Step 2. What are we measuring? Are we measuring it well?
Not everything that can be counted counts and not everything that counts can be counted. (Albert Einstein)
In the previous step, we discussed ways to move from the goals and objectives of the business to unbiased, specific and measurable laboratory objectives. Now we
need to make key decisions on what to measure at each stage of the experimental project so that each data point will advance the laboratory and business objectives. For that to happen each data point must be accurately measured – so your measuring tools must be precise and accurate!
HT Measurement Flowdown
The easiest way to assure this is to use the Flowdown tool discussed in Step 1. By adding a “Measurements” column to the right, you can assure that each of your flowdown goals is measured appropriately.
The peril is that of measuring something that is easy to measure, but not important to the customer. Remember the drunk searching for his keys under the lamppost – because the light is better there?
At each step in the chart, we must ask what is being measured – and does this measurement relate to the measurements higher up?
An example for GE’s Diphenyl Carbonate (DPC) process might be:
|General public||Better, cheaper plastics products||$/lb|
|GE Plastics||Lower cost raw materials||DPC process vs. Cl2-based process|
|LEXAN business||Cheaper route to DPC||Process efficiency|
|LEXAN research||New DPC Catalyst||Yield, throughput|
|Process R&D||New search strategy||Size & complexity of spaces searched|
HT Measurement Strategy
You can increase your chances of picking the right HT measurement by using the following measurement strategy.
High throughput experimentation puts a whole set of new constraints on measurements. HT screening is often envisioned as a funnel (FIG) in which starts with a very large number of samples that are tested for only the one or two most crucial responses. The winners are then moved to a (still rapid) secondary screen, which then generates candidates or “Chits”. In planning the measurements for this funnel, speed is of the essence. The pioneers of HT expressed it as “Analyze in a day what you make in a day.” (P. Cohan Symyx Technologies).
If the measurements are truly high throughput, they cannot be hand checked. A system must be in place to bring “Six Sigma” quality to the measurements. For example, a chromatography system that relies on a single internal standard is unlikely to be good enough. In the DPC project, multistandard calibration with an AI peak recognition system was used.
What kind of measurement are you making? Is it:
- Binary – an on-off, yes-no measurement. It broke or it didn’t.?
- Subjective – a one-to-five estimate of a quality like a popcorn taste test?
- Integer – counted defects like the number of delaminations in a sample of plywood?
- Real numbers – actually measured with an instrument, like percent sulfur in an oil sample?
Each of these types of measurement requires a different experimental plan and data analysis!
Remember: Information… is a difference which makes a difference. (Gregory Bateson)
You now have a DOE design that
- is relevant to the organization’s goals and objectives, and
- captures the right measurements
In the next section, we’ll look at the independent variables (factors).
If you want to jump right to the whole strategy, contact me at +1 413 822 5006 or email@example.com!