Cost of not testing your solution will propagate and snow ball

Today I've a got a bit of relief at work, as we will have some free time for the next two days.

Today I will settle some things after the house moving. One of them is exchanging bulb. We need to replace 10 bulbs on the old rented house that has been damaged during our 2 years tenancy. The store didn't provide a testing platform for the bulbs, yet they provided one-to-one replacement guarantee. I bought 6 x SL twisted neon bulb, 2 x square-based SL neon bulb, 1 x 32w round TL neon bulb, and 1 spot light. From the 10 replacement bulbs I bought, one of them didn't really worked (defect). The store promised me for replacement when some of them are defect products. "Just don't throw the boxes, they said". Today is Monday, I bought the replacement last  Friday. Because I don't come to office on weekends, I have to postpone visiting the store to reduce cost (while losing time).

Hmm, even though I have one-to-one replacement guarantee, I still have to go down town to the store I bought from, and it's only just for this one bulb! For the replacement of this single bulb I have to go down town to the store near my office on off working day, pay the MRT (train) fare there, waste some time there, and then pay the fare for returning home again.

This event reminded the cost of not testing your solution thoroughly. Once it's rolled out, the impact wouldn't be just what you will feel at your coding room where you can simply fix, compile, and run again. The cost of replacement stuffs in production will create a huge snowball as it goes down hill. The issues, the delays, the lags will be propagated far.

Where as:

• we are not perfect -- we produce bugs
• the requirement guys are not perfect -- they produce defective or incomplete or inconsistent requirements
• our customers are not perfect -- they might ask us to do some impossible or contradicting stuffs

I believe it still better when we do test the right way before we roll out to production.