Special Report: File-base workflows and what you can learn from Toyota

In the second part of his article about QC in the media industry, Bruce Devlin, CTO for AmberFin, looks at how other industries, in particular the automotive industry, have successfully implemented process automation as part of quality control.  Here, he argues that much can be learned from companies such as Toyota and that their revolutionary QC concepts can be applied to a file-based media workflow. By applying some of the key principals of waste reduction and monitoring the quality of processes, a similar revolution in the way we QC media may be possible.

Historically, media facilities were built around people – there was little if any automation; operations were tape-based, people carried tapes around the facility and problem solving tasks were achieved within peoples’ heads. Fundamentally, this is an inefficient means of operation.

Today, much of this media supply chain is not heavily automated. However when achieved, effective media process automation transforms the way in which your facility operates. Unlike our industry, car manufacturers have successfully employed process automation for many years. One of the most successful has been Toyota.

The essence of Toyota’s car production workflow centers on three concepts:

1 – Waste reduction – never do anything that doesn’t result in a better car.

2 – Jidoka – automation with a human touch. Toyota’s production philosophy is simple – only do what is needed, when it is needed and to a level that is needed. Connected to this is their view on analysis: implement just enough analysis, just in time and highlight/visualize your needs.

Toyota’s focus on quality is not so much on the quality of the end product but rather on the quality of the production processes that contribute to the production of the car. They believe that if you put perfect materials into a perfect process, you make perfect cars, perfectly.

When Toyota identifies a problem in the production process, it will stop the production line and fix the fault so that all the vehicles coming off the production line are top quality. Compare this with the traditional US or European maxim of “keep the production line going at any cost” and then retro-fix the sub standard vehicles at the end. A look at how these two philosophies have prospered in recent decades, since their divergent QC approaches have been employed, demonstrates which approach is more successful.

3 – Andon – means taking just one glance to see how the entire production system is operating. This typifies the change in methodology from performing full QC on the completed product to a situation where you perform QC on each process within the production line.

In the media world, we currently re-try and re-work failed media files. If this were the case in car manufacture, the industry would fail within a very short period of time. What we need to do is to take the QC central elements of car manufacture automation and integrate them within file-based media workflows to create Unified Quality Control.

The goal of “Andon” is to see the whole process in a single visual action. Creating a unified QC timeline gives an accurate and easy to use display of potential issues of any kind such as simple video and audio problems, file wrapper abnormalities, artifact detection, PSE (Photo Sensitive Epilepsy) Flash detection, loudness violations and potential content-related editorial issues. QC processes can be implemented at any point in the lifecycle of an asset, using the most appropriate technique. Simple and accurate presentation of the information helps operators to easily identify the source, nature and position of an error with a thumbnail of the frame where it occurred.

In the media world, we still tend to QC the individual files and not the processes that made those files. By measuring QC before and after a process, and storing the results with the asset, a Unified QC workflow creates knowledge about how a process affects the media. For example: The end of this file is black. Both an upstream SDI tool and a downstream file based tool can measure the video and know it’s black. A human operator can convert the status of these automatically generated events into an “OK” state, because only the human knows that the black frames were an integral part of the creative content. This human judgment has now been captured in the QC report so that the QC of a downstream transcode can automatically perform one of the following actions:

a) if the downstream transcode detects black in the same location in the file, it can automatically propagate the “OK” decision to the downstream QC task via a MAM system.

b) if the downstream transcode does not detect black in the same location, it can flag an error that something has gone wrong in the transcode. In any case, we have the ability to use knowledge about the media so that the absence of a QC event is the error. This can only be done if the process is being Quality Controlled and not the file.

Since QC requirements vary from facility to facility and from project to project, each application requires a different level of manual intervention in order to optimize the level of “business insurance” that the QC process delivers.

Employing the principle of Andon and Jidoka, a single unified user interface provides the operator with a feeling of how good the media is. Automated devices continually make measurements but it requires a human touch (Jidoka) to make the judgment that can be trusted and propagated to downstream processes.

To read the first part of Devlin’s report click here.