Related to a well-defined objective is an actual sample of the key problem area response times. This will become your baseline. A baseline is used for comparison. For example, when someone says, "the query now runs 50% faster," you still want to know if the query runs in 2 seconds or 2 hours. It's similar to a commercial stating something contains "35% less fat!" A statement like that screams for a baseline, and is clearly intent on deceiving and manipulating. Don't fall into the same trap.

If you are not sure what the key business transactions are, then ask your manager. Even better, set up a meeting with the user representatives and ask their priorities. Keep it limited to just a handful of transactions. Let everyone know that initially your team must focus on the absolute business-critical transactions and nothing else. Reassure them that after these transactions are running well, you will move on to another tier of performance analysis.

After selecting the transactions, take multiple samples. It is best to have users run the transactions and have them watch you write down the timing result. This will increase their trust, which at this point is probably at an all-time low. You don't need to take 30 samples-3 to 5 will be enough to give you a good idea of the situation.3 Then write these down and perhaps post them so everyone knows the baseline.