I am working at a custom software development company. The number of incoming custom projects has increased over the last few years.
The first problem is that not all customers are willing to pay us for the development of the requirements specification. Thus, we are losing money due to the estimation of some projects, which we reject. The second problem is that projects are complex and do not have a lot of standard tasks.
Are there any tools helping to estimate a project with sketchy requirements? It would be perfect to have automatic estimation, but speeding up the process for a human would be enough.
When I can affect the type of contract - T&M is simplest solution, of course. But this is rare case because most of the projects are tender-based.
Usually requirements occupy a hundred pages, but are still sketchy and unclear. So I'm looking for ways to accelerate at least a primary estimation of the requirements.
I will be glad to hear any tips and ideas on this subject.
P.S. I don’t like the idea of rejecting every other project by default :) Saving Changes...
Are there at least some functional similarities among the customer requirements? Could you develop a Function Point estimation scheme? This won't be automatic estimation, but hopefully standardizes and speeds up your process.
Product Operations Program ManagerBarcelona, Cataluña, Spain
It is tough to provide an answer that falls into the "one size fits all" category. Time & Materials seem indeed the way to go, but then it clashes with the tender approach.
Whatever you do, don't imitate Sacyr and the Panama Canal scandal. Sacyr won the bid (lowest bid); after they got warded the contract, they realized that the budget was largely underestimated Saving Changes...
Thomas WalentaGlobal Project Economy ExpertHackenheim, Germany
The good message is that the number of requests is increasing.
There are many things you could do, I just ask some questions:
- have you developed a standard risk assessment that you could apply to all incoming tenders - then you can compare and be more selective, you also have a better understanding of your risk
- who says you could not bid T&M for a tender? Even T&M disguised as FP is possible if you put significant disclaimers on the main risks
- have you considered to go agile for some, meaning forget about the crap requirements and offer an interative and incremental project (that is what a company I know recently did successfully)
- did you try to establish trust on exec level first (even before there is a specific tender) and then think about jointly with the customer how to approach their problems (and avoid their tender efforts) Saving Changes...
Samuel, we thinking about more lightweight estimation scheme, because Function Point perfect for similar/standard projects. Ironically, for such projects we usually agree with a customer on the development of the requirements specification.
Eduard, thanks for the warning. Usually we compensate difficult projects with help of successful projects. Sometimes we can cut the requirements.
Thomas, unfortunately we can do all right things only with non-tender projects (and be happy with them). We periodically trying to do steps 1 and 4 (but rarely successful).
P.S. We do not aim to evaluate the project as accurately as possible. We want to be able to understand - whether it is necessary to assess the project accurately. Saving Changes...
Anonymous
I worked with one consultancy that did a few assessments in view of hazard.
They utilized an equation that would assess the accompanying:
Kind of client (e.g. how community oriented they were, had we worked with them some time recently, and so on.)
The appraisals were not exceptionally precise, but rather they were snappy and shoddy. The cost charged to the client was expanded to represent the additional instability of utilizing this approach.
A case:
A client reaches asking for some work to be finished. The necessities are not very much characterized, so the specialized group looks them over and gauges that it is a "medium" estimated extend. They commonly consider "little" under one month, "medium" as 1-2 months and "vast" as more than 2 months.
The venture includes no new advances and is a rehash of comparative work the organization has done some time recently. They rate it as "low" for specialized unpredictability.
In the wake of chatting with the client they observe them to be entirely protective and extremely legally binding (i.e. they demand a due date and full degree conveyed). They rate the client as 'trying'.
They apply this data to a spreadsheet utilizing a basic weighted equation. This outcomes in a venture gauge of 40-60k. As the client seems to be "testing" they choose to run with the upper end of the range and quote 60k.
etract all similarties and those can be parametric estimation, and
Model the system. A model might be a mental model, diagrams, or existing data records. Decompose this model and build estimates from the components. Assign values and error ranges (+/-) to each value.
off course it depends on your systems to analyse. Saving Changes...
Community Champion
