Part 5: Estimate the Cost Categories
Now we can use the cost breakdown structure identified in part 4 of this series to come up with a working estimate of costs for managing an architecture model. As with the benefits, it’s not going to be realistic to create a simplistic formula that we can plug numbers into – we are creating a framework.
A brief word on estimating staff costs. Depending on the jurisdiction we can assume a 40-hour week and say 47-50 weeks in the year. This works out to 1880-2000 hours in a year. Dividing an annual salary by this amount gives us an hourly rate for salary; we then multiply this figure by a multiple to impute overheads. I’ve heard 1.5 used, also 1.75. So an architect making $120k per year and working 2000 hours would have a base hourly cost of $60 per hour, multiplying by 1.75 gives $102 an hour or $816 a day. The precise figures to use are less important than that they should be ones that you can defend.
Data Gathering
Data Import.
This will be driven by what data format is available, what data exists, and so on. I might assume half a day to analyse an input file, the same time again to discuss and specify requirements, and transformation would depend on the format – writing simple Excel macros might only take half a day, creating an XSLT or a SQL extract would take longer, perhaps a few days. We can multiply this by the daily rate for the relevant person to estimate a cost for each interface. For example, for 2 interfaces, each taking 4 days at $800 a day works out to $6400 in cost.
Import and conversion of existing diagrams. Again, this will depend on circumstances, but if we are given Visio diagrams to work with, I would estimate 2 hours per diagram as Visio 2013 has a shape replace feature. Other tools may take more or less time. Let’s assume that we import 100 diagrams at 2 hours per diagram; at an hourly rate of $100, it costs us $100*2*100 = $20,000.
Data Discovery.
Estimating this cost will vary widely based on what needs to be discovered and documented. We estimate the costs involved by the time of the modelers and the time of the people that they need to engage with. For example, say two process analysts will work at $800 a day for 3 months, and 20 staff at an average of $600 a day will need to give 5% of their time to these analysts. Then the costs of the data discovery will be 2*60*$800 + .05*20*60*$600 = $132,000.
Data Rationalization
Data cleaning.
As for data discovery, this will depend on the number of systems, how ‘clean’ the data is. If we assume that one data analyst at $800 a day will need 3 weeks to clean legacy data, we would spend 15*$800 = $12,000 on data cleaning.
Naming conventions.
We’ll assume a core working team agrees a naming standard; this team consisting of 2 leads, meeting twice each with 3 groups of domain owners (each sending 2 representatives) for an hour each time, with the leads performing an hour follow up after each meeting to address action items and so on. So we have 10*10*$100 + 2*2*8*$100 = $13,200 to establish processes.
Resolving overlaps.
As for data discovery on several factors, primarily it will depend on the level of overlap. If we assume that two SMEs at $800 a day will need a week to agree on and implement a mechanism to resolve conflicts, we would spend 2*5*$800 = $8,000 on data conflict resolution.
Communication and roll-out
Obtaining buy in.
In general I would budget 8 hours of two people’s time to create materials for a road-show, plus an hour for each stakeholder group, with a further hour for follow up each time. Assuming that we have 10 stakeholder groups with 2 representatives each, this works out to 8*2*$100 + 10*2*2*$100 = $5,600.
Establishing Processes.
We’ll assume a core working team agrees processes; this team consisting of 10 stakeholders, meeting 10 times for an hour each time, with 2 people performing 2 hours follow up after each meeting to address action items and so on. So we have 10*10*$100 + 2*2*8*$100 = $13,200 to establish processes.
Training.
Based on past experience, training in a modeling standard will generally take a couple of days to full proficiency. Training on processes and working practices will take half a day.
So, for 20 architects, 2.5 days at $800 a day works out to $800*2.5*20 = $40,000 initial training costs.
Tooling
Tools.
This will, naturally, depend on the tool, the number of seats, the precise options chosen and so on. Most tools that I am aware of will cost five or six figures, quoted in dollars, once you’ve paid installation, year 1 maintenance, etc.
Tool consultancy.
Again, this will depend on the vendor and the scope. Most vendors that I know rates for charge their consulting out at $1500-$3000 a day, and an engagement will last between a couple of weeks and several months. So we could be looking at a five figure sum, say $30,000.
Governance overhead
Architecture review.
In general, this actually imposes less overhead than might be expected, depending on whether the organization in question already has an architecture review mechanism in place. We’ll assume that each project requires a 1 hour review meeting of 5 people and 1 hour preparation for the architect. If an organization has 100 projects completed annually, this makes 1200 hours*$100 hourly cost = $60,000 ongoing.
Updating the model.
Here we need to estimate the extra hours this adds to workload, to merge changes into the production model. We’ll estimate 2 hours for each of 100 projects, which comes to $100*2*100 = $20,000 annually.
We’ve now taken some benefit categories and cost categories, looked at components for each of them, and used this breakdown to consider how to estimate benefits and costs.
In the last two posts of the series, I’ll apply this structure to a couple of examples that are based on real implementations that I’m aware of, applying the formulas above and from part 3 of the series.