0.8 Taking the speak n spell into production  (Page 3/5)

Selecting the testers

The next issue was choosing and qualifying the testers who would live out their days in the dog houses. We took the time to do hearing tests on all of the production people to help us pick the appropriate testers, just as we had done for the editing and QC staff who had created the vocabulary. But, once we got the results of the tests, we asked ourselves a hard question: “Do we chose those with average hearing, good hearing, or great hearing? We ended up throwing out the test results and picked people who seemed to catch on quickly to what we were looking for.

This did backfire for us one time. On a particular day we were having units make it through test that weren’t talking. This was an issue as they should have been caught by the testers. So, as part of our trouble shooting method, we asked each of the testers how they were testing the units. The first couple of testers explained what they were doing and it was exactly what we had told them to do. When we approached the last tester to ask how they were doing the test, the tester didn't respond to our request. Yes, the person was deaf. So, we learned that, although the tester didn't need to have perfect hearing to do the testing, they did need to be able to hear. Go figure!

Consistent quality

Finally, the familiarity issue was the hardest to solve. Typically, a quality issue on a production line became easier to detect the more one became familiar with it. For example a scratch in the plastic case became easier to spot with experience. In the extreme, the scratch goes from unnoticed to an issue to a reason to shut down the production line in a matter of days. But, just the opposite happens with speech quality.

Several years prior to the Speak N Spell, I was on a desktop calculator production line. We had on the product in production, the TI-450, two different shades of gray keys to distinguish the numbers keys from the function keys. One day the line came to a halt because it was noted that one of the keys had yet a third shade of gray. This quickly became a reason to shut down the production line until the issue was resolved. I decided to run an experiment to determine the severity of the problem. I took one of the offending calculators out into the hallway of the building (the S/C building in Dallas for those of you who are interested). I stopped people at random and asked them to tell me what was wrong with the calculator. No one noticed anything wrong. I further pointed them to the keyboard as the area where the problem existed – still not an issue. Finally I would point them to the offending key and the one next to it. At this point most, but not all, would say “oh, they're not quite the same color”. That is when I learned how familiarity increased awareness on a production line. Once pointed out, the color difference was obvious.

But that wasn’t so for voice. As we had learned from creating the words, familiarity decreases the sensitivity to quality rather than increasing it. Our solution was to continually change the people doing the testing. As you might expect, this was well received by the testers who were spending all day in the dog houses. You might say we solved two problems with one solution. We rotated the testers to keep them fresh – where fresh had two different concepts. It also added to the variety of comments written on the walls of the dog houses.