I was hired by Callaway Golf and relocated from Milwaukee to San Diego in January of 2001 for a Research Scientist Position. This was an R&D position within the Research – Advanced Technologies Group that was in support of the Product Player Matching initiative. The original job description had 3 main areas of responsibilities:
- Will lead the application of science based analysis and measurement techniques to the development of new and improved golf equipment.
- Develops and implements new test methods for understanding the science of golf.
- Supports the design, development, and evaluation for new golf club products, conducts research planning, and guides development based on scientific knowledge and golf industry experience.
The Product Player Matching initiative consisted of 3 specific synergistic focus areas. They are listed below with the essential job duties and responsibilities for each:
- Player Profiling – principal contributor to the characterization of players based upon swing, impact conditions, and resultant golf ball trajectories; perform data reduction and statistical analysis of test results and conduct comparisons with simulation results to characterize golfers based on physical measurements, player, club, and ball kinematics, and/or kinetics of the player and club; lead the development of future golfer characterization measurement systems based on accurate interpretation and communication of analytical and experimental data.
- Digital Human Modeling – responsible for leading development of analytical and experimental tools for the design of golf clubs based on extensive relevant knowledge; use experience and demonstrate foresight in improving tools and processes to better understand the interaction between golfers and golf equipment; develop a unified theory of player-club interaction based upon simulations and experimental regression models; invent golf club design concepts that leverage knowledge of golfer’s individual characteristics.
- Club Fitting – responsible for leading development of analytical and experimental tools for the fitting of golf clubs; lead consumer and customer insight research in support of customized products; write reports on strategic opportunities and threats relevant to player characterization, player-club interaction, and customized products; plan for and lead player and robot testing in support of customization efforts.
When I arrived at Callaway, they had a number of advanced measurement systems that were already purchased or developed in-house, including:
- NDI Optotrak – an optoelectronic optical motion capture system. This was going to be used for swing analysis studies in the soon to be developed Player Performance Bay in the RCH Test Center. This system was already purchased just prior to my arrival but was not yet installed.
- On-Board Diagnostic (OBD) Shaft-Based Measurement System – the Callaway Measurement Systems group had developed a number of self-contained, tetherless instrumented clubs that contained instrumentation capable of measuring, processing, storing, and downloading data for golf swing analysis research studies. The shaft-mounted instrumentation system was a very complex data acquisition system capable of performing many of the same functions as a laboratory based data acquisition system. When I arrived at Callaway, the Measurement Systems Group was on version OBD II. The original OBD I version housed accelerometers in the head of the club, strain gages at the tip of the shaft, and also had a longitudinal rate sensor in the butt end of the shaft. The OBD II version eliminated the accelerometers in the club head and the longitudinal shaft rate sensor, and had strain gages at both the tip and butt end of the shaft.
- Callaway Performance Analysis System (CPAS) launch monitor – this was a dual camera based 3D launch monitor system that could measure and output both club head and golf ball kinematics. There were a number of different variations of launch monitors that Callaway R&D designed and developed. Some were specific to R&D use for Player Testing, while others were developed for Pro Tour and Club Fitting use. All were dual camera based systems, but some were on vertical stands while others were built in portable units. The original CPAS systems were designed for capturing and analyzing golf club information at impact and post-impact ball flight. The golf club impact information included golf club head orientations (dynamic lie, loft, and face angles), golf club head velocities (head speed, path and attack angles, and downrange information), and golf club spin (ball velocity, launch angle, side angle, ball orientation including true spin, back spin, and side spin) characteristics.
These systems could all provide a lot of data to be used in support of any of the 3 focus areas of the Product Player Matching initiative. CPAS was a very powerful launch monitor system that provided impact conditions and initial ball flight data for all player and robot testing. In my world, this was very powerful data but only represented a very small time frame of the dynamic swing and only provided results for the result or effect of the swing. I needed tools and systems that could provide information on the entire swing and provide insight into the cause of the resultant CPAS data. OBD II could provide that information in the form of measured strains at both the tip and butt end of each instrumented shaft; from that information, all 6 independent loads and moments could be calculated. Once the motion capture system was installed and setup, we would have kinematic data for all body segments for any golfer that was suited up and evaluated in the Player Performance Bay.
That is a ton of data, almost an overwhelming amount of data. When I came to Callaway, the OBD analysis was affectionately referred to as squiggly line analysis. People were used to seeing the dynamic outputs from the OBD; they just didn’t really know what they meant or even provided. The following is an example of 2 of the 6 channels from the butt end of the shaft (2 of 12 total channels per swing) for 2 different swing types.
One can imagine how much data there would be for all 12 channels for a swing typically ranging from 1 – 1.5 seconds in length. That doesn’t even include player kinematic data yet from the motion capture studies. It was my job to make sense of all of this data and meet the job requirements outlined above.
At the time, Callaway like many other companies was implementing Six Sigma philosophies and processes throughout the company in an effort to instill data-driven decision making at all levels. However, one of the key philosophies of Six Sigma is you have to analyze the right and appropriate data in order improve any process. That is the big question that I had with the existing technologies that had been developed – were we even analyzing the right data?
