sales engineers, product managers, anyone who demos a product with charts in it. Time: under 2 seconds.
The declaration#
Subscriptions move through a state machine: trial, active, past due,
churned. 60% are active and 26% churned, exactly. Payments come mostly from
a handful of large customers, and preset: demo moves every date so the
latest one is today.
# misata.yaml
name: demo_tenant
seed: 404
preset: demo
tables:
customers:
rows: 1200
columns:
customer_id: {type: int, unique: true, min: 1, max: 1200}
company: {type: text, text_type: company}
industry: {type: categorical, choices: [software, retail, health, finance, logistics], probabilities: [0.34, 0.22, 0.16, 0.16, 0.12]}
signed_up_at: {type: date, start: "2023-01-01", end: "2025-09-30"}
subscriptions:
rows: 1200
columns:
subscription_id: {type: int, unique: true, min: 1, max: 1200}
customer_id: {type: foreign_key, unique: true}
plan: {type: categorical, choices: [starter, growth, enterprise], probabilities: [0.55, 0.33, 0.12]}
status: {type: categorical, choices: [trialing, active, past_due, churned]}
started_at: {type: date, start: "2023-01-01", end: "2025-09-30", weekday_weights: [0.19, 0.2, 0.2, 0.18, 0.15, 0.04, 0.04]}
activated_at: {type: date, nullable: true}
past_due_at: {type: date, nullable: true}
churned_at: {type: date, nullable: true}
payments:
rows: 18000
columns:
payment_id: {type: int, unique: true, min: 1, max: 1000000}
customer_id: {type: foreign_key, sampling: pareto, alpha: 1.1}
amount: {type: float, distribution: lognormal, mean: 900, std: 2200, min: 40, max: 60000, decimals: 2}
paid_at: {type: datetime, start: "2023-01-01", end: "2025-10-20", weekday_weights: [0.17, 0.17, 0.17, 0.17, 0.16, 0.08, 0.08]}
relationships:
- customers.customer_id -> subscriptions.customer_id
- customers.customer_id -> payments.customer_id
lifecycles:
- name: subscription
table: subscriptions
state_column: status
start_column: started_at
initial: trialing
states:
- {name: trialing}
- {name: active, timestamp: activated_at}
- {name: past_due, timestamp: past_due_at}
- {name: churned, timestamp: churned_at, terminal: true}
transitions:
- [trialing, active]
- [trialing, churned]
- [active, past_due]
- [active, churned]
- [past_due, churned]
weights: {trialing: 0.07, active: 0.6, past_due: 0.07, churned: 0.26}Run it#
pip install "misata[duckdb]"
misata generate -c misata.yaml -o out --attestWhat the demo shows#
- A legal history on every subscription. A churned subscription that was active carries its activation date, then its churn date, in that order. A trial that churned has no activation date. Nobody has a churn date without having churned.
- Whales.
sampling: paretoon the foreign key concentrates payments: in this run the top 1% of paying customers bring 32.8% of revenue, and the median payment is 339 against a mean of 904. - Dates that end today. Signups run up to today, so "last 30 days" filters show something on the day of the demo.
- Weekly rhythm. Subscriptions start mostly on weekdays and payments
dip at weekends, as declared by
weekday_weights.
The check#
misata realism out -s misata.yaml
misata audit out -s misata.yamlRealism score 0.96 across 13 checks: 12 pass, 1 warn, 0 fail
Coherence: clean, no reader-visible contradictions.The realism report looks for the tells of generated data with no real data to compare against: flat weekdays, uniform amounts, every customer with the same number of orders, emails unrelated to names. The audit checks what a reader would notice: a churn date on an active subscription, a payment before its customer existed. The one warning is the weekday rhythm of the lifecycle's activation dates, which no declaration in this file shapes.
preset: demo makes the output depend on the day you run it. Drop the
preset when you need the same bytes every time.
Related#
- Use-case presets: what
preset: demochanges - Realism report: every check behind the 0.96 score
- Processes and event logs: when the history needs loops and dwell times
- Synthetic data for BI demos: more demo shapes
Files#
The recipe folder runs as it is. Its files on GitHub:
Prefer generating data visually?
Misata Studio lets you turn natural language prompts into relational schemas with exact outcome invariants, live preview tables, and one-click database seeding without writing code.