Time Study Data Model: Observation Sheet Fields to Standard Time

Key takeaways

• A time study produces one defensible number: the standard time per piece. Routings, costing, line balancing and capacity all sit on it.
• The chain is observed time × rating ÷ 100 = basic time, then basic time plus allowances = standard time.
• Two choices inside that chain move the answer, so write both down: which basic time carries the allowance, and whether the percentage is of basic time or of total time.
• The rating is a judgement, not a measurement. Store it in its own column so a reviewer can challenge it alone.
• Cycle count is set by variability, not habit. Below, one unexcluded interruption turns a 9 cycle study into a 382 cycle study.
• Never feed a padded standard into OEE as the ideal cycle time. The same shift reports 86.6% performance on the standard time and 76.0% on the basic time, and lower still against a true ideal cycle time.

What a time study produces

A time study is a direct observation of one operator doing one operation, broken into elements, timed cycle by cycle and converted into a standard time.

The output is not "how long it took". It is how long the job should take, at a defined pace, with defined recovery time.

That number drives the routing (run time per piece, setup per order), costing, capacity, line balancing and the staffing a cell needs to hold takt time.

It also feeds the performance factor of OEE, but only after a correction most plants get wrong.

A standard is only as good as the method it was timed against, so it must point at a specific standard work revision.

The observation sheet as a data model

Most plants run studies on paper, then type the result into the ERP. The raw readings are lost, and nobody can recompute the standard two years later.

Split the sheet into three tables: study, element definitions, individual readings.

1. Study header

One row per study, and the part an auditor reads first.

Record the conditions. A study on a new die and a study on a worn one are two different studies, and only this field says which is which.

2. Element definition

One row per element per study. Elements are the units you reuse, compare and move between stations.

The break point is the field people skip and the one that ruins comparability. "Until the press bang" and "until the part is clear of the die" are different elements.

3. Observation readings

One row per element per cycle. This is the evidence, and it is what spreadsheets usually throw away.

is_valid plus exclusion_reason is the honesty mechanism of the model. A reading removed without a written reason is a number someone decided you should not see.

Snapback or continuous timing

The two stopwatch methods store different things, and converting between them works in one direction only.

With continuous timing the observed time is this break point minus the previous one: readings of 0.15, 0.55 and 0.77 give element times of 0.15, 0.40 and 0.22 min.

That matters for audit: every second is accounted for, interruptions included, because the readings must always increase.

Snapback cannot be rebuilt into a timeline, because it never records a cumulative clock and each reset quietly loses a fraction of a second. Time continuously whenever the study has to be auditable, and keep snapback for quick checks where the timeline does not matter.

From observed time to standard time

Five steps, in this order. Each gets its own stored column so the arithmetic can be checked.

1. Observed time and rating

The observed time is the raw elapsed time for one element in one cycle. The rating is the observer's judgement of pace against a defined standard performance.

The common scale sets 100 as standard performance, so a rating of 110 means a pace 10% above it. Standard performance is the rate a qualified, motivated worker can hold for a whole shift without over-exertion, and that sentence has to be written into the study or the number means nothing.

Record which scale you used, because more than one is in circulation and they are not interchangeable. A rating of 80 is above standard on one scale and well below it on another.

2. Basic time

Basic time = observed time × rating ÷ 100. A rating above 100 raises the observed time, below 100 lowers it.

Apply it cycle by cycle, then average. Rating every reading and averaging the results is not the same calculation as multiplying a mean observed time by an average rating, and only the first one survives a review.

3. Frequency

Occasional elements do not happen every cycle, so multiply their basic time by the frequency.

Changing a tote every 20 pieces at 0.500 min contributes 0.500 × 1/20 = 0.025 min per piece.

4. Allowances

Allowances are added time for things that are real but are not the work itself.

Allowance percentages are set by each plant, not by this page. The relaxation allowance tables that circulate in work study textbooks are illustrative examples, not an issued standard, and no body certifies the numbers in them.

Treat any published table as the opening position for a local agreement. The values you apply are your own, and they must be agreed, dated and documented.

The percentages below are example inputs only, not standards.

5. Standard time

Standard time = total basic time × (1 + allowance fraction). Store the allowance percentage next to the result, never folded into it.

Check what the percentage is a percentage of. The formula above reads 14% as 14% of basic time, but many agreements express rest as a share of total working time instead.

