Using videos to improve operations | Part 6 – Quick simograms

Here, we finally start collecting measurements from the video, focusing on what we can collect while watching without stopping. In this mode, we can break down operator time by broad categories like  “Waiting,” “Walking,” or “Assembling,” but we don’t have the time to name each task and collect comments or improvement ideas. This will require a more detailed and time-consuming analysis. 

One method, developed by Christophe Caberlon, involves two analysts, one viewing the video and the other one filling out an electronic spreadsheet. Instead of looking for state-change events in the video, we look at it in 5-second increments. Every five seconds, the analyst viewing the video calls out the state the operator has been in since the previous call. Each 5-second. Interval is assigned one column in the spreadsheet and there is one row for each state. Based on the call, the second analyst switches the color of the cell for the state and time interval.

Counting in 5-second intervals involves aliasing, but it is not a problem for a rough-cut estimate. The rows in the spreadsheet do show the state transitions in a Gantt-chart like format called “simogram,” and can summarized into proportions of time spent in each state, as in the following example:

Simogram example

This example uses cell background color to express content, which is not generally recommended because Excel does not provide built-in tools either for quick input or for analysis. The result, however, is graphically much more attractive than filling the cells with Xs. Changing the background color of a cell requires multiple steps, which cannot be repeated every five seconds. These steps, however can be recorded as a Macro. In this example, the macro has Ctrl+q as a hot key to mark a cell and Ctrl+w to unmark it. Also, each 5-second time segment must be assigned to one and only one category. When working your way through a video, it is impossible to avoid cases where one segment will be missed and another accidentally assigned to more than one category.

To detect these errors, we need to count the gray cells by column, and to summarize the times into relevant aggregates, we need to count them by rows. While Excel provides no built-in function to do this, you can find add-on modules to do it. The modules used above are due to C. Pearson

This method is also restricted in the number of states to track. It is feasible for two or three but not fifteen. With the limited number of choices, it is a good idea to include an “Other” state. The states should also be clear and unambiguous, such as:

  • Walking: the operator’s legs are moving.
  • Working: the operator’s hands are moving.
  • Waiting: all the operator’s limbs are still.
  • Touching: One of the operator’s hands is touching the product.

Categories that are abstract and subject to interpretation, like “Value-added” should be avoided. Note also that an operator who is Working or Touching, may be handling the work piece or transforming it, and we don’t have enough categories at this level to make the difference. 

Timer Pro provides a method called “Non-stop timing,” in which the analyst simply clicks on a category when observing a state transition, and the time since the previous click is automatically assigned to this category. This eliminates the aliasing due to using 5-second intervals, and relieves one analyst from the task of clicking the right spreadsheet cell every 5 seconds.

Using videos to improve operations | Part 4 – Watching as a team

Starting with dissecting the video second by second is a recipe for getting bogged down and never finishing. The point of analysis at the gesture level is to answer quantitative questions about the operation’s improvement potential but, at the outset, we don’t even know what questions to ask. A video is a rich data source; we can zoom in anywhere in it, but first we need to identify where it is worth doing by viewing it end-to-end as a team, breaking it down into major phases, and stopping along the way to collect explanations, comments and ideas.

In a video of an assembly operation involving a machine, the phases might be:

  1. Place components on fixture.
  2.  Unload previous unit from the machine.
  3. Load fixture with new unit on the machine.
  4. Start the machine cycle.

If the video shows museum attendants setting up barriers of stanchions and belts to channel the flow of visitors, the phases might be:

  1. Bring carts of stanchions from storage to the reception area.
  2. Erect the barriers.
  3. Return empty carts to storage.

This first viewing of the video is an opportunity for the operators to see their work as a third-party would, which differs from the way they perceive it as they do it. The other participants in the review not only learn how the process is actually done, and to discover steps that are not in the specs.

One operator adds a cleaning step to the job of attaching stickers. When asked about it, she readily admits that she does it to fill up her time: her station has been assigned less work than the upstream and downstream stations, and she is embarrassed at having to wait. That comment identifies this assembly line as a target for rebalancing, and for the detailed analysis that supports it.

