The Hindu Business Line : Cracking the process code

A scientific bent of mind, clarity of thought and a charismatic personality make Ananth Krishnan an interviewer's delight. In an hour with eWorld, he engaged us completely in how Tata Consultancy Services is involved in industrialising software process delivery and building non-linearity (ensuring that the employee base does not grow in the same proportion as revenue growth) into the offshoring model.

To make an otherwise dry and technical sounding subject come alive is probably not a surprise at all. As a member of TCS Corporate Think Tank since 1999 and an invitee to the quarterly management review with the TCS Board, strategic thinking must be par for the course for this M.Tech from IIT, Delhi. Currently serving as Vice-President and Chief Technology Officer of Tata Consultancy Services, Ananth Krishnan shares his insights in the first of a two-part interview:

How has Tata Consultancy Services gone about industrialising the IT delivery function? The context is how is TCS building non-linearity into its offshoring business model...

What we are doing in the TCS context, in the industrialisation of IT... ... . The first thing we have looked at is repetitive tasks in information technology, to see whether they can be automated to a certain extent. The first candidate for this was programming language transformations, which started way back in the 80's itself. This transformation was pretty successful. The unique thing that we've done in the last 3-4 years is that, we've created a tools foundry. In the past when you had to convert programming language A to programming language B, you would write a tool to do the conversion, something like translating English into Tamil, then you would write an English parser, then you would write a Tamil parser, you would resolve some of the semantic differences and then you would say that I have succeeded in translating language A into language B. But if you had to do English to a Chinese parser then you would pretty much start from scratch.

So what the tools foundry does is it says that it treats language conversion as a generic problem. Then you feed in the semantic and syntax of each language of source and target language to the tool foundry. The tool foundry outputs a tool and then the tool works. So you don't have to write a tool for every pair of translations that you do.

So that's one kind of innovation that's happened. This has happened in our Pune research centre. That was what enabled us to treat one aspect of IT in a repetitive manner.

The second set of things that were done was in relation to what is called model level development. Now model level development has been the holy grail of IT for many years. You specify in some high level manner as to what exactly should this IT system do and through some magic the IT system is actually built with the end need specified in some formal way. So there have been a lot of models developed in the last 30-40 years, but the state of the art right now, is something called Unified modeling language (UML). The UML has been around since the early 90's and three of the big UML proponents - Grady Booch, James Rumbaugh and Ivar Jacobson have brought together the modelling philosophies that created UML. And UML was successfully commercialised by a company called Rational, which is now part of IBM.

What TCS has done around UML is to look at the traditional weakness of IT. It is almost like saying, in house construction the architect's role is the most critical and any workman can put one brick over the other as per the architect's decision. But we all know that it is not that easy even when you construct a house. The architect's plan might be great, he might have an exact understanding of what the house owner wants or the end customer wants, but if the workman has done a lousy job, the house is not going to be good. What we have done in TCS is to build a very strong link between UML specifications and actual construction of code. So this is something that is a high productivity enhancer. This is built out in our tool set called MasterCraft. Now MasterCraft was designed in the mid 90's to build our own products, whether it is our banking products or financial services products and so on. From 1998 onwards we've been using it in our service engagements. Today MasterCraft is used in about 70 or 80 projects of TCS worldwide in the market space and it gives us about 5 to 10 times construction efficiency.

Can you give us a live example where MasterCraft was used by TCS?

For an insurance customer, which was selling insurance in all 50 states in the US, we were rebuilding their automotive insurance applications. They wanted to specify the new system in UML and take advantage of all the capabilities of UML to articulate what they wanted. Once that was done, this new application had to be built for the next 10 years for this company to get its entire automobile insurance application up and running. So we brought in MasterCraft to capture the requirements in UML. Once it was captured, to say that this is what has to be built, traditionally it would have taken five years to build that new system, we did the whole thing in two years, in terms of time, it was a huge acceleration. In terms of cost to construct such a system, it was one-fifth the total. In terms of actual man years, it would have taken something in the region of 500 man years to build; we built it in about 120 to 130.

Now in a traditional job model, we would have assigned 500 people. The client will have to pay me X rupees per hour. We would have actually made more money by not using the tool because it then becomes a labour deal. In this particular case, we went and told the customer, right up front that we are going to use this framework, it is going to give you tremendous advantage in terms of cost, we can also guarantee a certain quality of the delivered code because in many ways it is untouched by human hands, where the possibility of errors is reduced. Here is what it can cost you if we don't use the tool and here is what we can give you as the core proposition. Now there is obviously that much savings in order that we can recoup on investments in the tool, we want to share it and it is fine with the customer.