While all of this technology was very impressive (and very expensive), they were all projects that were setup independently by R&D personnel that had no need to collaborate with one another as there was really no end user for all of it prior to my arrival. So I started digging into the data knowing that there were some issues I had with how each of the areas of research was currently configured. The main issues I had with current status were as follows:
- OBD II: while an impressive engineering feat, there were a number of problems from my standpoint as an analyst of the data.
- There was no database put in place to handle the vast amounts of data; rather it was currently configured to download the data into Microsoft Excel for analysis.
- Each club was a dedicated instrumented club as strain gages had to be applied which limited how many clubs it could be tested on at a time; for R&D purposes this was fine, but it would limit functionality of the technology for any club fitting efforts.
- Most importantly, while it was providing dynamic swing data, it was providing me with the effect data in terms of strain levels on the shaft. Through mathematics, I could calculate loads and moments at the shaft. However, I wanted the cause (input data) at the player-club interface – how was the player actually moving and loading the club.
- NDI Optotrak – the motion capture system was not even setup and I had some reservations about the system; it was not the system that I would independently chosen, but it could still provide meaningful data if configured properly and used the right way. Some of the main limitations included:
- No real-time video playback capabilities; all swings had to be played back after data collection and only in marker triad form.
- Data collection required a lot of markers in cluster designs in order to minimize data occlusions to the 3 cameras and to insure 6D rigid body measurements that were needed for each body segment.
- As it was an active optoelectronic system, it required cables and strobe boxes to be put on the player during the swing. Subject setup took quite a bit of time as a result to insure that cables were not flapping around during the swing. This required lab-friendly golfers to use the system.
- Reduced overall sampling rate due to the number of markers needed as described above with this active optoelectronic system.
- Swing Model – there were some swing modeling efforts already put in place with an external consultant. However, there were a number of issues with the format of this program:
- The cost was extremely high because of the software used. While very powerful, it was not a great choice for the proof of concept work that was being done. Even if it was successful, it could never be implemented in an in-house design program because of the cost constraints.
- Using an external consultant we were always dependent on their time constraints. While they did a very good job and provided good results, it was difficult to coordinate the program management of all of the synergistic studies tied to Player Product Matching initiatives.
- The biggest issue I had was how they would handle the motion capture data. Their process treated the data as motion influencing; I wanted the data treated as motion governing. We were going through tremendous efforts in subject setup and calibration to provide individual motion outputs to feed into a custom dynamic swing model. I wanted to have high correlation coefficients between measured 3D rotations and the simulated output 3D rotations. This was not possible based on their implementation.
So one of the first things I had to do was to address these problems and provide better alternatives. It was not any easy discussion as the initial shareholders had their motivations for why it was setup that way initially. By suggesting alternatives that provided increased visibility to non R&D personnel, more accurate data results, much lower per unit costs for new implementation, improved scalability for testing and future club fitting efforts, and much lower analysis costs, my managers eventually agreed that the changes were necessary and beneficial. It’s always a hard sell to Senior Management that the high capital costs to date did not produce any tangible results…especially when they were not my choices to begin with.
While this was a rather delicate discussion, thankfully I had very bright technical managers that were also very good leaders. They empowered me to make well thought out decisions and agreed with my plans moving forward. While much work had been done to provide the first data collection tools, much more work was necessary in order to achieve the Product Player Matching goals. A big factor for consensus, was the idea that big data did not necessarily mean the right data.
The changes that were necessary and were made moving forward were:
- OBD II CIS – change the instrumented clubs from dedicated strain-gaged shafts to a modular shuttle design using a 6 DOF and 9 DOF custom inertial sensor (CIS) designs. Details of the design changes and examples of the outputs can be found at the OBD II CIS link.
- Electromagnetic Mocap – Purchased an electromagnetic 4 sensor golf system for real-time analysis capabilities for our professionals and VIPs in for club fitting and swing analysis. This was a much lower cost system but provided real-time visualization and biofeedback capabilities in a much short swing session when compared to the NDI Optotrak sessions used for Digital Human Modeling work.
- Swing Model – brought dynamic swing model development back in-house for Digital Human Modeling work. Used external consultants for some development in order to provide more time efficient results, but maintained invisibility wiht our external consultants for the separate work so that we did not lose any secrecy of our R&D efforts.
- RBC CPAS – independent of the changes that I had requested, our Measurement Systems group had produced a rigid body code (RBC) CPAS system. The RBC system was a dual camera system like the original CPAS system, but it measured pre-impact club head conditions and predicted ball flight from the measured conditions. This was much more in line with my Product Player Matching work which was aimed at developing tools and simulations based on input conditions to predict the output conditions.
Taken together, all of these changes provided an extremely powerful set of advanced measurement technologies necessary to address the requirements of the three focus areas of the Product Player Matching initiative. All of the systems were now set in place to develop dynamic simulation capabilities using measured swing input data to predict resultant ball flight conditions. Using multi-sensor swing studies (MS3) we could measure and collect the necessary data for both modeling inputs as well as data for validating simulation results. More details for subsequent work and results for Player Profiling, Digital Human Modeling, and Club Fitting can be found at the respective links.