BI: the problem of choosing a solution and an example of one failure
In this article I want to talk about how the reporting system in one rather big company has changed, what products have been compared in real conditions (the harsh realities of interaction with our managers) and why the implementation of a fairly convenient and beautiful BI system failed.
Separated from a larger company, the mighty and terrible Cognos BI 8.3 was inherited. The reporting portal already had a certain set of reports used by the old team, but the team was changing, the commercial service was growing, which also acquired its own analytics service, which gained access to create new reports in the system. At a certain moment, the number of reports, objects in the data mart (tables, views, stored procedures), which was used by the analytics service, began to amount to hundreds, with about 70 people in the company's central office. Plus (although this is a minus), the analytics service prepared a large number of reports for top managers using kilometer macros in Excel, loading reports from the same Cognos and data from the storefront into it.
With the introduction of SSAS OLAP cubes, the number of reports used in Cognos began to decline (from two hundred to one), but the macros have not gone away. The reason is simple. The management refused to look at the cubes themselves. People are used to receiving reports from analysts by mail, and especially zealous immediately to a color printer in their office. The analytics service began to use them as an alternative to semi-finished reports as a data provider. In addition, having felt their advantage in the speed of data output, and armed with Excel 2010, which can “swallow” a million records, they began to manage to send a OLAP server requests for unloading this million records with 10-15 attributes into a spin.
A person came to the company who worked with a major integrator and said that he knows how to fix everything, or rather sold this idea to high authorities. And so we bought QlikView. The product is really interesting, but along with it the philosophy of its use, which the integrators gave us as “DAR” - Dashboards Analysis Report, is also for sale. In short, the first two levels (Dashboards and Analysis) are control panels and a set of graphs or small tables for top and middle management that see beautiful and informative control panels using convenient (really very well done) filter panels to monitor business conditions on the plots entrusted to them. The third level is a detailed display of data (tables, pivot tables) for analysts. Surprisingly, the commercial director, the main consumer of the printed reports, actively participated in the development of the application. He told what he wants to see panels and graphics, at each stage of development he looked at the result and expressed his wishes for improvement. In the final, saying that everything is very cool, he called the head of the analytics service and ordered to give him data to the printer from this miracle, since he himself cannot use the filters. Hand face.
Analysts didn’t need any informative graphs, and it’s more convenient to extract detailed data from OLAP using familiar Excel.
')
$ 20,000 (licenses) and 4 months of work to nowhere.
So what is to choose from this small zoo systems for the further development of reporting? To answer this question, you must first get answers to a number of other questions in the company:
If the company has more or less settled business wishes for what data they want to see and, for example, the commercial service KPI system does not change once a quarter, as in our case, then all three can be considered (Reporting Services also comes with SSAS ) and much more on the market. If a business does not have a clear understanding of what information they need for daily (weekly) viewing, then they are not particularly interested in dashboards. Throw out from consideration everything that relies on user convenience, consumers will be those who prepare business presentations, etc., and flexibility is important to them (OLAP is their help). Rely on the choice of analytics service or who is preparing presentations there.
If directors and other tops are not just ipad fans, but know how to use them, cannot live without mobile Internet and can manage data filters themselves (advanced users), then QlikView and others like them will like it very much and they will be ready to take part in the implementation (well, I think so). If, at best, the business is ready to look only at what has arrived at the post office, then only that which can send this mail (QlikView did not know how, we can wait for the new version at the beginning of 2014, it may be there), and not just send it, also do it on an event (for example, by the middle of the month the plan was fulfilled only by 30%). SSRS, Cognos and the like can send mail, and the second one can send everything to the printer at once (this may be important, although for me it is a schiz).
If the data volumes are quite large, then Cognos and everything that relies on its ROLAP models can let you down, and you have to buy in addition either IBM's TM1 or some OLAP server, which leads us to the next question.
Consider not only the cost of licenses, but also the availability of product specialists on the labor market, the availability of courses and training materials. We have decided to abandon Cognos because of the fact that at some point there was no one to even invite an interview.
Lastly, one more piece of advice: do not shoot the gun on the sparrows.
If you need only a single corporate reporting portal, do not buy something big and expensive for this. Bought Cognos - use everything (Event Studio, Analysis Studio, Planning and many other things there), if only a portal, then many products are cheaper, the same SSRS goes to the load to SQL Server. Looking into the future is, of course, good, but a cheap and angry product will be suitable for starting, and a business will be trying out BI only in a year or two, and there new products will mature and many other things may change.
