In these days, I’m starting to use Gemini and I’m discussing about it in dedicated CTP newsgroups. I recently wrote a message that I’d like to share here, but not before some considerations.

First of all, like Chris said, Gemini is less than a threat to (present and future) SSAS solutions. Far less than a threat. We already had this thought when Gemini was publicly presented for the first time nine months ago, but now we have some evidence.

Gemini feature set is not complete in this first release. The data cleansing tools will be not part of this first release. Data models must be very simple and cannot afford the complexity of the real world. And this make me think that, at the end, Gemini will not reduce the requests for data extractions made by power users: almost any operation made by an ETL that translate a data model into one another is not possible using Gemini. So, probably Gemini will increase the number of requests for data feeds by users that are not able to get data by themselves.

That said, Gemini is an interesting tool, that can fill the gap between the absence of a report and the delivery of a report build on top of a data warehouse with data validated and certified. In some way, it is an attempt to take control of existing spreadmarts. And, as I said at the beginning, it is interesting to consider who will be the Gemini users. Of course, my view is relative to the customers I know better that cannot represent the whole set of real world users. However, even if I will need to see how the end users will really approach to Gemini, I started to figure out that we will have several scenarios:

  1. REPORT VIEWER users – they are not able to build a data model in any way, but they can (and sometimes want) to navigate into data by changing some filters in the report. They will use Gemini as end users, but they will not create any new model.
  2. REPORT BUILDER users – they are not able to build a SQL query without a query builder, but they can put every data into Excel and already use simple VBA functions to make the calculation they want. They will build simple models with Gemini and their ability to access new data will be somewhat limited by their skills. They will ask to IT for data they are not able to get or to calculate.
  3. DATA MODELER users – they are able to build a complex data model, today they use Access to build their data models overcoming the limits of Excel (they started to do that years before Excel broke the 65535 rows limit). They are not programmers nor DBA, but they are the sort of “advanced users” that will adopt Gemini as their primary tool. They already provide reports to their colleagues and/or managers. Often, they build so complex models that the work to industrialize them in a cube is very hard, because it requires a tough reverse engineering process.
  4. SSAS DEVELOPERS users – they are, like me, happy to use Gemini as a tool to navigate into data the first time, to create prototypes and to understand if a model is good for an end users before building the complete DW-ETL-SSAS solution. Today these prototypes requires hours to days of work, I expect we will be able to lower this time and, most important, to apply changes to these prototypes in a faster way.

I like the utopist vision that Microsoft have about Gemini and its SharePoint integration – however, I’m not sure we will see this vision applied 100% of the times – especially in the first years, just because SharePoint adoption has a wider impact than the installation of the client side of Gemini (and Excel 2010 – which can affect only a few PCs in a company).

Finally, I read someone asking for a tool that will migrate a Gemini model into a SSAS solution. Well, this is not on top of my
priorities. If you migrate to a managed SSAS solution (managed in the sense you will have a DW-ETL-SSAS daily process), you probably have other requirements (data cleansing, data modeling, …) that requires to build a new Gemini/SSAS model anyway. The only exception would be the Gemini model created on a Data Mart (regular star schema), but I think that if you have a
star-schema data mart, you probably already have the corresponding regular SSAS cube…

A more important piece of software would be a way to create a documentation of the Gemini model, in terms of metadata descriptions, that would help me to create the corresponding Data Mart and SSAS cube models and the necessary ETL steps. As you may think, my interest about Gemini is mostly about the SSAS developers type of user!