Work Measurement and Time Study

Amongst other things, being an industrial engineer (MIIE(Aust) resigned 2014) I don’t dismiss time casually, so when I say we don’t sell time it doesn’t mean we don’t pay attention to time.

When conducting a time study, it is typically done using a decimal minute stopwatch (if you can find one in SA), so instead of measuring minutes and seconds, measure minutes and 1/100th of a minute. Time studies are however of less importance than work studies, and importantly a work study needs to be conducted before a time study can take place. The purpose of the work study is to identify individual activities within a task, and more importantly identify end events or break points. For example picking up nut and placing on end of bolt is an activity, the nut resting on top of the bolt is an end event. An event is an instant in time, if we had a camera we could take a snapshot of the state of the system at such instant in time. It is the duration between the end events which add up to the total time for the task. Time measurement of the task can be done by measuring the duration of each activity between the break points, but it is unlikely the operator is fast enough to stop and start the stopwatch and record the time. The alternative approach is continuous timing, rather than stop/starting simply record the time at each of the break points, then after the study subtract the adjacent event times to calculate the duration of the activity between the events. Now the important part about the time study is that recording times to 1/100th of a minute is important. For such to be important a task has to be only a few minutes in duration, definitely less than an hour. Also of importance is that the task is expected to be repeated several 1000 times.

Office work in general may involve repetition and various short term tasks, and similarly fabrication, construction and building also may involve short duration tasks and repetition, in the main however the tasks are not consistent. Whilst in the main we can identify a general pattern of activity, and say some generic task is repeated, the actual activities within a generic task will change and be dependent on a set of conditions being met. For example the task of installing a beam may required installing 4 bolts or it could require installing may be 20 or more bolts. Similarly this variation in quantity of bolts varies the time required to produce a drawing of the connection and the time required to determine the forces in the bolts. Furthermore it isn’t simply proportional, having 10 bolts instead of 1 bolt, in a connection doesn’t result in the connection taking 10 times longer to assess: the time required depends on the pattern of the bolts which means the connection needs to be drawn first.

In a general office, fabrication workshop and on a construction site, the activities which anyone is pursing at any point in time is variable: they are not manufacturing the same widget everyday using the exact same process. In such situation rather than time study of the individual task it is more productive to carry out activity sampling. In which case can either choose to monitor an entire day from start to finish, or check the activity at random times of the day. Basically if random sampling indicates the same activity throughout the day, and it is known the activity should change, then it is better to monitor the whole day if there isn’t a schedule of work activity. The activity sampling can cross days, weeks and month: to get a better picture of activity throughout a year.

So from working elsewhere on contract and timing my work I know that I can transform architects CADD drawings of large buildings into structural drawings in around 5 hours. Refining these drawings into final production drawings can take another 15 hours or more. Basically I found most tasks could be broken down into 5 hour blocks or less.

On the other hand when we started the business, producing residential footing construction reports (FCR), we were typically given papr drawings, for which it can take 5 hours to produce a CADD drawing of a floor plan before can start drawing residential footing: thats a big drain on time compared to a tradition of using tracing paper or a lightbox to produce a structural drawing with pencil and paper.

Anycase what we have are projects and tasks measured in hours not minutes. On the other hand in the age of computers, the time for a computer to execute a task can be measured in a fraction of a second, as computers can execute millions of instructions per second. It was my argument back in the early days of CADD that we didn’t need faster computers, rather we needed to let the computer do the work. In these days, our drawings took 30 minutes to open and 30 minutes to regenerate every time we zoomed to a view which caused regeneration (we could minimise such regeneration by using AutoCAD’s zoom dynamic command which identified if regeneration would be required). For certain the computers we were using were slow, and replacement of the computers dropped the times down to between 2 and 4 minutes. But these drains on time were once a day activities, unless working on multiple drawings, the big issue was and still is the time the computer spends doing a task versus the time an operator works on a task.

