Build Operate Transfer: Tensilica's Experience in Offshoring Engineering to India

Tensilica is a private, venture-capital financed semiconductor intellectual property company headquartered in Santa Clara, CA. It was founded in 1997 and currently has about 100 employees worldwide. We license configurable and extensible embedded processor technology to the leading system and integrated circuit suppliers of the world. Our technology includes both hardware and software tools.

In August 2003 Tensilica decided to explore low-cost, offshore engineering resources to supplement the engineering team in Santa Clara. We started an offshore development center in Pune, India in November 2004. We chose a build-operate-transfer (BOT) model, leveraging the infrastructure and expertise of eInfochips, located in Ahmedabad, India. This company was established in 1994 and focuses on three main technology areas: ASIC design, embedded systems, and application software. It has 450 engineers, an office in Silicon Valley and customers that include Agere, TI, Broadcom and Transwitch.

This article describes the philosophy of our approach to off shoring, the models we explored, how we chose the work to offshore, the costs we were aware of (and those that we discovered), the opportunities and challenges we have faced and what we have learned so far in this ongoing process.

Initial Stages: Why, Where and How

Tensilica is a very technology- and engineering-focused company. Its founders and engineering team has vast experience in microprocessors, compilers and electronic design automation (EDA) technologies. They have worked at major US firms such as Intel, Synopsys, MIPS and SGI. Due to its intellectual property business model, Tensilica relies on intensive, sustained leading-edge technological innovation by its engineers.

Our business model has two main expenses: employee compensation and EDA tool licenses. During the downturn of 2001 and 2002, the engineering budget wasn’t growing, while Tensilica’s business was slowly but surely growing. With more licensees of our products and more widespread use of our technology, we had to invest in increasing the ease of use and robustness of our products. However, allocating resources to this purpose would have to come at the expense of further technological innovation, which would be dangerous given the ferocious competition in our industry segment.

In August 2003 we decided to explore low-cost alternatives to address these issues.

Though Tensilica is a technological innovator, we didn’t want to be pioneers in offshoring. We wanted to leverage the experience of other technology companies and follow a well-trodden path. The choice of doing this work in India, therefore, was easy. There were many companies, small and large, with engineering development centers in India. Plus, we had contacts, could attend seminars and could use many other means of gathering information about working in India. The method with which to secure an engineering team abroad, however, was less clear.

We explored four different approaches:

1. Outsourcing

Contracting a set of tasks to a service company in India. We talked in depth with three companies about a specific project. As we discussed this in detail, it became clear to us that though the service companies would be able to provide expertise to do this and complete the task, our investment would be too high. Due to the high engineering content of our technology, there would be a delay while we trained their engineers to do the task. Moreover, the contractors couldn’t guarantee us access to the same engineers we had trained for subsequent projects. This made outsourcing a difficult choice as we might lose the value we created in the training process. Also, since IP is the core of Tensilica, its protection is a major concern. For this reason we chose not to outsource certain projects.

2. Set up a subsidiary

This is a very of attractive choice, since it addresses the concerns of IP protection as well as the training investment. However, it comes with an initial investment in infrastructure and organization overhead, which we deemed to be too high for the scale of operation we contemplated. Since we are an IP company, with no silicon tapeouts, testing or manufacturing, we don’t need a large number of employees. It would be adequate to have about 15 engineers during the first year. So, though a subsidiary was an eventual goal, we searched for a model that wouldn’t have a large up-front investment.

3. Acquisition

This choice has the very real advantage of bringing in a team that already exists. As is clear now, building up a well-qualified engineering team in India is a difficult and long task. We explored this option in detail, but eventually determined that no prospect was found that met our needs in terms of engineering expertise, location and cost.

4. Build-Operate-Transfer

This was our eventual choice. We decided to hire and build a team dedicated to Tensilica projects in partnership with a company in India that had domain knowledge, local infrastructure and expertise. We chose eInfochips to be our Indian partner and are working with them to build a team in Pune and operate it under their aegis for about a year and a half. The transfer to Tensilica occurs when there is a “critical mass of engineers to justify the cost overhead of infrastructure and additional services.

During the “build and operate” portion of the agreement, Tensilica pays eInfochips a fixed per-month per-engineer fee. The transfer will occur, at Tensilica’s discretion, in 18 to 24 months and will entail a pre-negotiated transfer fee.

The advantage of this method is the almost zero initial investment, low financial risk and quick start up. We have a team of engineers whom we have hired and trained in collaboration with eInfochips and who are dedicated to Tensilica projects. They’re aware that within a period of time, they’ll become Tensilica employees.