Another operator, as a fender subassembly station, is recorded walking to the front of the fender, running his hand over the surface and looking at it intently. This is not in the spec either, and he explains that he has noticed scratches, that he thought were generated when the front of the fender rubbed against the floor of the carrier in which it came from the paint shop. This is a quality problem no one else is aware of, and foam strips are promptly added to the carrier floors to keep it from recurring.

This viewing is also an opportunity to comment about sharp edges, heavy parts, missing tools, slippery surfaces, obsolete instructions, poorly located resources, etc.,  and to brainstorm on remedies. The observation that the operator below appears to spend half her time reaching for parts in distant bins leads a discussion of possible means of providing these parts closer to the fixture.

Automatic riveting station

In the next example, three museum employees move a cart of 16 stanchions from storage to deployment: one to push , one to steer and one to open doors in front of it and close them after it. The explanations are as follows:

  1. The cart is tall and the pusher can barely see over the top.
  2. The four wheels swivel freely, which makes the cart difficult to steer.
  3. Without a door-person to clear the way, the cart would have to stop and go multiple times.

Concourse setup Cinepak

Figure 2. Cart with stanchions at museum

 During the erection of the barriers in Figure 3, the team makes the following comments:

  1. The stanchions are held horizontally in the cart but used vertically, which requires extra handling of these 30-lb parts.
  2. Rather than moving carts around to where stanchions are dropped, the carts stay in one location, and the employees carry or roll them to their destinations.
  3. Pulling the belts between stanchions is done separately from setting them down.
  4. Some stanchions are not located properly and rework is necessary.

This slideshow requires JavaScript.

Figure 3. Erecting the barriers

 The ensuing brainstorm comes up with the idea of using smaller carts handled by just one person, with fixed rear wheels, each carrying the stanchions needed for one barrier, held vertically, and with the belts pre-connected. One door opener/closer suffices for the entire group. Each employee then erects his or her barrier by simply dropping the stanchions in sequence at the assigned locations, of which a map is mounted on the cart.

At the end of this process, we have many ideas, some of which can quantify by drilling into the details of relevant sections of the video. It may be hard to quantify safety or quality improvements, but we can measure how much time is being spent on tasks we propose to eliminate. We now have questions that a targeted detailed analysis can answer.

The frequency with which this work is done also plays a major part. The assembly work in Figure 1 is done every 17 seconds; setting up and taking down the stanchion barriers in front of the museum is done once a day. The stakes in improving both are obviously different. In the first case, it is using one operator instead of two, and having fewer quality problems at the end of the shift due to operator fatigue.

What about the museum? What does it gain if the barriers are erected in three minutes instead of nine?

The employees who do this work are from Guest Services, which means that they sell and process tickets at the gate, give maps to visitors, direct them to exhibits, restrooms, the store or the restaurant, and reunite stray children with their parents or teachers. They are the face of the museum to the visitors who represent the bulk of its revenue. The Guest Services employees contribute most to the visitors’ experience through direct contact, and there is therefore value in reducing the amount of time that they spend on tasks like setting up barriers, which do not involve such contacts.  If this were the only such task, improving it would make little difference, but they are involved in other, similar tasks throughout the day, to which similar ideas apply.

When brainstorming, you want to keep all the ideas already expressed in full view of all participants, to stimulate further idea generation. For future use, you want to record these ideas electronically, and linked to the corresponding video scenes. The first objective is met by flip charts on the walls all around the room; the second, by attaching comments to video segments using annotation software. If you use only flip charts, then someone will have to transcribe their content after the fact, which will add a delay and introduce errors. It is done faster and better during the meeting by having a dedicated scribe who is proficient in the use of the annotation software.

Unlike stopwatch time studies or motion studies based on predetermined time standards, videos not only allow us to do a coarse analysis of some sections and a fine analysis of others, but also to change our minds and go back to a segment we skipped over and examine it in detail.

