Loop

Find out whether the changes you make actually work — a baseline from the day each change took effect, a score after, and a named owner who decides what to keep.

MangoApps

Book a walkthrough Ask AI Download datasheet

Overview

Most workplace changes are never checked. Someone rewrites a procedure, fixes a request type that keeps bouncing back, or tries a new practice at one site — and nobody ever finds out whether it worked. Loop measures it: name the change and the metric it should move, and it takes a baseline from the day the change took effect, then scores the same metric after.

Changes made elsewhere start measuring themselves. A completed safety corrective action opens a recurrence review; a Service Desk fix is watched from the day it went live; answers banked from a won bid are tracked each time they get reused.

Every result carries its receipt — the comparison windows, how much data it rests on, and how many other changes touched the same metric in the same period — so a number can be argued with. Windows that close without enough signal are reported as inconclusive rather than dressed up as a win.

A good number promotes nothing on its own. A named owner decides: keep it, adjust it, pilot it at another location, retire it, or roll it back — with the reasoning recorded. Kept changes are rechecked later, so a change that quietly stops working reopens itself for review.

Productivity Utility

Highlights

Watch any change you made: describe it, pick the metric, and Loop captures a baseline from the day it took effect — no automation required to get your first measurement.

A result is not an improvement until a named owner decides: keep, adjust, expand, retire, or roll back — recorded with rationale, and adopted changes are rechecked later for regression.

Every result ships with its receipt: the comparison windows, the sample size, a guardrail volume metric, and how many other changes touched the same metric in the same window.

Journeys start in the app that owns the work: completed safety corrective actions open recurrence reviews, banked RFP win answers get per-entry reuse tracking, Service Desk fixes get hold-rate measurement from the day the fix went live.

A leadership portfolio grouped by what needs doing — regressed, awaiting decision, piloting, holding, retired — plus pilots that measure a new location separately before a practice spreads.

Honest by construction: thin windows close as "inconclusive", failed measurements say so and retry, and sample stories are always labeled — they fill the totals only until your first real measurement, then step aside.

Capabilities

Measurement Ledger

Pre/post measurement rows for every intervention (baseline, after, delta, window)
Cross-domain: declared changes, lessons, corrective actions, banked answers, acted-on suggestions, engagement actions
"Watch a change": declare a change you made, pick the metric, measure from its activation date
Per-row receipt: method, windows, sample size, guardrail volume, concurrent changes
Direction-aware verdicts — a resolution time that FELL reads as improved, not regressed
Inconclusive state for windows with insufficient signal (no noise deltas)
Failed measurements show their reason, retry nightly, and can be retried on the spot

Decisions & Adoption

Owner decision on every measured change: keep, adjust, expand, retire, or roll back — with rationale
Adopted changes are automatically rechecked; regression reopens review and notifies the owner
Result-ready and regression notices to the accountable owner via notifications
Pilot a proven practice at another location, measured separately for that location
Leadership portfolio: regressed / awaiting decision / piloting / holding / retired, with owners
Twice-proven practices surface as platform candidates with an evidence brief handed to App Builder
Decisions write back to the originating Muscle Memory lesson, so Ask AI distinguishes observed from adopted

The Loop, End to End

Detect: suggestion services counted, with how many carry measured outcomes
Remember: distiller roster and the active-lesson count, linked into Muscle Memory
Measure: registered scorer domains across the platform
Adopt: a named owner decides on the evidence; positive deltas alone graduate nothing

Automations

Schedule a playbook to run unattended, or fire it on an event
Run a playbook on demand without waiting for its schedule
Pause and resume an automation without deleting it
Set the input values a playbook runs with
Require approval before an automation's risky steps execute
Install a suggested automation from a system template, paused for review

Run History

Every playbook execution with its per-step results and timings
Retry a failed run — completed steps are skipped, not repeated
Cancel a run that is stuck or no longer wanted
Filter by status, playbook, and date range
CSV export of run history for compliance review
Rollup of the tools failing most often across recent runs

AI Impact & Oversight

AI outcomes reported as business results delivered, not activity counts
Tune which AI suggestions fire — rules, thresholds, and how each is performing
See what the AI agents learned from user corrections

Access & Mobile

Business admins and Loop app-admins always have access
Optional read-only manager access (one setting, off by default)
Mobile measurement ledger at /m/apps/loop
One-click sample stories spanning the full lifecycle — including changes that didn't work
Sample rows stay badged and stop counting the moment a real measurement lands

Use cases

Prove the AI is earning its keep

Leadership asks what the AI layer actually changed. The ledger answers with measured before/after deltas per intervention — not usage counts.

