Release judgment criteria・Estimate the number of call center calls from the remaining bugs
Estimate how many call centers will be called in a month due to residual bugs
In the previous article, I introduced the idea of Father Gutara that even if there is a residual bug at the time of software release judgment , the release of the software is approved with little influence on the service provision according to the purpose of the software. In this article, let’s take a step-by-step introduction to how to estimate and calculate the number of phone calls to the support center , which is used as an index to measure the degree of impact on service provision .
First, I will introduce the entire calculation formula, and then I will explain the contents of the calculation formula in order.
This formula calculates the cumulative number of calls to the support center per month for one remaining bug . In the actual software release, if there are multiple remaining bugs, the number of declared calls is estimated and calculated using this formula for each of them, and the total value is used to determine whether or not the software can be released. Was there. I will explain how to make such a judgment again after explaining the calculation formula.
First, let’s calculate the estimated number of calls to the support center per month for one remaining bug.
Estimated number of calls to the support center per month (times / month) =
(A) Estimated number of problem events per month (times / month) X
(B) Declaration rate for each degree of impact X
(C) Country-specific adjustment factor
You’re just multiplying the three values A, B, and C, so let’s look at how to calculate these three values in order.
(A) Estimated number of problem events per month (times / month) =
(A1) Number of occurrences in 1 unit / day operation (times / unit / day) X
(A2) Total number of units in operation per month (units / day / month)
(A1) Number of problem events that occur in the operation of one unit / day (times / day) =
(A1-1) Probability that the conditions for occurrence will be met by operating one unit per day (times / day) X
(A1-2) Probability that a problem event will occur when the above conditions are met.
(A2) “Total number of products in operation" in the market per month (Taiwan) =
(A2-1) Monthly average number of units in operation X 30 (Sun)
First, let’s estimate and calculate the probability A1-1 that the conditions are met.
Somehow, it looks complicated. However, if you look at them one by one, I haven’t said that it’s that difficult, so please be patient for a while. First, I will explain the contents of item (A). For now, forget about items (B) and (C) as they will be explained later.
First , (A) Calculate the estimated number of problem events (times / month) per month . This is easy, just estimate and calculate the number of problem events that will occur if one product operates in one day in (A1), and multiply this by the total number of operating units per month in (A2). is.
(A1) The calculation of the number of problem events that occur in the operation of one unit / day is further decomposed into the contents. ) And (A1-2) The probability that a problem event will occur when the above conditions are met is divided and calculated by multiplying them.
First of all, (A1-1) is the probability that the occurrence conditions will be met by operating one unit per day . When some problem occurs, it is usually when multiple conditions overlap. For example, let’s consider an example in which a problem occurs when a button to start another function B is pressed while using one function A. And let’s assume that the average user uses this device for an average of one hour a day.
Then, (A1-1) The probability that the occurrence conditions will be met in one day’s operation can be calculated by multiplying the following three establishments.
a) Probability of using this service in one day (24 hours). For example, assuming that a typical user uses this feature for one hour a day, the probability that the service will be used is 1 (hours) / 24 (hours).
Next b) Assuming that a typical user uses function A for 10 minutes while using the service for 1 hour, the probability of using function A while providing the service is 10 (minutes) / 60 (minutes).
Finally, c) If the probability of pressing the button to start function B while using function A is once every three times of function A, it is 1 (times) / 3 (times).
In this case, (A1-1) 1 unit, probability that the occurrence conditions will be met in one day operation = 1 (hours) / 24 (hours) X 10 (minutes) / 60 (minutes) X 1 (times) / 3 (times) ) = 1/432 = 0.0023.
This (A1-1) is the probability that the logically determined conditions for occurrence of a defect are met . And, apart from the probability of occurrence that can be calculated logically, there is a problem event that occurs very rarely when it is repeated tens or hundreds of times, although it is difficult to explain it logically . This corresponds to the case where the internal movement of the software is affected by some timing and the problem event occurs or does not occur.
Estimate and calculate the probability A1-2 that a problem event will occur when the conditions are met.
The item for incorporating the occurrence frequency of this timing event into the calculation formula is (A1-2) the probability that a problem event will occur when the above occurrence conditions are met . If it happens all the time, it’s 1.0, and if it happens once in 1000, it’s 1/1000. This value may be calculated by desk calculation when the related timing is simple, but in many cases, it is often calculated from reproduction tests performed thousands or tens of thousands of times.
The total number of units in operation per month in (A2) can be said to be a calculation item that adjusts the quality required for determining whether or not to release depending on the number of units in operation in the market . To put it plainly, if the same problem occurs when only 100 units are operating in the market and when 1 million units are operating in the market, the latter case is 10,000 times the former. It is important because the number of service calls will be generated.
Now that you have calculated the A1 item, the next article will introduce you to the remaining A2, B, and C calculations.