Using videos to improve operations | Part 3 – Shooting shop floor videos

Following are a few recommendations on the art of taking shop floor videos:

  1. Special requirements on shop floor videos. We have already seen that the requirements for shop floor videos differ from those of other uses of this technology. If you shoot a family or sports event, you will naturally want the highest resolution you can get, which would be counterproductive here. Likewise, shooting a video for the purpose of data collection is different from doing it for art or entertainment.For example, the Youtube video of a NASCAR pit stop looks somewhat like a shop floor video but isn’t one. It is entertaining and dramatically shot, but not usable for analysis. In fact, a shop floor video that captures everything that is needed for analysis is likely to bore anyone who is not directly involved with the target process.

    This needs to be considered when deciding who will be holding the camera. You will naturally prefer someone who is already handy with it, and that is likely to be from experience capturing family occasions, sports, or from making movies as an amateur. The ability to keep a camera steady and pay attention to lighting, composition and focus is valuable, but the camera operator will have to be coached on the specific objectives of shop floor videos.

  2. Applications to setup time reduction or to the improvement of a work station. the camera needs to be looking down at the operator’s hands. In short operations, it can be done by holding the camera with a raised arm, and using the swiveling LCD screen for control. This gets tiring quickly and requires standing in such close proximity to the operator as to possibly interfere with his or her movements.
    Many plants have mezzanines or catwalks that provide a view from above. Being observed from such a place, however, may be uncomfortable for the operators, as well as too far to zoom in on the hands and capture any voice comments. The middle ground is to shoot from the top of a stepladder located within zooming and hearing range of the operator station, just far enough to avoid any kind of interference

    Amin recording operation

    Shooting a video from a stepladder.

    This works, until the operator leaves the station to walk beyond the reach of the zoom, at which point getting down off the stepladder to follow the operator while recording causes a few seconds of the action to be lots. A better solution is to hand over the camera to another team member on the ground, or even to involve more than one camera. In any case, this needs to be planned. Image stability is not an issue on the stepladder, but it is when following an operator’s movement across the floor, and you do not want a video that will make participants sea-sick during review. While professional tracking shots require equipment that is not available in a factory, some amateurs have supplemented the camera’s own image stabilization by shooting from a wheelchair.

  3. Fixed position on a tripod for time-lapse videos. Setting the camera on a tripod in a fixed position is not appropriate for this kind of analysis, but is when taking time-lapse videos of a large area for purposes of work sampling.
  4. Recording the position and orientation of the camera. It is also necessary to record on a layout of the shop floor the position and orientation from which the video is shot. The point is to return to the same location to shoot another video to document the improvements once implemented.
  5. Number of repetitions. Traditional time studies involve taking measurements on the same operation 6 to 10 times, for the purpose of improving precision when setting standards of operator performance. But our purpose in recording operations is not to set standards but to change processes to make the work simultaneously easier, safer, less error-prone, and faster.
    All we need for this purpose is one representative execution, and the operator can tell us if there is anything special or abnormal about it. If possible, we just take it into account during the analysis; otherwise, we make another recording. To make sure we have one complete execution, we start recording a few seconds before the operation starts and stop a few seconds after it ends.
  6. Scale. The presence of people in the videos gives us at least a rough sense of scale, but sometimes we would like more precision, for example to know how far an operator has to reach for a part, or how fast a cart is rolling. The following shots show the extreme measures the Gilbreths took for this purpose, with a gridded background. The picture also shows a large and precise timer, which was necessary because they used imprecise hand-cranked cameras.
  7. No editing. We do not edit the shop floor video, except possibly to add a title and administrative data at the beginning, Otherwise, we use it in the analysis exactly as shot. It is raw data, and we want to keep it that way.

Sports video analysis software used for motion studies in manufacturing

From the press:

The adoption of video technology in the improvement of manufacturing operations proceeds at a glacial pace. A recent article from the Financial Times describes the application of video analysis developed for sports to motion studies in manufacturing.
Via Scoop.itlean manufacturing
Dave Westphal’s 13-year-old daughter, Rachel, is a competitive figure skater. She is also the inspiration for a manufacturing improvement initiative at Nexteer Automotive, a leading US-based maker of steering and driveline systems for the car industry, where her father is director of lean manufacturing.
Via www.ft.com (You have to register on the Financial Times site to retrieve it, but a free registration will do.)