As my simple AutoCAD speed test shows the manual time for the simple drawing task is 26 seconds, but if program the task with a script then the time drops to 0.094 seconds. So if have 480 minutes in a working day, and it takes 30 seconds to input each set of instructions, then can input 960 instructions each day, each of which takes the computer less than 1 second to execute. So maximum time is 960 seconds of computing time: or 16 minutes. Therefore total task time 496 minutes. Or refining this can complete 929 instructions, which the computer will take around 15.5 minutes to execute in a 480 minute day: with operator taking 464.5 minutes to input instructions.  So if we make the computer 10 times faster, then get 1.6 minutes of computing time for 957 instructions taking 478.4 minutes to input. The improvement in the number of instructions executed is minor. But let’s try making the computer 100 times faster: 0.16 minutes to execute 960 instructions taking 479.84 minutes to input. The improvement in number of instructions executed is relatively insignificant, we are not completing any more work. What we need to do is reduce the number of human instructions input before the computer goes ahead and executes them. It isn’t a faster computer which is needed but automation of the task: that is we need to program the application.

Programming the application requires work study not work measurement. A 5 hour task (300 minutes) can often be reduced to 40 seconds of computing time which takes less than 2 minutes to input the instructions. The data returned by the computer is not in decimal minutes its in fractions (1/1000th) of a second. But we have a problem, to collapse the 5 hour task to 40 seconds can take 40 hours or more of programming time. If the task isn’t repeated then it isn’t worth the effort to automate. However if can make the automation program flexible, then it can be used for similar tasks by inputting different operational parameters.

So when it comes to selling time, the problem is that a task which took 40 hours is reduced down to 5 hours, and the market price is initially locked at 40 hours multiplied by some hourly rate. If the hourly rate remains the same, then the price offered by new players using new technology can drop to 1/8th of the original price. On the other hand the new players can increase their hourly rate slightly and still use the reduced time for the task: for example they could simply half the price. The issue is what price is the market willing to pay at a given point in time, how much profit do they consider acceptable for the task?

Anycase the basic task, whether it is a set of drawings or a set of calculations takes around 5 hours. So a set of calculations and drawings takes around 10 hours. As things get more involved then the time required increases in 5 hour blocks, not minutes. A typical project involving interaction and feedback is likely to take a minimum of 20 hours, and involve doing the drawings and calculations at least twice. So if take $20/hour as a minimum rate, then typical project will have minimum fee of $400 (excluding GST).

Now I can sit down and break the task down into activities and I can estimate time for each activity in minutes, no need to measure them. Though often I do measure activities. For example setting up the simple formula M=wL^2/8 in a spreadsheet takes from 29 to 42 seconds: the 99.9 percentile value is 54.1 seconds, from which can conclude takes approximately 6.02 seconds/cell to build a calculation spreadsheet: ignoring long descriptive cells and ignoring vba source code. My largest workbook has 39,703 used cells, thus estimate takes 66.4 hours to produce. Thus at minimum rate it cost $1327.85 to produce, allowing maximum sale of 1000 units, it could retail for $1.32/unit. On the other hand the workbooks remove the need for others to spend time building the workbooks. My estimate is it roughly takes the same amount of time to do calculations by hand as it does to set them up in a spreadsheet, though double the amount of time may be required to improve the presentation of the spreadsheet. But basically 40 hours of hand written calculations can be reduced to a spreadsheet which takes less than an hour to work through if used regularly, or around 5 hours to read through in detail if used occasionally. So the typical buyer is likely to be reducing 40 hours of time down to 5 hours, thus saving 35 hours at their rate of around $36.79/hour, so saving $1287.65. So if  the spreadsheet priced at $20 it would cost less than an hours worth of work and still save $1267.65, and that is only considering a single project. It is thus strange that more consulting businesses do not give their personnel the opportunity to create design tools, and pay for the development in-house.

I can do calculations with pencil and paper, and if I am not clear about what I need to calculate then I usually start with pencil and paper. If however the requirements are relatively clear, I don’t need to rearrange formula and I don’t need illustrative diagrams, then I can just go straight to using a spreadsheet. If I do need diagrams I can still do those with pencil and paper to guide what I am doing in the spreadsheet. All in all the spreadsheet just replaces my pocket calculator: only instead of writing formula on paper and then punching numbers through the calculator and possibly making a transcription error, the spreadsheet allows the same numbers to be used in the calculation as are displayed in the report. As indicated elsewhere I find spreadsheets preferable to MathCAD/SMath, since every number I put in a cell is a variable and can be used elsewhere. I don’t have to spend time isolating variables, and subscripting them. Anycase whether use a spreadsheet or Computer algebra System (CAS) getting the calculations into the computer and creating a report shouldn’t take much longer than using pencil/paper and pocket calculator. When finished however the numbers can be more readily changed than when using a pocket calculator. The first project would take say 40 hours, just like if used pencil and paper, but the next project will only take 5 hours, and the one immediately after that will take 1 hour. Then depending on a lack of variety between projects, with some 90% of the parameters staying unchanged from project to project, each project could be reduced to 5 minutes.

