Bottom Up And Top Down Approaches To Estimating Costs For A Single BI Report

How much does it cost to produce a single BI report? Just like typical answers to most other typical questions, the only real answer is “it depends”. But let’s build a few scenarios:

Scenario 1: Services only. Bottom up, ABC approach.

Assumptions.

 

  • Medium complexity report. Two data sources. 4 way join. 3 facts by 5 dimensions. Prompting, filtering, sorting ranking on most of the columns. Some conditional formatting. No data model changes.
  • Specifications and design – 2 person days. Development and testing - 1 person day. UAT – 1 person day.
  • Loaded salary for an FTE $120,000/yr or about ~$460/day.
  • Outside contractor $800/day.

Cost of 1 BI report: $1,840 if done by 2 FTEs or $2,520if done by 1 FTE (end user) and 1 outside contractor (developer). Sounds inexpensive? Wait.

 

Scenario 2. Top down. BI software and services:

Assumptions:

  • Average BI software deal per department (as per the latest BI Wave numbers) - $150,000
  • 50% of the software cost is attributable to canned reports, the rest is allocated to ad-hoc queries, and other forms of ad-hoc analysis and exploration.
  • Average cost of effort and services - $5 per every $1 spent on software (anecdotal evidence)
  • Average number of reports per small department - 100 (anecdotal evidence)

Cost of 1 BI report is $8,250. Still sounds inexpensive? Let's keep going.

Scenario 3. Top down. Information management software and services.

Assumptions:

  • Average BI software deal per department (as per the latest BI Wave) - $150,000
  • Average ETL software deal per department (as per the latest ETL Wave) - $150,000
  • Cost of 1Tb DW (as per latest DW Wave) - $50,000. Let's also figure in a 3x for production + development test + DR environments.
  • 50% of the software (and DW) cost is attributable to canned reports, the rest is allocated to ad-hoc queries, and other forms of ad-hoc analysis and exploration.
  • Average cost of effort and services - $5 (anecdotal evidence)
  • Since the effort of building full blown DW/ETL environment usually go beyond one department, let's increase the number of reports by 3x to 300

Cost of 1 BI report also comes out to $8,250 since we are now using economies of scale for ETL and DW accross several departments. But that's not the end of it.

Scenario 4. Top down. Fully loaded:

Assumptions:

  • 10% of enterprise operational data quality cost of $1M and 2x for services
  • 10% of enterprise operational MDM cost of $5M and 2x for services
  • 10% of an enterprise portal fully loaded (hardware/software/labor) cost of about $3M (as per Matt Brown)
  • Not counting (since I don't have good numbers - if you do, please provide) a) 50% of enterprise performance management applications b) servers (physical or virtual) for production, development, test, UAT and DR regions c) additional data center support and infrastructure

Cost of 1 BI report is now $15,250. Still waiting to collect more data on the 4th bullet point above, but I can easily see how the truly fully loaded cost might grow to $20,000.

Of course this is only one side of a complete picture, cost/benefit equation, since the cost of producing a BI report cannot be looked without referencing the value it provides. After all, if that one report produces $100,000K in cost savings or in additional revenues, then even $20,000 becomes very attractive.

Suggestions for bringing the cost of BI down?

  • Since services and support are major parts of the cost equation - enable end user self service
  • 100 reports is a very low estimate, most organizations have more. Practice agile BI to reduce the effort it takes to develop each report, and ensure that the requested reports are actually used.
  • Leverage other best practices for reducing your BI costs found here.
  • In certain specific use cases, consider BI Saas.

I welcome and encourage comments and will update this blog with all relevant and useful suggestions.

Comments

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris,

This post is thought provoking and I am sure that you will get a lot of discussion going.....I personally think that you need to create metrics which will help you normalize cost/value/usage across the reports and help you create fair comparison across all scenarios you mentioned.... here are my thoughts

Value: Looking at reports without value they create might result in comparing apples and oranges. Obviously your scenario’s 3 and 4 are using lot of resources because they need that level of sophistication to product those reports. I think looking at just cost without factoring in value created by reports does not give you a fair picture. Possibly a composite metric which could capture “Reporting $ required for per $ of value created” would be a better measure to compare investments in reports across these scenarios. That way a light weight report (scenario 1) can be fairly compared with heavy weight report scenario 3 or 4).

Audience/# of users: We need to factor in the # of users who are going to consume this report. Normalize the cost of the reports across # of users. So a metric like Reporting $ per user for a report would help normalize comparisons across reports.

Also, in scenarios 2 through 4, it seems to me that your are assuming a complete analysis area and not just “single report”.(basing this assumption on fact that you are factoring in the other investments done in the enterprise data warehouse). In this case…. There are many reports/report outcomes which will come out of those activities not just single report (For example building sales by territory report which allows you drill downs across sales territory, verticals etc…. will result in many reports).

let me know what your thoughts are

Vish Agashe
http://vishagashe.wordpress.com

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Vish, very constructive suggestions, thanks. I redid the numbers a bit and that lowerd the fully loaded cost to $13K-$20K. Also, yes, it definitely makes sense to mention the value of these reports, but it's not the purpose of the blog to try to build a BI ROI business case. That's a tough proposition that I reviewed here http://www.forrester.com/rb/Research/business_case_for_bi_now_more_critical/q/id/54955/t/2

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris:

Excellent analytical approach. But some things are still missing from the analysis.