We actually, with less people deployed on the project, made more money than we would have made otherwise. This customer is now with TCS forever, they absolutely love the framework. So this is the kind of non-linearity that is possible in large development.

Is there any latest process industrialisation work that you are involved in?

Another example of non-linearity is something that we are working with some of the industry leaders. Jacobson is one of the people we are associated with. This is to look at ways in which we can further industrialise the process of software development. We have seen the way in which people actually write applications. In IT software, you've learnt it by reading a book, you must have gone to college or done a training course and so you have some knowledge yourself. What this knowledge gives you is a certain set of skills and certain knowledge of what you don't know, which is a larger part of what you actually know.

Now when you hit something that you don't know, what do you do? You ask your friend, you ask your boss or go on the Internet or post a question and so on. You ask how do I do this particular task, then you get advice, this is what all of us do. Jacobson has a very interesting company built on this concept of software that mentors software writers and we have tied up with them a couple of years ago. It's a very good experiment and it is called active guidance. Active guidance as a concept has been built by this company called Jaczone, which is Jacobson's company. They have built it for one particular platform - IBM Rational tools, so if you have Rational tools then you use Jag zone's WayPointer. It prompts you, it gives you guidance and if you're stuck, here are four things you should try. It becomes like an expert mentor sitting inside the software, giving you prompts on what to do next. What we have done is we've extended this to two or three other platforms, so now instead of just giving guidance on Rationale tools; it gives guidance on some of the other tools of interest. We are now extending it for general purpose documentation, doing it for testing, even for enterprise software such as SAP, Oracle and so on. These are in the same state that IT was twenty years ago, you need an expert; you need a very expensive set of consultants to come and tell you how to do packaged software deployment. Say, in those days if you wanted to process payroll or something, the payroll expert or guru will come and do some magic and the payroll will work. Today's big software packages are like that.

It is now ripe for disruption in the sense that, how do I now say that with something like active guidance I can take an X or Y and call him a SAP expert. I can do that if he has an effective tool in his hand. All that he's got to do is read, follow the problems, what the tool is telling you. You know something about SAP, Oracle, Siebel and the tool has got the knowledge, it will tell you now do this or this and you just follow that. We are still far away from that goal, but we have started doing some research in that area to bring about this kind of productivity increase and non-linearity with this kind of technology and guidance. So that's one of the future steps that are being targeted.

The last example I would like to give for non-linearity is that we have looked at our own record of execution. Lot of people come back and ask us, people who have gone through the CMMI (Capability Maturity Model) process maturity, execution maturity and so on and ask: What does it actually give you? So over the last 10 years or so, that TCS has been in the CMMI kind of space, there is enormous amount of metrics that has been gathered on our own projects, literally thousands of projects. Metrics are available because that's one of the key CMMI process. We've gone back and looked at these statistics: quality control, process control, execution, quality, delivery or customer satisfaction metrics and drawn lots of useful correlations on which kind of work, which kind of technology, which kind of execution process results in what outcomes. And like all statistical processes, there is a variation in each of these. Even though the CMM file reduces the variation that you are on time, let's say 97% of the time but there would be 97 per cent plus or minus 3 per cent so what is the difference between the one sigma on this side and one sigma on the other side. Now we can apply control techniques and say that if we have the best practices from this side, the most successful projects are applied to the less successful projects, then I can bring down the less successful projects and then the average will keep moving up and getting better. So this kind of continuous improvement or mindset on a process side is also something leading us to go to forward. It is not as glamorous as putting out a tool or as glamorous as saying here is a fantastic invention which is resulting in dramatic changes, but the core delivery acceptance message is what is shifting the bulk quality towards the upper end, moving the averages, improving productivity techniques, all of which would result in dramatic shifts in a non repetitive fashion. If there is a project that has not applied active guidance or which is not using a tool or MasterCraft. How do you shift 10,000 projects is a much more engineering focused kind, almost Toyota like invention.

(To be concluded)

eworld@thehindu.co.in

More Stories on : Interview | Software | Tata Consultancy Services Ltd

Article E-Mail :: Comment :: Syndication :: Printer Friendly Page

Stories in this Section
Map mania

Powering training

Recording TV programme

In tune with change

Of deal sizes and offshore billing rates

Cracking the process code

Quiz

`Just enough to get by quickly, cleanly'

Cartoon

New range

For a better picture