On that second reading the divisor changes: standard time = basic time ÷ (1 minus 0.14), which turns 0.8095 min into 0.9413 min rather than 0.9228 min. Store the base in its own field beside the percentage, because the two readings differ by about 2% on every piece.

Worked example you can check

One press operation, four elements, decimal minutes. This plant has agreed 12% relaxation and 2% contingency, 14% in total.

Check element 1: 0.150 × 0.95 = 0.1425. Check element 3: 0.220 × 1.10 = 0.2420.

Now the allowances. 0.8095 × 1.14 = 0.9228 min, which is 55.4 seconds per piece, or 65 pieces per hour.

The decision that moves the answer

Many plants apply relaxation allowance only to the manual basic time, because the operator is not working during the machine element.

Manual basic time here is 0.1425 + 0.2420 + 0.0250 = 0.4095 min.

Then 0.4095 × 1.14 = 0.4668, plus the machine time 0.4000, gives 0.8668 min: 52.0 seconds and 69 pieces per hour.

Two methods, 65 and 69 pieces per hour, and they are not equally defensible. Relaxation allowance pays for recovery from effort, so putting it on time the operator spends watching a press is a convention of convenience rather than a rest requirement.

Write down which one your plant uses, because mixing them makes work centres impossible to compare. If the operator is genuinely idle during the machine element, that unoccupied time belongs in its own column, not inside a relaxation percentage.

Rating is a judgement, machines are not rated

Rating is the only step in the chain that is not a measurement. Two trained observers watching the same operator will not always write the same number.

A rating corrects pace, never method. If the operator walks three steps the layout does not need, rating at 100 prices those steps into the standard for as long as the standard lives.

So store it separately rather than hiding it, and a reviewer can accept the timings while challenging only the judgement.

Two controls cost nothing: rate per element per cycle rather than once for the study, and have two observers rate the same recorded cycles each quarter.

Machine controlled elements must never be rated. The press takes 0.400 min whoever stands in front of it, so a rating of 95 would cut the basic time to 0.380 min and delete 0.020 min of real machine time.

A rating of 110 does the opposite and invents 0.040 min that nobody spent. Either way the error is silent, because the machine element looks like every other row.

Set ratable = no so the system refuses any rating but 100, and apply the same rule to conveyor, oven and curing paced elements.

How many cycles should you observe?

The honest answer is that variability decides, not a round number. The formula for a sample size at a stated confidence and precision is:

n = ( z × s ÷ (a × mean) )²

Here z is the confidence factor (1.96 for 95% confidence), s the sample standard deviation, a the relative precision (0.05 means plus or minus 5% of the mean, not 0.05 min) and mean the mean observed time.

Three assumptions ride inside it. The readings must be independent draws from one roughly normal population, the normal z is used even though s comes from a small pilot, and the answer sizes one element, not the study.

So run it element by element and take the largest answer. Then recompute s once the extra cycles are in, because a ten reading pilot can understate the requirement badly.

Worked sample size

Ten readings of element 3, unload and deburr, in decimal minutes: 0.21, 0.23, 0.20, 0.22, 0.24, 0.21, 0.25, 0.22, 0.20, 0.22.

They sum to 2.20, so the mean is 0.220 min, and the squared deviations sum to 0.0024.

The sample standard deviation is the square root of 0.0024 ÷ 9, which is 0.0163 min.

Now the formula: 1.96 × 0.0163 = 0.03195, and 0.05 × 0.220 = 0.01100.

0.03195 ÷ 0.01100 = 2.905, and 2.905² = 8.4, so 9 cycles. The ten already taken are enough.

Note how hard this bites. Double the standard deviation and the required cycles quadruple, from 8.4 to 33.8, so 34 cycles.

The rule-of-thumb tables, honestly

Many plants never run this calculation. They use a published table saying something like "cycle under 0.10 min, take 40 readings; cycle over 2 min, take 10".

Those tables are a shortcut used in real factories every day, and also a guess about your variability made by someone who never saw your line. Take the table's number, then compute s and add cycles only if the formula says you must.

Mean, median, standard deviation and exclusions

Store all three statistics per element. They answer different questions, and the gap between them is a signal.

When mean and median separate, something happened that is not the job. Take the same ten readings, but cycle 7 now reads 0.62 min because the operator waited for a missing tote.

The median is still 0.220 min. The mean jumps to 2.57 ÷ 10 = 0.257 min, which is 16.8% higher from one reading.