But even though reduced the primary task to 5 minutes, still have around 30 minutes or an hour or more talking to the client determining their requirements before can decide that can use the tools available for rapid calculation. Usually from project to project something will change, which makes the existing tools unsuitable and thus require modification of existing or creation of new tools from scratch.

Consider the following simple equation:

Vz =Vr. Md (M[z,cat].Ms.Mt)

This simple calculation can take less than a minute to calculate, and for a typical project that is probably all that is required. However the calculation needs to be calculated for 8 compass directions, still if all assumed to have the same conditions, then still less than a minute to calculate. However, depending on the site, it can take an hour or more to calculate each of: M[z,cat] considering terrain averaging, to check the topographical factor (Mt) if not on flat ground and conditions not really in scope of the code, and to check shielding (Ms). Then depending on the project requirements I may have to vary Vr by changing the probability of exceedance and life expectancy of the structure. So a 1 minute calculation can increase to 4 hours or more. Consider that similar situation can occur for any of the 100 or so calculations making up a project. So if the calculations are quick then 100 minutes (1.67hours), if slow then time shoots up to 100 hours. That poses a problem because the effort required to reach a result has significantly increased but the value of the output has not.

So say that a wind speed assessment has a market price of $55, as long as it takes less than a minute to calculate and around 30 minutes to discuss requirements and create the project then potentially acceptable: roughly $106.45 /hour.  But if we get upto 4 hours for the task then rate drops to $13.75/hour which is less than federal minimum wage $18.93 and likely not acceptable. However if less than 10% are of this nature and $55/hour is an acceptable rate then it’s not really a problem on condition that the longer projects are randomly distributed and do not all occur at the beginning of a business: and noting typically getting 1.94 times desired rate. Assuming 2000 hours/year, then can complete minimum of 500 such projects in a year.  So $55×500 = $27,500 which is a low salary compared to federal minimum of 38x52x18.93=$37,405.68. So not good to be doing all year, but minimum duration is 31 minutes, so maximum projects is 3870 projects, giving $55×3870 = $212,850. {Maybe I should get bored to death doing wind speed assessments all year}.  The actual income is somewhere between these two extremes: which can either solve algebraically or just formulate and use Excel goal seek to solve. Taking this into consideration we get 2081 short projects and 231 long projects, giving a total of 2312 projects at $55/each: $127,160 which seems a reasonable income when compared to the federal minimum. Though compare against industrial award for registered nurse level 5 grade 6, $2025/week => $105,300, and it’s still looking reasonable. Or professional employees award level 4, $72,704. Further consideration $55x38X52 = $108,680 and expected other operating costs less than $30,000, so total revenue required $138,680 and now it looks too low. Except desired income can be varied, on the other hand , the other operating expenses could probably be dropped to around $15,000 working from home, so total becomes: $123,680 and we now have some surplus. But I’m just pulling numbers out off a hat. The desired income is really dependent on whether want $20 shoes or $200 dollar shoes: as long as there is choice then low incomes are acceptable. When only have the option of $200 shoes then the income needs to rise to accommodate. As a supplier we likewise have a choice as to whether to cater to those on low incomes or only those with extortionately high incomes. Not everyone can be supplier to those on top incomes, there are very few of them, so most have to supply to those on lower incomes. All the business advice however comes from those with top income clients: its mostly unrealistic.

So further reason not to sell time, is that the activities under consideration have such a variability in time requirement, such as from 1 minute to 4 hours ( 240 times longer). Such time difference is often not apparent until start the job and working through it: as the time change is dependent on conditions met as work through the job. So in working out fixed fees, have to account for such likely variability in time.

Alternatively, I can conduct a work study and specify the conditions which cause the work effort to change, and then create job specifications which exclude such variability. For example if I can specify the wind assessment in detail, then I can exclude the long duration projects and charge a fee of $55x(31/60)=$28.42 for the typical wind assessment project.

