LoE

[icecream-guy]

How does one determine the Level of Effort for estimating the number of hours needed for a project?

say customer wanted to move 3 or 4 VLANS behind a firewall,
or wanted to connect a new router for a new C2C VPN,
or say needed some traffic analysis

I've never worn these shoes, I am not an estimator, but the task comes with the new Team Lead.
so I need to figure out how to estimate.

[wintermute000]

break down the tasks
assign mid-range hours per task
contingency 10%
admin overhead 3%
profit

e.g. new router

- design 16 hours
- BOM 1 hour
- place order 15 minutes
- draft config 2 hours
- draft change 2 hours- take delivery, unbox, stage 1 hours- submit change 15 minutes
- change meeting 15 minutes
- peer review 30 minutes
- 3rd party coordination 1 hour
- implement change 120 minutes after hours x1.5
- CMDB update 15 minutes
- doco update 15 minutes

[icecream-guy]

so I need a SoW before the LoE...

[deanwebb]

so I need a SoW before the LoE...

Yep. But you need LoE estimates to include in the SoW, it's a vicious circle. The LoE will tie into the SLA, which is why I love "best effort" the best. That means if a reboot doesn't fix it, oh well.

[Otanx]

Requirements. That is what you need. So I hear. Never actually seen a requirements document that details what is needed. You take that and build out the high level steps. Estimate how long each will take, and who is involved. Then build a SoW, and project plans from that.

How to do the actual estimation is experience and is very dependent on each organization. One organization my take a week to deploy a new switch with all the change control, manual config, rack stack, cabling, etc. All done by different teams. Another organization it may be 2 hours with no change control, and one guy goes down and does the work.

-Otanx

[Dieselboy]

This is why I've always enjoyed working for a tiny (<50 person) SME. The AM walks over to my desk and says "how long will this take, roughly" and I say "ooh 2 hours, 3 hours max" and they disappear and return some time in the future with a job 

[icecream-guy]

ended up figuring it takes me 30 mins to research, develop and deploy a single firewall rule,
so I estimated 200 devices, x5 rules each,  1000 rules  ~500 hours,  plus time to research develop and plan,
discussion TRB CCB processing, etc.

[Dieselboy]

Another way to think about it is, you do 30 minutes research and 10 mins documentation then future effort doesn't require so much research when you have a validated design (copy/paste)

[Otanx]

To add to what Dieselboy said you will find that some tasks can be done in bulk faster than doing the task again. As an example it takes 6 hours to get a server build done. This is installing the OS, patching, installing applications, etc. However, if I need to build 5 servers the time it takes isn't 6 x 5. I can start the OS install on one, and move to two and start the OS install. Then on to three etc. Once I get server 5 started server 1 is waiting for input. So I can say 1 server takes 6 hours, and up to 5 servers takes 8 hours.

-Otanx

[wintermute000]

iTs super hard to estimate correctly esp with the factors mentioned. Its an art not a science esp. when you consider different individuals do things at a different pace...

[deanwebb]

iTs super hard to estimate correctly esp with the factors mentioned. Its an art not a science esp. when you consider different individuals do things at a different pace...

That's why you need to follow the Mr. Scott rule:

https://www.youtube.com/watch?v=t9SVhg6ZENw

https://www.youtube.com/watch?v=8xRqXYsksFg