9.4 Creating Estimates

< Day Day Up >

Unfortunately, there are no tricks to solid estimating other than covering the bases as thoroughly as possible. My personal expertise, if you want to call it that, is in the hardware and networking infrastructure arenas, which are simpler than some activities because the bulk of expenses are components with known costs. With this sort of estimating, you or your team can work closely with the manufacturer or qualified reseller to make sure that all:

The right parts and pieces are considered in proper combinations and quantities.
Extras such as cables, memory, software, engineering support, and maintenance are included.

In the area of labor estimating, making an educated guess and then doubling it is not a bad way to start. For example, if I were to recall that it takes 2 hours to upgrade an old personal computer (PC) with random access memory (RAM), a new hard drive, and a new operating system, from an estimating standpoint I would figure twice that, or 4 hours.

Why is that? Simply put, worker productivity is far short of 100 percent. Time is lost every day with meetings, vacations, tardiness, chatting, and shopping on the Internet. Because workdays are finite (i.e., people tend to go home at a regular time unless they are under the gun), you recognize that in your estimating by assuming longer work cycles than logic might suggest.

I do have a simple algorithm that is useful for sizing resource requirements; in fact, it is also helpful for planning purposes. It is depicted in the following illustration as an unadorned arithmetical exercise of deducting weekends, holidays and so forth from the 365 days in a calendar year to show maximum hours available for work. It ignores overtime. I provided columns to distinguish between 7- and 8-hour workdays, because organizations are not consistent in this way. You can make your own table if you have different work rules. Take a look at Exhibit 2, and then we will see how it can be used.

Exhibit 2: Resource Availability Rule of Thumb

	8-Hour Workday	7-Hour Workday
Total days in year	365 days	365 days
Less holidays	-10 days	-10 days
Less vacation	-10 days	-10 days
Less weekends	-104 days	-104 days
Less personal days	-5 days	-5 days
Net available days per year	236 days	236 days
Net available days per month	19.7 days	19.7 days
Gross hours per year	1888 hours	1652 hours
Less 1-hour lunch and break per day	-236 hours	-236 hours
Net available hours per year	1652 hours	1416 hours
Net available hours per month	138 hours	118 hours
Net available hours per week	31.8 hours	27.2 hours
Net available hours per day	6.4 hours	5.4 hours

In an 8-hour workday environment, you can anticipate, for each resource, the following availability factors.

Workdays: 236 per year or 19.7 per month
Work hours: 1,888 work per year, 138 per month, 31.8 per week, or 6.4 per day

Now, I will provide the promised useful application of all this. Suppose a team lead requests that you fund a consultant for a deliverable estimated to take a thousand person hours. Most people would assume a 40-hour workweek, divide that into 1000 hours, and come up with a calendar time of 25 weeks to get the task done. That is represented by line 1 in Exhibit 3. Line 2 documents how I calculate the resource requirement. It is based on the previous table, which, to repeat, says there are 31.8 work hours available per week over the long term, not the 40 hours that people generally assume. For dramatic effect, I have postulated that the consultant to be hired for this function is available at a rate of $750 per diem. The third line shows the delta resulting from the contrasting resource availability assumptions under consideration.

Exhibit 3: Total Hours in a Real Workday

Line No.	Hours Needed	Hours per Week	Weeks Required	Days Required	Cost per Diem	Extended Cost
Line 1	1000	40	25	125	$750	$93,750
Line 2	1000	31.8	31.5	158	$750	$118,500
Delta	n/a	-8.2	6.5	32.4	n/a	$24,750

Compared with the traditional 40-hour workweek assumption made in calculating labor durations in line 1, my method in line 2 says you need 6.5 more weeks, or 32.4 more work days, at an added cost of nearly $25,000, to engage that resource to complete the 1000-hour task.

Remember the reality of engaging resources. These are not day laborers you send home due to inclement weather or some other condition that would render them temporarily unproductive. In the professional environment, they are on your payroll for the duration, ^[3] and your project coffers will bleed that money accordingly. Based on that workplace reality, it makes sense to have a realistic estimate of that cost. This is how I thumbnail the cost, assuming, of course, that the team lead's estimate of 1000 hours is credible.

While we have this tool in our hand, let us look at another way to use it, even though it is slightly off topic. Suppose we add a time constraint to the thousand-hour task scenario, specifically that the work has to be completed within a 7-month (i.e., 28-week) window because of some milestone or dependency. If you used the 40-hour workweek model given in line 1, you would say, "No problem, I only need 25 weeks, so I have 3 weeks to spare." Using my approach, you would say, "Oops, I need 31.5 weeks, so even if my 1000-hour estimate is dead on, I would be nearly 4 weeks late." If you buy into this model, you would then either work overtime with that consultant or add a second professional to make sure the work is done by your "must finish by" date.

You can shave this process any way you like. What I have recommended has worked fairly well, given the imperfections of the workplace and the occasional intrusion of misfortune. Intuitively, it would appear that the tardiness and cost overruns endemic in large IT projects are partly due to managers underestimating the time and dollars required to get work done well in a timely manner. Counting routers can be far simpler than guesstimating time. Hopefully, the thought process we have just discussed will minimize your exposure in this area going forward.

^[3]Most would walk out on you if you tried micromanaging their time sheets in this manner.

< Day Day Up >