The reality however is juggling a whole product mix of different projects throughout the year, and making sure the mix generates an appropriate income. If all too low then none of it balances and fail to make an acceptable living. If all too high then probably won’t stay in business for long.

Considering the wind assessment again, why is it needed? On the one hand it is needed because legislation requires a wind classification so that appropriate components can be selected for a house. On the other, it is useful to reduce costs. Without the site specific wind assessment likely stuck with a wind class N3 requirement, but with assessment possibly show that wind class N1 is adequate: this may save from a few hundred dollars to a few thousand.  So assume we decide that 10% of saving is an acceptable fee. So $900 saving and fee becomes $90, thus 1.63 times higher. Or if saving in the thousands, say around $2000 then fee goes upto $200, which is 3.64 times higher than our default $55.

That’s 1 minute of calculation potentially worth around $200, the 30 minutes of project administration not worth much of anything, its an unwanted expense to remove. The problem is need to be able to design the two options and cost the two options so that can determine the saving between the two: such cost comparison rarely occurs in practice. But it is a point to note, in terms of the region N3 tends to be the upper limit experienced, though the upper wind class is N6 (or C4). If don’t want a wind assessment then just specify components for wind class N6, if know the region typically has a wind class no higher than N3 then specify for N3.

I don’t need to measure time, because I spend a lot of time reducing the time taken for a task, by using computers to automate the task, then the whole benefit is lost when the next project comes along and it introduces additional parameters. The 5 hour task is reduced to 5 minutes, then increased to 20 hours. By not reducing the project fee based on the new time, funds are made available for when the project duration increases again. In short all clients pay what is reasonable in order to fund that where the fee would become unreasonable but the cost none the less exceeds the fee.

As indicated at the beginning before can measure time first need to conduct a work study, or activity sampling to identify how time is being used. By incremental refinement of the activity descriptions, the actions get small enough that they take less than 1 minute to perform. For a task which activity sampling indicates takes 5 hours, an estimate or guess of 1 minute for component actions is good enough: as that would require dividing the task into 300 activities which is highly unlikely. The task is more likely to be divided into activities which can vary between 5  minutes and 90 minutes in duration.

The variability is too great and therefore need to estimate an upper and lower limit to the durations. Then can either take the average of the two values or estimate the frequency of occurrence of the two values. If using frequency of occurrence then can consider more than an upper and lower limit, can consider multiple durations with differing frequencies. We can then start to get a better feel for the work mix or product mix. With clearly defined products, can then define price blocks and allocate each product to a price block. For such purpose it is important to know the difference between a thing called engineering and technical design. The thing people seek when told to go get “engineering”, is not engineering, it is technical assessment: a rational scientific appraisal of fitness-for-function. Whilst there is uncertainty in the time required to do job there is little uncertainty in the outcome: either assert a proposal is fit-for-function or refute such assertion.

So the time to complete a broadly defined task is highly variable, whilst the hourly rate people charge to complete such task is whatever they pull out off a hat, the income they desire. Whilst the market price is the amount the customer is willing to pay, they can afford, and otherwise a number they pull out off a hat. The task is being able to guess what number the customer is going to pull out off the hat, and hope that is higher than the costs.

Though an important factor is that employing organisations have locked labour rates, whilst a sole practitioner can vary their labour rate. For the sole practitioner a labour rate is not a cost, they have a minimum income, which they need to generate, to cover their actual living costs, and then the income they would like to generate, to improve their standard of living. As long as the total value of projects for the year can generate the requirements to cover actual living costs, the pricing of projects and volume of projects is adequate: assuming that weekly requirements are also adequately met as the peaks balance the lows.

So simple rule: if cannot break task down into activities which take less than 1 minute, then don’t really need time study. If the time for a task is highly variable then also don’t need time study. In both cases do not need to progress further than work study. Once have a documented work study can work on simplifying the task, to reduce time, or automating the task to reduce time. There is no need to measure the time to prove it has been reduced, its usually blatantly obvious. Furthermore rather than drop price, the typical business is more likely to increase labour rates because they can do the work faster: faster typically has higher value than slower.


Revisions:

  1. [2/10/2018] : Original