Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
Respect the customer: Uniquify’s Bob Smith offers rational advice
July 24th, 2014 by Peggy Aycinena
The rumors are flying fast this week about Apple’s next product announcement. Many believe it will be a wearable, possibly a watch-like device. Happily, I’m one step ahead of that Silicon Valley-based behemoth with months of research into my own wearable, which will undoubtedly swamp the market and outsell Apple’s wearable by orders of magnitude.
I had a chance to discuss my project with Uniquify SVP Bob Smith in a recent phone call and started by asking how he felt about the IoT, the Internet of Things. Is it simply a trendy phrase emanating from tech sector marketeers?
Bob said no, and recounted the delight of a friend of his who owns an IoT-enabled crock pot. “The thing has WiFi connectivity,” Bob said, “which allows the guy to turn on the crock pot remotely, and at the appropriate hour, so dinner’s ready on time, but not overcooked.”
“Sounds like you’re okay with the whole IoT thing,” I responded, “so how about some feedback on my Dick Tracy keychain. It’s going to allow me to have keyless entry and ignition for my car, to open and close the garage door, to know if there’s sufficient milk in the fridge, and to also tell the time. In other words, it’s got a limited feature set, but importantly nobody will ever get locked out of their car again because the keychain will be strapped to their wrist.”
Bob commended the designated feature set, noted its simplicity and usefulness, and then agreed with one of my conclusions: After many months of conversation with IP companies about developing my product, the Product versus Services & Products is a legitimate topic when discussing the IP business model with vendors.
Per Bob, “Certainly, the IP business is about both products and services. The more complex the IP, in fact, the harder it is to get it to work right when meeting the application requirements, and the more it has to be customized [by the IP vendor] to match the design ideas of the customer.
“At Uniquify, we’re in the technology development space and we understand well the [issue of Product verses Product & Services]. Our original IP, the DDR 1 or 2, you probably could have used straight off the shelf in a design. But with our DDR3 or DDR 4, these products are super high-speed and customization is critical to the success of these blocks in the overall design.
“Getting the DDRs to perform properly means they need [to reside] inside of a great layout and have interconnects with the rest of the chip. To accomplish this, needs a lot of attention, both from the customer and from Uniquify.
“In our business today, there is no way we would hand our PHYs off to a customer. We absolutely need to do the layout for the customer. The people at Uniquify have done a ton of these PHYs and know exactly what it takes to get it right. We don’t want to just stand behind the quality of our IP, we want to be there with our customers right through to the I/O design.”
“And by the way,” Bob added, “if you look at any complex IP sold by any vendor when it involves a design that’s got higher speed, things like Ethernet, MIPI, PCI express, or SerDes, unless the customer plans to do something at really low speeds, they absolutely have to have some customization from their IP vendor. And when that happens, voilà, the IP vendor is offering services along with their products.”
“So then,” I asked, “is there such a thing as plug-and-play in the IP business?”
“The answer is yes,” Bob replied, “as long as it’s lower complexity and lower speed. Then it can be considered ‘commodity’ IP, something simple like a UART that’s relatively small. Those types of IP you can probably use right off the shelf and are often [recognizable] because they’re delivered as RTL code. For us, we only deliver RTL for controllers, because these are generally not too hard for anyone to use in a design.”
“But actually,” Bob added, “I should make a note. Even when we deliver the RTL for a block, we also deliver services. Customers, for instance, might say they need a particular kind of arbitration when the signal into the controller we are supplying, so then we are providing services and advice [to optimize design decisions].”
“The next big question, as always,” I said, “is how much do you charge for services?”
Bob said, “If you license our controller and it’s targeted at a process for which we don’t have the I/O, we can focus on the PHY. If you buy the PHY from us, we bundle into our pricing the hardening of the IP, but there are still some options. If the customer asks us to harden the PHY and I/O, then we have fixed costs.
“[Alternatively], if the customer says they don’t know much about the DDR, they just want a hardened black box that includes a controller, a PHY, and a DDR, then there’s a different fee for that. Or if the customer has a whole bunch of things that need to be added, in the end we will figure out a Time & Materials pricing. In that case, it’s moving much closer to the model you describe as a full-service design house.
“And how is all of this sorted out?” I asked.
“It’s a complicated endeavor,” Bob said. “Everybody involved has to sit down together, say here are the specs and what the product needs to do, and here’s the IP needed. But even then, chip design never follows a simple path from A to B to C. It’s pretty rare that there’s consensus on how to do things.
“But we are always building project based on customers’ concepts. If there’s a delay on their side, for instance if they’ve just realized that their test strategy is flawed, they might put us on hold for a week which means our people are kind of idling.
“We can tolerate that for a while, but it’s also built into our contract that if the customer goes beyond a certain grace period, we have to put our team on [an alternative project] and need 3 weeks notice [to get the team back on the first project].
“Of course, every project is going to need some kind of slack [in the relationship] between the vendor and the customer. There are always going to be the inevitable delays. Although, because we have good mechanics built into [our business model and our technology], our customers often know what they want at the outset because of the detailed form we ask them to fill out. Once we get going, they can come back with changes but there are far fewer surprises.”
Impressed, I asked Bob if Uniquify could build my whole chip for me.
He said, “Actually, we’re not in the ASIC business model. If you want that, see Jack Harding at eSilicon. They may, in fact, subsidize the cost of a design because they believe they will make so much money manufacturing the chip. In other cases, they may not.”
“In our model, however,” Bob continued, “because of our roots in the business and because we are self-funded, we will design your whole chip but we will separate out the [parts]. We would prefer not to touch the IP design, because then we can be specific: Here’s what it’s going to cost and here is how we will make some margins.
“Margins in the service business, however, are not super good. The customer may say, here are all the specs, but [continue to refine] the specs as they move forward. Sometimes [those change] discussions happen all at once, but often not. It can be complicated to re-negotiate the costs once the job is already underway.
“Nonetheless, it’s not uncommon for a customer to change their mind mid-project, perhaps they need more of a certain type of IP, or a different type of IP that what has already been set up. Then we have to rework our costs, and sit down with the customer and amend the contract, especially if we’re going to be delayed because of their changes.”
Bob offered optimistic note: “The best defense for a business like our is a good offense. Generally, the people we meet in the chip-design world are reasonable. They know we’re not McDonald’s or Starbucks, that we’re not selling the same thing over and over again.
“And we know that every DDR3, for instance, is slightly different and each customer has different needs. We also know that some projects can take on a life of their own, but there’s always a schedule that we’re working to.
“In the end, it’s all a matter of negotiation. We’re not in the business of skewering our customers, where we take it all if the customer [changes the project]. We do ask that people be honest with us, however, and fess up to the changes. If changes are needed, generally we find they’re rational changes because the customers are rational.”
2 Responses to “Respect the customer: Uniquify’s Bob Smith offers rational advice”