There are some concerns with this model, of course. Will we have the freedom and flexibility to inculcate Tensilica culture and values into this team? Also, we have seen that recruiting is potentially harder into this model than directly into a subsidiary might have been. There are risks associated with events in the future too. Will the transfer occur smoothly and on time? Will all the employees transfer over, or will some want to stay on at eInfochips?

We also explored three locations for this offshore center: Hyderabad, Bangalore and Pune. The final choice — Pune — was based on the higher retention rates of engineers compared to the other cities, a fairly good infrastructure, good weather and quality of life.

We knew that the initial recruiting in Pune would be harder than in Bangalore, since there is a much larger pool of qualified engineers available in the latter city. However, we decided that the combination of good engineering colleges and universities in the Pune/Mumbai area and the lure of opportunities closer to their hometowns for many engineers currently transplanted to Bangalore or Hyderabad would increase the available talent pool.

Choosing the Work To Be Done

Our philosophy on choosing the work to be done offshore consisted of the following:

1. Only R&D work to be offshored initially (in other words, no customer support, IT services, etc.). It may make sense, once the operations are firmly established, to do other kinds of engineering work in India. For example, support of customers in Asia or Europe might be more conveniently done from India rather than California due to closer time-zones. We have decided, however, to focus first on R&D operations.

2. Do additional work there; in other words, supplement the work done in Santa Clara, not replace it.

3. Do testing and QA, but also development, in India.

4. Fundamental innovations will continue in the US. We see this as imperative due to tight coupling with other functions within the company and the unique skills in the existing team.

Tensilica’s IP consists of both hardware and software components. The greatest need for increased ease of use and robustness was on the software portion of the product. We began by taking two different approaches to the work we decided to do in Pune.

Task 1: Additional development work for a portion of our product. It has aspects of new feature development as well as testing, requiring skills in both graphical user interface development and in embedded software tools. Task 1 is an absolute necessity for the company, and there is no viable alternative to doing this in India.

Task 2: A QA-only project that was seen as more experimental than absolutely required. It was chosen by the US manager as a means of “trying out” the offshore model, and there was skepticism as to whether it would be a success.

Over time other tasks have been added. They share the major attributes of supplemental work and include aspects of development and testing.

What we learned in terms of task allocation was to put only “must have” tasks on the list to offshore. Task 2 was only partially completed. The large amount of effort required on the part of the US manager, with a long learning curve of the engineers in India, eventually led the US manager to decide that his efforts were better spent elsewhere.

Expectations vs. Reality

Building the Team

We expected to have a team of about 10 engineers in Pune by mid-2005. By that time we were only at five engineers. Recruitment has been a major issue. The recruiting of more seasoned engineers (with five or more years of experience) has been particularly difficult. There are various reasons why this is particularly hard.

First, there isn’t a large pool of well-qualified, experienced embedded software engineers available in India. Embedded software and systems engineering has come to India only in the last few years, so the population of experienced engineers in this field is still developing.

Second, the BOT model is a barrier. There’s reluctance among the type of engineers we are trying to hire to join a services company. They want to work on a product and have ownership of their work. Though they work on Tensilica’s product, the appearance of working for a service company where the transfer to Tensilica isn’t guaranteed is a hurdle to overcome.

Third, name brand matters a lot. This is a very important but often overlooked point. Tensilica isn’t a well-known, large multinational company. It’s well-known, for its size, in California, but it’s hardly known in India. Engineers need to be able to explain to their parents, spouses and neighbors who they work for — and the name needs to be instantly recognized. This is true also of eInfochips. Though the company has been in India for about 10 years, it has only recently expanded to Pune and isn’t a large company.

We’re addressing these issues by resetting expectations to some degree, in that we’ll have to hire relatively less-experienced engineers and train them. It’s important, though to have at least one local manager for each group or project. Without this daily supervision and help, it’s almost impossible to manage the team from the US.

The BOT model issue is being addressed by doing things to make the team feel part of Tensilica. They’re given Tensilica stock options and receive other incentives. We’re actively working to increase Tensilica’s name recognition in India. For example, we proactively participate in conferences and are forming relationships with top-notch universities to expose the embedded systems and digital signal processor communities to our technology.