First, you only include upfront costs of developing reports, but you don't include life-cycle costs of report maintenance, which probably are proportional to report complexity, number of fields, number of joins, and number of sources. Perhaps you should include a life-cycle cost multiplier in your equation, factoring in report complexity. That would increase report costs, of course, while delaying the break-even date for each report.

Second, you amortize "average BI software deal" cost only by the number of reports in a given department, but fail to include the fact that this cost is also spread across other BI apps, such as ad-hoc queries, dashboards, intelligent search, and interactive visualization. If you include those other apps in your analysis (I like to refer to this as "economy of scope," that will reduce the cost per BI report.

Third, you can't realistically estimate the per-report value by speculating on the possibility that a single report costing, let's say, $13,000, leads to a business opportunity worth, let's say, $100,000. I refer to that as the "lottery value" of BI--in other words, the possibilty that some scrap of actionable info led to a breakthrough insight that unlocks unlimited potential value. By that same token, you could justify anything as as the potential "ticket to paradise." (Keep in mind that the only person who ever cashed in a ticket to paradise had 2 of them, and that he already had money in his name).

Jim

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris:
Interesting analysis. I'm curious why open source BI & ETL would not be a consideration for lowering costs? Typically a support and update subscription to OSS is significantly less than traditional solutions, while self support versions are completely free. Also, might be helpful to represent your data as a table to read more easily.

- Mike
http://www.jaspersoft.com

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris,
Great work on showing some different ways of determining the cost of a report. It proves that developing a cost model is an art, not a science. But there is some science to it.

I know you are not trying to use this as an ROI model, but that seems like a good test for the numbers. So, my thought is to apply it to one of our customer scenarios.

A couple of years ago, we helped one of our customers set up a Business Intelligence Competency Center. As a result of new governance models and better cross-polination between business and BI, they have reduced the number of reports being generated by analysts from 6000 down to 180. The best part of this story is that with fewer reports, they are exceeding the expectations of the business compared to pre-BICC.

So, here's the math...

If you take the low end of the bottoms up approach, they would be saving $10.7M. If you take the top down approach, they would be saving $48M. Either way, it's a very significant savings.

I would love to send your data to our customer and get their feedback. Even if the end result is not accepted by them, the approach is certainly one they should consider. Thanks for your work on this one.

-John

re: Bottom Up And Top Down Approaches To Estimating Costs For A

John,
It took me a while to have any faith in the numbers, so I went ahead and did my own rough calculations based on a large report development effort I participated in over the past year. The per report numbers (using a highly normalized database, rudimentary reporting tools, and waterfall methodology) come out to be about $2,000 of effort per report. Thanks for taking the time to work through some of the other math.

The analysis begs the question, though - how much can an organization expect to save by following the advice you provide?

Self Service yields a 20% savings?
Agile BI practices is a 15% cut in development time?

Any metrics on that part of the conversation or what highest cost components are in the development costs. Technical skills. Subject matter expertise.

Thanks for the thought provoking numbers!

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris, this is great. I find it remarkable that more has not covered on this topic in the industry – I think you’re on to something.

Based even on your conservative estimates, one can see that the cost of running a traditional BI deployment can get significant – quickly exceeding the cost of the software and hardware itself. With high per-report costs like this, it’s no wonder IT organizations force users to really think through their questions and requests before embarking on report writing efforts.

But if users are forced to map out their precise questions in advance, what room is left for users reacting to interesting things spotted in reports, exploring follow-on questions, or asking an unanticipated question? While the costs alone are an argument for more “self service” BI and analytics, the real benefit for more user-driven models is that users themselves can ask and answer their own questions as they come up without going through IT for every new request. This leads to new insights and better decision making.

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Paul,
It was the business to IT interaction, taking place in the BICC, that lead to the savings. The IT side learned to ask questions about what the business actually needed and the actual use of the data. And the business finally understood more about what the BI team could produce. When these two worlds came together, many of the reports were eliminated because they were no longer necessary. It's the process of information governance in the BICC that leads to the savings. Tom Gransee wrote a great white paper on "Building a Business Intelligence Competency Center" that might interst you. It's one of the best I've read: http://ow.ly/10saR Tom used to teach at TDWI.

re: Bottom Up And Top Down Approaches To Estimating Costs For A

L.s,

Good article. In the costs however i miss the DB License and maintenance costs.... which can make a huge difference.

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Thanks for the thought-provoking article Boris. Of course there are often more factors that can be brought into a method for estimating costs for BI report development. Some big ones are: pre-packaged enterprise data models (denormalized views/queries and dimensional models), pre-packaged ETL scripts for popular ERP/CRM applications, pre-packaged reports/dashboards as well as offshore report development. I like John's comment about the BI Competency Center as well...building fewer of the highest value reports would definitely help.

re: Bottom Up And Top Down Approaches To Estimating Costs For A

Boris, just posting a reference to Roman's article about your calculations. Would be curious to see your reply as well.