Close the loop on a bouncing problem

Service Desk resolutions for one request type keep getting reopened. The lesson that flagged it carries a measurement; thirty days later the delta says whether the fix held.

Decide what becomes standard practice

An intervention measured positive. The accountable owner reviews the receipt — sample, guardrail, concurrent changes — and records keep, adjust, or retire. Kept changes are rechecked automatically; a practice that also survives a separately measured pilot becomes a platform candidate.

Prove a safety fix actually prevented recurrence

A corrective action completes and Loop opens its recurrence review from the completion date. Sixty days later the incident page shows whether the same problem came back — and the printable evidence pack carries the whole chain for an auditor.

Audit the suggestion engine's value

Suggestions get acted on — then what? Acted-on suggestions from services that declare an outcome get their own pre/post windows, so action-rate stops standing in for impact.

Run a playbook unattended — and see exactly what it did

Schedule a playbook, fire it on an event, or run it on demand, with approval required before its risky steps execute. Every run keeps its per-step results and timings, a failed run picks up from the first incomplete step instead of repeating work already done, and the history exports to CSV for compliance review.

Resources

Loop datasheet (PDF)

Datasheet

Capabilities, use cases and limits in one page — generated from this listing.

Loop app overview

Help Article

What the learning loop is, what the ledger shows, and how the absorbed admin surfaces fit together.

AI Outcomes explained

Help Article

The outcome metrics behind the ledger — what counts as a business result vs. activity volume.

FAQ

Detect → remember → measure → decide → keep watching. Apps detect opportunities, distillers turn trusted outcomes into lessons (Muscle Memory), changes get pre/post measurements anchored at their activation, a named owner decides on the evidence, and adopted changes are rechecked so regression reopens review. Loop is the workbench where that cycle runs — execution stays in the apps that own the work.

Muscle Memory is what the business REMEMBERS — evidence-backed lessons with confidence and decay. Loop is how the business LEARNS — the measurement ledger, the automations that act on what was learned, and the admin surfaces for outcomes, suggestions and corrections. Lessons appear in Loop's ledger only when they intervened and got measured.

Two ways. The platform seeds them at intervention moments — a safety corrective action completing, answers banked from a won bid, a Service Desk fix recorded against a bouncing pattern, a suggestion someone acts on. And you can declare a change yourself with "Watch a change": pick the metric and the day it took effect, and Loop measures from there. Scores always come from scorers; no delta is ever hand-entered.

The measurement window closed with too little signal to say anything honest — for example, too few resolved tickets of that type. Loop reports that plainly instead of publishing a delta that would just be noise.

Yes. One click loads a set of sample stories covering the whole lifecycle — a change that held on recheck, one that regressed, a rollback, a retire on "no movement", and a window that closed inconclusive. Every row stays badged "Sample" and the totals count them only until your first real measurement arrives, at which point everything reverts to real-only. Remove them any time from settings.

No. "Watch a change" takes any change you made — a rewritten procedure, a new staffing rule, a fix you shipped — and measures it from the day it took effect, whether or not AI had anything to do with it. Tenants running the AI layer get more measurements seeded automatically, but the ledger, the receipts and the decision workflow work the same either way.

Business admins and Loop app-admins always. One setting (off by default) extends read-only overview access to managers; the interior admin tabs stay admin-only regardless.

Loop is included with the platform tiers alongside Muscle Memory — it isn't sold à-la-carte. Like every app, it provisions disabled and an admin opts in from the Apps Marketplace.