Close and very proactive involvement with the entire recruiting effort is required. Though the BOT model provides a ready-made infrastructure and direct recruiting help, we have found that our staffing specialists and technical managers have to invest significant time and effort not only in interviewing, but also sourcing and closing candidates. We update the recruiters in India with our job specifications, requirements and roadmap of activities to show to the candidates and specifics on the technology. Getting each individual excited about the company and the work is important. Due to the highly active employment market in India and the multiple offers (and offer acceptances, it seems) that each potential recruit has, we remain in constant touch even after an acceptance, to increase the probability of the person starting at the company. (We have about a 50% success rate so far!)

Management and Startup Efforts

I can’t overstate the level of management bandwidth that goes into starting an India design center and building it up. Currently our technical managers spend an average of 10 to 15 hours a week on various India-related activities. Of course, this is over and above their work that needs to be completed in the US. The time zone difference exacerbates the stress of the effort. Being on the phone or sharing workspaces on the computer after dinner or early mornings with the India team takes its toll.

On the other hand, successful completion of projects and increased productivity at low cost are the fruits of that effort. Tensilica is beginning to see some positive results, but we still don’t have enough of a history to definitively declare the India initiative a success.

Build-Operate-Transfer Model

The BOT model came with the promise of low initial investment, low financial risk and quick startup. Certainly, the first two have borne out so far. The quick startup hasn’t occurred. Though, we didn’t have to spend any time with the initial setup of a legal entity, getting office space or selecting the accountants, we also didn’t succeed in getting an effective engineering team in place quickly. That effort continues.

The long-term risks of a smooth transfer also remain to be seen.

Effect on the US Team

When this effort was begun, we were aware of the importance of getting the US engineering team on board. Though the team was fully bought into the idea at an intellectual level — clearly it was good for the business of the company — it has been harder to get them to embrace the idea.

Due to the nature of the work we do and the high level of expertise required, the engineering team isn’t insecure about their jobs. In fact, they clearly recognize that there is truly much more work that should be done than can be done by the current US resources. The doubts and skepticism stem from two main sources:

1. The ability and productivity of the Indian engineer. Doubts remain about how much money the company is saving when it takes time to hire and train them. The staff is unsure about their basic engineering ability. The approach could potentially use more engineers than required in the US. It comes with communication issues that hamper efficiency.

2. Lack of growth potential in the US. With most of the increase in headcount in the engineering department currently occurring in India, there’s doubt about career growth opportunities in Tensilica US.

Tensilica decided from the early stages to be open and communicative about the entire process, plans and status with its engineering team. As the team interacts more with the India engineering team, considers them more of an extension of the team in the US, and sees personal growth opportunities in a global workplace, the emotional acceptance of this effort will also be there.

The Lessons We’ve Learned So Far

Tensilica is still at an early stage of setting up operations in India. We’re trying to create an excellent engineering team in India working on high-technology areas of embedded microprocessors and systems-on-chip design. In this short period there are several important lessons for us.

1. Select the engagement model carefully.

Build-operate-transfer is a good model to start with low initial investment, low risk and quick infrastructure and legal setup. If the projected growth in the first couple of years is less than about 30 engineers, it makes perfect financial sense. However, you mustn’t rely too much on technical input or technical managers from the BOT company, especially in a complex engineering arena like embedded processors. Be prepared to be heavily involved. Be sure that the BOT company is trustworthy and open about communication and has the bandwidth to service your needs.

2. Select the work done offshore carefully.

It’s difficult to attract the right talent in India if you’re only doing QA work. There must be a mix of development and testing projects. Select projects that are as self-contained as possible. This reduces communication issues. Don’t do purely optional projects offshore. The overhead of training and communication is too high to keep the US team motivated to complete such tasks.

3. Recruit the team leads first.

Though it’s easier to find junior engineers, having a team in India without a leader to guide them on a daily basis is counter-productive and leads to instability.

4. Recruiting is hard!

Closing a candidate and ensuring that he or she shows up is harder still.

5. Building close rapport between the teams and understanding communication styles goes a long way.

Travel and working together physically helps engineers understand each other and opens communication channels. Travel should be both ways, with US engineers going to India and vice versa. Getting full commitment from the US team is hard. Being open, communicative, and providing opportunities for global teamwork is required. It’s absolutely essential that the US manager with projects being done offshore be fully committed to the effort. Without this commitment, the burden of the work and the inevitable hiccups are easy scapegoats for failure.

Tensilica actively continues its efforts on the India design center. In fact we have recently increased our commitment and investment there. We’ve hired our first Tensilica employee in India, the director of the center. Increased demand for our products, more demanding customers, price and competitive pressures and cost efficiencies possible in India make it necessary for us to succeed in this endeavor.

Useful Links