You're right. If the product doesn't perform as expected after researching it, get a refund.
As for numbers that don't lie, I'm unfamiliar with Crittercisms methodology so I won't throw a rock.
But my job is to analyze economic / financial data, and interpret it per a clients request. In 36 years I've never encountered data, private or public sector, that couldn't be manipulated to obtain a desired outcome. The canvass is blank. You paint the picture you want. I know because I do it. You can look at identical data from multiple angles and reach different conclusions. The final result is the one the client is paying for.
I do know this. My adult daughter has a Samsung Galaxy II and her Android apps crash or glitch with such frequency, I need to pick up some drywall for her place. My iPhone 4 (got it the week it launched) runs efficiently - and I experience few app crashes. My iPad 2, which I've owned since August, performs just as well. If find Safari browser a bit twitchy, but not objectionable. At least for my purposes.
AA
Somewhere between selling Megahertz Corporation to U.S. Robots then to 3Com and finally to MSL, we had to explain our data gathering methodology. We had to explain how we interpreted our data. Those that were buying us wanted to confirm our numbers and it was a very serious effort to capture as much data as possible. We had to look at who used the smoking area and how many times they passed through one area into another. We had to look at every employee and put a value on their contribution. We had to move crap that went directly from the production line to rework. In other words, we build it, fail it, and send it to rework where they tried to fix it and quite often, stored it away in a little room. We discovered that people were hired to repair these devices, but they were discontinued long ago.
We hid things like that from the new buyers so the buyers would not ask why 250,000 or so units--some still being reworked, yet discontinued a year before--were stored in the old part of the building. Later, someone discovered that you can look at the number of PC Board we purchased and compare that number to how many we shipped. There was a serious investigation because we were only shipping something like 40% of the products we supposedly built. Without data, we would never had known.
At the time, this data collection was a serious priority and scary because careful study showed us to be very bad at what we did. Or that is what the data would lead you to believe. Most of us still did a great job. But we could not hide from the data.
In one case, we were having huge production run pass rates. We were successfully passing huge numbers of modems through test with a 98% (or so, I forget) success rate. Far too many; at the time, we knew something was going on because we did not always get it right. One particular supervisor wanted our test programs modified because his crap was being failed.
What we found out later was modems manufactured on production lines with the highest numbers were being returned to RMA. In some cases, modems went through test and into shipping and returned to us not long after the customer discovered connection problem. RMA started telling us that they were seeing modems without PC Boards inside the cases or failures caused by integrated chips put on wrong, or resistors/capacitors missing or traces routed at the edge of the board.
I still have a modem I used for training. It had almost every component replaced several times. The Rockwell chip sets were replaced half a dozen times. Parts were added to pads that should not have been added and pads on this board were gone; it is what happens when you re-solder a connection multiple times. I calculated the shop time and cost of components and it was a multi-thousand dollar modem. The problem was damaged traces at the edge of the board. These issues came to light when we delved into the data. And it was one of hundreds of products that cost us thousands to build and sold for a fifty bucks at the time.
When these issues were brought to light, suddenly, good numbers were suspect, and rightly so. When you make 500 units and 450 units are returned, suddenly, your production lines look incompetent. And rightly so.
I sometimes think some of the most scariest things can be found lurking in the data. Like comparing RMA numbers to test pass rates. My problem was I always overlooked something and I had problems discovering what was buried in the data. I was not a trained in data analysis. We were likely worse than we/I thought. Or perhaps better.
