A Data Science Central Community
I have twice the experience of building an Analytic team, one with my first job and second with my current one. I left in my first job just that I was frustrated by the organization's lack of commitment to Analytic, or in other word: I found that the Senior Management did not appreciate the development of Analytic. As I learnt it after years of contemplation, the problem lie not with the company, nor the boss: It has more to do with YOU, the lonely Analytic guy out there in the organization. And this is exactly how I think it may work ( and it's working so far).
1. Build up the Quantity
In the financial services industry, there are only two things in the organizations: human and data. For middle level professional or lonely analytic geek, building up the data to support decision making in the organization is key. It builds up a solid foundation for any future analysis and modeling work.
2. Set up the Real Database
From my experience, data accumulates from an excel file, expands to several excel files, then moves on to a MS Access file, then several Access file... To make sure the senior management enjoys the benefit of bigger data. As you may reach to the critical point of 2GB per file ( or per table, the technical limitations of MS Access), it's about time for you to ask for a real database. Make sure you get it.
3. Use the database and do routine things faster
As you have just been awarded the budget to build the database, it's like positive feedback, make sure your report readers feel the improved effectiveness and efficiency of the database. At this stage, operational efficiency is key. Recommend you to document your process and streamline it.
4. Vitalize the operational data
If you start with reporting the post-Mortem (after the fact) business results, your next move should also report and analyze about the day-to-day activities of the firm. Find ways to help your leaders to understand the day-to-day businesses and use them in their business review meetings or even in appraisal. Due to its use, the leaders will drive the data quality of the normal data contributors/the inputers real nuts as the demand for quality is higher with higher usege.
4. One-client view
As most of the mature database suite has the power to integrate data, once you have successfully use the database and gradually moving more and more data to the database platform, you can start integrating the data to form the client view. If your works on reporting is appreciated you will have more and more request to integrate the different information by the end users. Try to join them, build an integrated relationship model. I tends not to use the word OLAP as I believe the data can or cannot be online. It can still be very useful even if it's offline ( or updated regularly).
5. Predictive Analytic for [ for a purpose]
It's often suicidal for you to spell out the word prediction for your boss if you are the only living geek in the planet of your firm. Prediction means possibility of failing to predict. Therefore you need to think about the existing process that can be "enhanced" by predictive model, not creating something new. It could be new clients acquisition, cross-selling or attrition prevention. Anyway, please listen and smell the need of your organization. Very often, going back to your department charter and link the relevant questions that could be fulfill by predictive analytic "Ask not what your company can do with Analytic, ask what Analytic can do for the company."
Wait a minute, to do 5. do not forget about your connectivity within the organization on a day-to-day basis, which should normally be built up with 3. One client view and predictive analytic means there will be output and new insight, finding a proper business knowledge expert are important. With the proper input, your models will fit the expert's understanding about the organization. It's often very dangerous to bring a foreign sharp object called "prediction" to the organization without the support of the subject matter experts: identify them and try to cyrstalize their models in form of predictive models.
Still not getting positive feedback from your boss? Get someone across the organization or from outside to speak about it. To do things properly with the right persuasion skills, you need external forces. Get them and make them push for you. In my example, I get my existing external data provider to speak about targeting and then I ask other countries' predictive modeling expert to share about his great works. Combing these two information, it spells out naturally that predictive analytic is the way forward.
In short, it does not get straight to the point where you want in day one, make sure you have the patience and skills to educate the people slowly and build up their confidence in you to deliver Analytic in all levels.