Initial state of affairs
Separated from a larger company, the mighty and terrible Cognos BI 8.3 was inherited. The reporting portal already had a certain set of reports used by the old team, but the team was changing, the commercial service was growing, which also acquired its own analytics service, which gained access to create new reports in the system. At a certain moment, the number of reports, objects in the data mart (tables, views, stored procedures), which was used by the analytics service, began to amount to hundreds, with about 70 people in the company's central office. Plus (although this is a minus), the analytics service prepared a large number of reports for top managers using kilometer macros in Excel, loading reports from the same Cognos and data from the storefront into it.
Add to this OLAP
With the introduction of SSAS OLAP cubes, the number of reports used in Cognos began to decline (from two hundred to one), but the macros have not gone away. The reason is simple. The management refused to look at the cubes themselves. People are used to receiving reports from analysts by mail, and especially zealous immediately to a color printer in their office. The analytics service began to use them as an alternative to semi-finished reports as a data provider. In addition, having felt their advantage in the speed of data output, and armed with Excel 2010, which can “swallow” a million records, they began to manage to send a OLAP server requests for unloading this million records with 10-15 attributes into a spin.
And a little about the beautiful BI-system
A person came to the company who worked with a major integrator and said that he knows how to fix everything, or rather sold this idea to high authorities. And so we bought QlikView. The product is really interesting, but along with it the philosophy of its use, which the integrators gave us as “DAR” - Dashboards Analysis Report, is also for sale. In short, the first two levels (Dashboards and Analysis) are control panels and a set of graphs or small tables for top and middle management that see beautiful and informative control panels using convenient (really very well done) filter panels to monitor business conditions on the plots entrusted to them. The third level is a detailed display of data (tables, pivot tables) for analysts. Surprisingly, the commercial director, the main consumer of the printed reports, actively participated in the development of the application. He told what he wants to see panels and graphics, at each stage of development he looked at the result and expressed his wishes for improvement. In the final, saying that everything is very cool, he called the head of the analytics service and ordered to give him data to the printer from this miracle, since he himself cannot use the filters. Hand face.
Analysts didn’t need any informative graphs, and it’s more convenient to extract detailed data from OLAP using familiar Excel.
')
$ 20,000 (licenses) and 4 months of work to nowhere.
So what is to choose from this small zoo systems for the further development of reporting? To answer this question, you must first get answers to a number of other questions in the company:
1. Who is the system intended for (who is the main consumer)?
If the company has more or less settled business wishes for what data they want to see and, for example, the commercial service KPI system does not change once a quarter, as in our case, then all three can be considered (Reporting Services also comes with SSAS ) and much more on the market. If a business does not have a clear understanding of what information they need for daily (weekly) viewing, then they are not particularly interested in dashboards. Throw out from consideration everything that relies on user convenience, consumers will be those who prepare business presentations, etc., and flexibility is important to them (OLAP is their help). Rely on the choice of analytics service or who is preparing presentations there.
2. What are the company's preferences in how to deliver data to the consumer?
If directors and other tops are not just ipad fans, but know how to use them, cannot live without mobile Internet and can manage data filters themselves (advanced users), then QlikView and others like them will like it very much and they will be ready to take part in the implementation (well, I think so). If, at best, the business is ready to look only at what has arrived at the post office, then only that which can send this mail (QlikView did not know how, we can wait for the new version at the beginning of 2014, it may be there), and not just send it, also do it on an event (for example, by the middle of the month the plan was fulfilled only by 30%). SSRS, Cognos and the like can send mail, and the second one can send everything to the printer at once (this may be important, although for me it is a schiz).
3. What volumes of data will you have to deal with?
If the data volumes are quite large, then Cognos and everything that relies on its ROLAP models can let you down, and you have to buy in addition either IBM's TM1 or some OLAP server, which leads us to the next question.
4. How much the company is ready to pay for all this and spend on further development and maintenance?
Consider not only the cost of licenses, but also the availability of product specialists on the labor market, the availability of courses and training materials. We have decided to abandon Cognos because of the fact that at some point there was no one to even invite an interview.
Lastly, one more piece of advice: do not shoot the gun on the sparrows.
If you need only a single corporate reporting portal, do not buy something big and expensive for this. Bought Cognos - use everything (Event Studio, Analysis Studio, Planning and many other things there), if only a portal, then many products are cheaper, the same SSRS goes to the load to SQL Server. Looking into the future is, of course, good, but a cheap and angry product will be suitable for starting, and a business will be trying out BI only in a year or two, and there new products will mature and many other things may change.
Source: https://habr.com/ru/post/207788/
All Articles