Product details
What you're looking at, and the four-step wizard that creates one
The mockup above is a project running a single scheduled job — a daily comparison on the ANALYTICS plan, set to auto-apply drift within a row limit. The table at the bottom is the receipt book: ten nights of automation runs, all completed clean. Below, the cards walk through the active row, the AUTO-APPLY semantics, the Recent automation runs feed, and each of the four wizard steps that creates a new schedule.
The active schedule, read top-down
The single row at the top of the mockup is the whole job description: the plan name (ANALYTICS Prod → Dev) with its calendar icon, the cadence (Daily at 02:00) and behavior summary (auto-apply) packed under the title, an amber AUTO-APPLY mode pill, a NEXT RUN cell with the friendly 'in 17h 48m' countdown next to a green liveness dot, and the LAST RESULT pill from whatever last fired. Action icons on the right run it now, pause it, or delete it without leaving the row.
What 'AUTO-APPLY' actually means
The amber mode pill is the difference between a schedule that just reports drift (Compare-only) and one that closes the gap automatically (Auto-apply). When AUTO-APPLY is on, the nightly run executes a Compare and then immediately runs Reconcile against the result — but only within a Max planned rows ceiling the operator sets in the wizard. Small drift gets repaired quietly; if a night's count crosses the limit, the run completes Compare and stops, waiting for a human to look. That single guardrail is why teams are comfortable letting it run unattended.
The receipt book underneath
Every scheduled run lands in Recent automation runs in chronological order — date stamp, plan name, status pill, source (scheduled vs manual), outcome (executed vs skipped). The ten green rows in the mockup are ten clean nights; an amber or red entry in the same column would point at the exact night something diverged. The receipt book is what makes the morning standup answer for data consistency a fifteen-second glance instead of a question.
Step 1 — Plan
New schedule opens on a screen that asks two things: name the schedule, and pick the saved Plan it should run. A small info bar at the top reminds the operator that backend automation schedules run comparisons against a mapping Plan — Auto-apply lives in the next step, not as a separate kind of schedule. Naming matters because the name shows up in the receipt book later.
Step 2 — Cadence
Four big tiles — Hourly, Daily, Weekly, Monthly — pick the recurrence. Daily picks a Time of day; Weekly picks days of week; Monthly picks day of month; Hourly lets the backend pick the next hourly slot. The cadence string ('Daily at 02:00') gets stamped on the schedule row in the list so the team never has to open the wizard to remember when it runs.
Step 3 — Behavior (the auto-apply gate)
Two stacked cards. Auto-apply drift after comparison flips the schedule from Compare-only to Compare + Reconcile. Underneath it sits a 'Max planned rows' input (5,000 in the screenshot) that caps how big a drift the auto-apply will fix without escalating to a human. The warning card next to it spells out the trade-off — each scheduled run can write repairs without an operator confirmation — and recommends a conservative cap. An 'Initial state' checkbox decides whether the schedule starts enabled right after Create.
Step 4 — Review
Last screen is the read-back. Schedule name, Plan name, Cadence string, Enabled yes/no, Max planned rows — all in one card next to a small banner that quotes the behavior in plain English ('This schedule will ask backend automation to repair drift automatically when planned rows are within the configured limit.'). Create schedule wires it up, fires the first run on the cadence, and the row joins the list.
Auto-apply with a guardrail
Step 3 of the wizard lets a schedule auto-apply repairs after the Compare run — but only up to a Max planned rows ceiling you set. The schedule will fix small everyday drift quietly; if a night's drift count crosses the limit, it stops and waits for a human. That single threshold is why teams are willing to let it run unattended.
A receipt book, not a black box
Every scheduled comparison lands in the Recent automation runs panel with a date stamp, the plan name, the status, and the source of the run (scheduled vs manual). Reading it backward is how the team answers 'when did this start drifting?' without paging anyone.
The morning report writes itself
An overnight schedule plus a glance at this page is the entire morning standup answer for data consistency. Green pills mean the night ran clean; an amber or red entry tells exactly which plan needs eyes before standup.
