Winning in the turns with "Analytics on Demand”
As the world comes out of the worst financial crisis ever, the world has seen two big turns in the space of the last 12 months alone. Lehman fall led to panic leading to the first big turn and the anniversary of the big fall was greeted by record stock market highs, possibly leading us to the next big turn. Organizational nimbleness has never been tested like this before, and, worst still, the global volatility, given the free web and higher degree of interdependence between multiple economies and events, does not show signs of ever reducing.
Understanding ever-growing reams of data with agility– both organizational as well as macro-economic, has assumed significant importance given the environment. While detailed and on time insights into the data around customers, vendors, productivity, profitability etc. enable the organizations to raise their ability to extract the best productivity from the resources deployed, these insights also provide them early signals on the business performance and bolster their capability to react quickly to the volatile business environment.
However, creating a world class “data to insights” capability is not easy. Even the best organizations have issues around lack of data integrity, lack of single version of truth, applications and data in silos and not talking to each other and last but not the least, lack of scalability around the talent pool required to analyze data. For some firms who have already have invested in BI and data warehousing applications, quick response to different situations requires highly agile analytical capabilities with aggressive timelines that cannot wait for the changes to be made to those BI tools and applications. For those, who haven’t got these tools in place, IT investments in BI tools, data warehousing and middleware initiatives are not easy to come by. They need to look at ways and means of achieving analytics capabilities without investing in the IT infrastructure as getting buy-in for these investments is a long drawn process and not a recipe for success during the turns.
Is there a middle path possible and if yes, what is that middle path? Is that middle path “Analytics on Demand”? The answer atleast to my mind is yes for 3 reasons.
Firstly, “Analytics on Demand” or “Analytics on the cloud” enables the organizations to leverage global talent pool on tap without making any significant investments upfront. This reduces the cycle time in terms of getting business case signed off, change management etc.
Secondly, an on demand model enables agility while responding to changing business and analytics requirements. It enables the organizations to experiment with different views of their data. While the organizations are trying to fine-tune their business requirements around reporting and analytics, an on demand model provides them ad-hoc capabilities around reporting and analytics using the people layer on top of the hard coded applications, more apt at providing well defined and configured analytics.
Thirdly, I have always believed that “IT is the most extreme form of process outsourcing”. This also implies that an on demand model can be used as an interim before business applications are developed. This approach provides quick time to market and an ability to perform quasi-user testing at the stage of requirements gathering! I have noticed that huge % of analytics is “disposable”. It cannot be defined and coded in a business application without running the risk of it becoming obsolete by the time the application is ready. While experimenting, once it is established that certain reports and analytics have a large count and can be standardized for business use, those can be hard-coded in the business applications.
As my colleague Santhana wrote on democratizing analytics (http://www.infosysblogs.com/knowledgeservices/2009/10/democratizing_analytics_is_it_1.html ), an on demand model puts the power of leveraging analytics in the hands of people who need it rather than bundling these needs in a central initiative, larger than life and something that has a part chance to see the light of the day.
Comments
Analytics on Demand is definitely turning out to be widely adopted. The verticals Web Analytics (Omniture, Coremetrics etc) and Social Analytics (Visible Technologies, Radian6, Nielsen BuzzMetrics) have a huge list of successful products.
Posted by: Sankara Narayanan | December 3, 2009 2:48 PM
The argument looks fairly logical, am just wondering if it is so logical, is it being practiced currently or not
Posted by: Rachit Dhir | December 8, 2009 12:11 PM
Yes, there are large numbers of analytics projects that have got outsourced. Analytics capabilities span across the entire analytics value chain ranging from data management to data analysis & reporting to predictive modeling. They also span across multiple areas of expertise like marketing, operational, financial and risk analytics and multiple industry segments like Banking, Retail, Insurance, Energy & Utilities, and Telecom. Analytics work happens in 2 models – a) short term projects; b) Annuity deals in terms of having a set of people dedicated for a client and the client providing ad-hoc plus regular work including model maintenance.
Having said this, it is important to bring this practice under the umbrella of “Analytics on Demand” to brand it correctly.
Posted by: Rahul Shah | December 10, 2009 10:40 AM
These types of projects would require data to be transferred from the clients? How is this done and how is the data privacy and confidentiality taken care of?
Posted by: Marshneil Pachori | December 10, 2009 10:42 AM
Yes, data transfer is one of the key requirements. It can happen through secured FTP servers and clients providing access, it can also happen through vendor’s FTP servers, through password enabled hard drives if the data is huge or over emails if the data is less than 10 MB.
In terms of maintaining data privacy and security, I think this is a non-issue given that ITES industry has been accessing clients’ data, be it while managing client applications or running the BPO work for them. Having said this, there are multiple ways to protect data privacy. Dedicated secured environment can be created for information exchange, outputs are delivered over mails that are exchanged in an encrypted format, restricted access is maintained to the Offshore Delivery Centers, printers & shredders are installed in the access controlled delivery huddle rooms and last but not the least, printout history/ logs are made available to the client
Posted by: Rahul Shah | December 11, 2009 5:56 AM
How does this work from the operations perspective in terms of the time a project takes, interaction with client stakeholders, ability for the vendor to provide domain expertise etc.
Posted by: Marshneil Pachori | December 11, 2009 5:57 AM
The time taken to complete a typical analytics modeling project is anywhere between 8 and 12 weeks. Data Collection & Preparation takes about 4-5 weeks, Profiling & Segmentation takes about 2-3 weeks and the modeling itself takes 3-4 weeks. Having said this, it is all project specific. There are multiple reviews with the client stakeholders to ensure that there are multiple reviews. During these interactions, clients’ domain expertise is leveraged to create a quality output.
Posted by: Rahul Shah | December 11, 2009 5:58 AM