The standard deviation jumps to 0.128 min, and the sample size becomes (1.96 × 0.128) ÷ (0.05 × 0.257) = 19.53, squared = 381.2, so 382 cycles.

That is the argument for exclusions. One unhandled interruption turned a 9 cycle study into a 382 cycle study, and would have padded every piece for years.

Do not go and time 382 cycles. The sample is no longer one population, so the formula is sizing mixed data, and 382 is a flag on your sheet rather than a work order for the observer.

How to exclude a reading properly

1. Keep the row. Set is_valid = false, never delete.
2. Write the exclusion_reason in words: "waited 25 s for tote, material supply".
3. Ask whether the cause belongs in a contingency allowance instead. If it happens daily it is not an outlier, it is the job.
4. Recompute from valid rows only, and store the count of valid and excluded readings beside the result.

"Mean 0.220 min, 10 readings" and "mean 0.220 min, 9 valid, 1 excluded, reason recorded" are not the same study.

Standard time, ideal cycle time and the padding trap

A standard time and an ideal cycle time are different numbers with different jobs, and plants routinely use one where the other belongs.

Our guide to cycle time defines the actual and ideal versions. The trap appears when the padded standard is loaded into OEE as the benchmark.

The same shift, two answers

A work centre runs 405 minutes and produces 380 pieces.

On the standard time of 0.9228 min: 380 × 0.9228 = 350.7 min earned, and 350.7 ÷ 405 = 86.6% performance.

On the unpadded basic time of 0.8095 min: 380 × 0.8095 = 307.6 min, and 307.6 ÷ 405 = 76.0% performance.

Basic time is still not the ideal cycle time. It is the time at standard rating, while the OEE benchmark is the fastest sustainable time, so the honest performance figure is lower again.

The padded benchmark hides 10.6 percentage points of speed loss, roughly 43 minutes of output per shift that nobody is looking for.

Keep both fields: standard time for money and people, ideal cycle time for OEE performance. The same warning applies to production efficiency, where a loose standard gives a flattering percentage and no action.

Time study, PMTS, work sampling and machine data

Stopwatch time study is one of four ways to get a time, and each is right somewhere.

These are complementary, not rivals. PMTS or time study sets the standard, and machine data collection checks weekly whether it still describes reality.

Storing revisions so a standard can be audited

Someone will eventually ask why a part is costed at 0.92 min and who approved it. Never update a standard in place: insert a new version and close the old one.

Two rules make this work. A method_rev change forces a new standard version, and a standard past its review_due belongs on a report rather than being silently trusted.

The same versioning applies to times carried on production orders, covered in our production work order data model.

Common mistakes

• Timing the fastest operator and calling it standard. The rating already handles pace, so this does the job twice.
• Averaging raw readings that include interruptions. That is the 0.257 instead of 0.220 above.
• Rating a machine element. It invents or destroys time the machine controls.
• Losing the raw readings. If only the standard survives, nobody can audit or defend it.
• One allowance for the whole plant. A cold store and a seated inspection bench do not carry the same fatigue.
• No method reference. The fixture changes, the standard stays, and the number drifts in silence.
• Feeding the padded standard into OEE. Performance looks fine while micro stops hide inside a loose benchmark.
• Restudying only after complaints. Reviews triggered that way only ever loosen standards.

How Fabrico helps

Fabrico is not a time study or work measurement tool. It holds no observation sheets, calculates no ratings or allowances, and produces no standard time.

Fabrico is an OEE platform with a full CMMS. It measures actual cycle times, short stops and downtime through PLC connectivity, IoT sensors and computer vision cameras, the options compared in our guide to machine monitoring systems.

That is the check your standard needs, and the comparison stays in your hands. The field Fabrico uses as the OEE benchmark is the ideal cycle time, so load that one and keep the padded standard out of it.

Then set your calculated standard beside the measured cycle times yourself, shift after shift. When the machine has quietly been beating the standard for a month, the standard has drifted.

Order data, including the product and its ideal cycle time, can be imported from your ERP, so the OEE benchmark is the one your engineers approved rather than a number retyped by hand.

On the maintenance side your team creates work orders, runs preventive plans and records spare parts used per work order from web or mobile. The AI assistant answers questions about a machine's history in plain language.

Want to see how your approved standards compare with what your machines actually do? Book a 30 minute demo with a Fabrico consultant, no commitment, or contact us with your questions.