Short answer
Labor time standards give each task a reference duration, such as a repair time in a flat-rate guide or a motion-based time from a predetermined time system. Attached to demonstration clips, they can serve as relative difficulty labels and help balance a task mix. They measure human time, not robot success, force or skill, and publishers often restrict redistribution, so rights come first.
The buyer's problem
A collection plan for maintenance tasks gets lopsided fast. Quick jobs dominate the footage because they are frequent; the long, hard jobs are rare and expensive to record. If you train or evaluate on a flat sample, the easy tasks decide your results. A buyer needs some way to say, before collection, how much harder one task is than another, and some label to stratify by afterward.
Labor times are a ready-made candidate. Repair networks, shops and industrial engineers have already estimated how long tasks take, for pricing, scheduling or work design. This page covers what those numbers are, where they come from, how they can be attached to data, and where they mislead. For choosing which tasks to collect in the first place, see prioritizing maintenance tasks for collection.
Two families of time standard
| Family | What it gives | Where it comes from | Who tends to hold it |
|---|---|---|---|
| Flat-rate or book labor times | One reference duration per repair operation, per equipment model | Publishers or manufacturers, often by subscription | Repair networks, dealers, insurers |
| Predetermined motion-time systems | Duration built up from basic motions | Industrial-engineering methods applied by analysts | Manufacturers, consultancies |
Flat-rate guides are commonly used to price repairs. Their terms, scope and update cycles are set by their publishers, and this page does not make claims about any particular guide. Ask each supplier which publication they use and what it covers.
Predetermined motion-time systems are better documented publicly. The MTM Association describes Methods-Time Measurement as a method standard for describing, evaluating and designing human work (MTM Association). Training materials and lecture notes, not the association's own page, describe MTM as analyzing manual operations into basic motions such as reach, grasp, move, position and release, each with a predetermined time, and give 1 TMU as 0.00001 hour, or 0.036 seconds. Treat those details as secondary descriptions and confirm them against the supplier's or the association's documentation before building on them.
The key idea for a data buyer is that a motion-based analysis breaks a task into motion elements with timing. That gives you something finer than a total repair time: a decomposition you can link to video segments. The supply line this belongs to, industrial-engineering and EHS consultancies, may be able to offer task decomposition, timing and posture data, but not robot actions.
What a time label can and cannot tell you
A time standard can support:
- Relative difficulty or effort. A task with a longer reference time is plausibly more complex or longer-horizon.
- Stratified sampling. Balance short and long tasks in collection and evaluation sets.
- Segment priors. A motion decomposition suggests where step boundaries fall and how long segments should last.
- Progress estimation baselines. Compare observed duration with the reference for a sanity check.
- Cost-benefit framing. Estimate how many human hours of work a task represents. See work order histories for pairing with actual durations.
It cannot support:
- Robot success or speed. A robot's time depends on its hardware, planner and safety limits, not on a human reference.
- Skill or force. Time is silent on how a task is done. Two technicians can hit the same time with different technique.
- Difficulty of perception. A short task may be perceptually hard, such as finding a hidden clip; time says nothing about it.
- Failure and rework. A book time assumes a standard path. Hard, unusual jobs may take far longer; see difficult and failed repairs.
- Control training. Time labels are metadata. Control still needs robot or teleoperation data, and human video alone cannot supply it.
Linking times to demonstrations
The value comes from joining a time record to a clip or a work order. Three joins are possible.
- Operation-level join. Link a repair operation in a labor guide to a recorded job of that operation. Needs a shared identifier or reliable mapping from the job description.
- Segment-level join. Link motion elements from an analysis to labeled video segments. Needs timestamps and agreed step names.
- Aggregate join. Use times only to stratify task categories, with no clip-level link. Weakest, easiest.
Observed durations in the video usually differ from the reference. Keep both. Their ratio is itself a feature: it shows how this execution differs from the standard, though many things can cause that, including camera placement, interruptions and technician pace.
Illustrative package
Synthetic; shows structure only.
| Field | Illustrative value | Note |
|---|---|---|
| task_id | op_0412 | Supplier or buyer taxonomy |
| equipment_class | class_X, model_group_3 | Scope of the reference time |
| reference_time_source | flat_rate_guide_A rev 2026 | Name of publication |
| reference_time_value | value with unit | Do not alter units |
| motion_breakdown | [reach, grasp, move, position, release] with element times | Only if an analysis exists |
| observed_duration_s | from clip timestamps | Measured, not assumed |
| clip_id | clip_00973 | Links to the demonstration |
| notes | free text on conditions | Access, rust, special tools |
Supplier types and rights
Repair networks, dealers, parts-and-service publishers and industrial-engineering consultancies may hold these records. The main issue is rights. Flat-rate times are often licensed products, so a repair shop may have a subscription that does not allow it to pass the content on. Consultancy analyses may be owned by the consultancy even when the underlying work happened on a client's floor, and the video of employees may carry its own consent conditions. See licensing terms for robotics training data and worker consent and identifiable people. Using numbers from a licensed guide as labels may still be a use the license addresses. Ask. This is not legal advice.
For route comparison, see existing assets, new collection or specialist partners. Automotive repair is a setting where flat-rate practice is often discussed, though suppliers should confirm what they use; the automotive service tasks page discusses that domain, and first-time fix and callback labels covers an outcome-based alternative.
Limits and extra preparation
- Averages hide variation. A reference time is one number for a family of conditions.
- Scope drift. Whether a time includes diagnosis, parts retrieval or cleanup differs by publication.
- Human pace. Standards describe human work and may assume a skill level.
- Unit and rounding conventions. Confirm them; do not mix systems silently.
- Mapping effort. Matching repair descriptions to operations is manual work.
- Evaluation. Do not tune on the same tasks you use to evaluate difficulty correlations.
Questions to ask suppliers
- Which publication or analysis produced these times, and which edition or date?
- What does each time include and exclude: diagnosis, access, setup, cleanup?
- Is a motion-level breakdown available, and who produced it?
- How were times validated against observed work?
- Can the times be joined to work orders or recorded jobs, and by what key?
- What do the license terms say about use as training labels or in derived datasets?
- Are times reproducible by the buyer without exposing the source's full tables?
- Who owns analyses done at a client's site: the consultancy or the client?
- Can the supplier share samples to test mapping quality?
Discuss your requirements
If task-time labels belong in your maintenance data plan, describe the tasks, time sources and permitted uses to SourceX. SourceX looks for US businesses that hold suitable material, checks the data and the supplier's licensing permissions, and follows up about potential sources; annotation and linking work are scoped case by case. A request does not guarantee a match. Start at the humanoid training data guide for the larger map or the maintenance cluster for neighboring pages, and see the catalog entry for maintenance logs.
Explore sourcing and licensing options with SourceX
Sources
- MTM ASSOCIATION e. V. (Methods-Time Measurement) — MTM as a method standard for describing, evaluating and designing human work; basic-motion and TMU details on this page come from secondary training materials and are hedged accordingly.