My comments:

Video technology is now so pervasive that it is nearly impossible to buy a phone that does not include a camera capable or recording footage that is good enough for broadcast news. Journalists use amateur videos to show storm damage or expose human brutality. We use it to identify improvement opportunities in business operations.

Motion pictures have a long history in manufacturing:

As is obvious from watching the Gilbreth films, where Taylor measured in order to control, the Gilbreths observed in order to improve. Taylor’s greater fame or notoriety, however, obscured this fundamental difference in the public mind, and made workers as wary of cameras as of stopwatches.

According to psychologist Arlie Belliveau,”The Gilbreths used workers’ interest in film to their advantage, and encouraged employees to participate in the production and study of work through film. Participants could learn to use the equipment, star in a film, and evaluate any resulting changes to work practices by viewing the projected films in the labs or at foremen’s meetings. Time measurements were made public, and decisions regarding best methods were negotiated. By engaging the workers as participants, the Gilbreths overcame some of the doubt that followed Taylor’s time studies.” In other words, these pioneers already understood that, unlike the stopwatch, this technology enabled the operators to participate in the analysis and improvement of their own operations.

Until recently, however, the process of recording motion was too cumbersome and expensive, and required too much skill, to be massively practiced either in manufacturing or in other types of business operations. In addition, most managements failed to use it in as enlightened a way as the Gilbreths, and manufacturing workers had a frequently well-founded fear that recordings would be used against them. As a consequence, they were less than enthusiastic in their support of such efforts.
Setup time reduction is probably the first type of project in which it was systematically used, first because the high stakes justified the cost, even in the 1950s and second because its objective was clearly to make drastic changes in activities that were not production and not to nibble a few seconds out of a repetitive task by pressuring a worker to move faster.

Technically, the cost of shooting videos has not been an issue since the advent of the VCR in the 1980s. Analyzing a video by moving forward and backwards on a cassette tape, while it appears cumbersome today, was far easier than dealing with film. The collection of data on electronic spreadsheets also eliminated the need to use counterintuitive time units like “decimal minutes.” Adding columns of times in hours, minutes and seconds was impractical manually but not a problem for the electronic spreadsheet.

With videos now recorded on and played back from flash memory, and free media-players as software, not only is moving back and forth in a video recording is easier, but the software maps video frames to the time elapsed since the beginning. We could manually transfer timestamps read from the bottom of the video player software window into electronic spreadsheets and have the spreadsheet software automatically calculate task times as the differences between consecutive timestamps.

While this approach has been a common practice for the past 15 years, video annotation software is available today, which helps break down the video into segments for steps, label them, categorize them, and analyze them.
You can also use it to structure the data and generate a variety of analytics to drive improvements or document the improved process through, for example, work instructions. Over the previous approach, video annotation has the following advantages:

  1. It automates the collection of timestamps. Reading times on the video screen and typing hem into an Excel spreadsheet is tedious and error-prone. Plowing through the details of a 30-minute is tedious enough already.
  2. Within the annotation software, each video segment remains attached to the text, numeric or categorical data you attach to it. One click on the data brings up the matching video segment.
  3. Using parallel tracks, you can simultaneously record what several people and machines do. Of course, you can do that without annotation software too, but it is more difficult.
  4. You can still export the data you collect and analyze it in Excel, but you can also take advantage of the software’s built-in analytics.

“Video time studies” is too restrictive a name for what we do with videos. It implies that they are just a replacement for a stopwatch in setting time standards. But what we really do with videos is analyze processes for the purpose of improving them, and this involves more than just capturing times. The primary pupose of the measurements is to quantify the improvement potential to justify changes, and to validate that they have actually occurred.

Putting this technology to use is not without challenges. Video files are larger than just about any other type we may use, be they rich text, databases, or photographs. And they come in a variety of formats and compression methods that make the old VHS versus Betamax dilemma of the VCR age look simple. More standardization would help, and will eventually come but, in the meantime, we have to learn more than we want to know about these issues. Functionally, the next technical challenge is the organization of libraries or databases for storage and retrieval of data captured in the form of videos. The human issues of video recording and analysis of business operations, on the other hand, remain as thorny as ever.