Director, Product Management, IoT Solutions for Retail / CPG (2014 to 2016)
By 2014 I had spent four years arguing that the website is the store, that content and commerce are one experience, and that the transaction should happen wherever the customer already is. So when Intel came recruiting, the pitch landed on a nerve I already had exposed. The Internet of Things was going to do to physical retail what the web had done to catalogs: every shelf, sign, cooler, and fitting room was about to become addressable, measurable, and personal. Intel had just formed a new business group to chase that future, and the team it was standing up deliberately recruited software-industry people; I came in through the software side of the house. The job description was almost funny in retrospect: a director of software solutions for retail, inside the company whose silicon sat behind America’s checkout lanes. I took the seat, and I wrote my own mission in my journal that spring: collaborate and innovate with retail and consumer product companies to reimagine the shopping experience, powered by Intel Inside.
This chapter is not a victory lap. There is no revenue number here that I would put my name on. What Intel gave me instead was two years inside a world I did not know, silicon economics, embedded systems, design wins, fleet-scale unit math, and what I gave Intel back was the thing I have always been able to give: an evidence-driven read of the market, even when the read was that we did not have the right pieces. Almost everything I concluded in 2015 about Intel’s IoT stack became, over the following few years, public record. Mostly this is a story about learning to do the math.
A different world
I had never worked in manufacturing, let alone a true “silicon” valley company, and I did not fully understand the position I had accepted. I thought I was taking a product management job. What I actually inherited was the Intel version of a product line manager, a role that runs from business development through sales and marketing, and a mandate that amounted to a box of Lego pieces and a challenge: find new ways to get Intel Inside the nascent Internet of Things. The pieces were real, and expensive. Intel had spent years buying its way up the stack: Wind River for the embedded operating system, McAfee for security, Mashery and Aepona for APIs and services, alongside the processors themselves. On paper it was the most complete collection of IoT ingredients in the industry. The company’s own marketing said so: only Intel is experienced from silicon to software.
I also got an education I could not have bought. I learned electrical engineering by osmosis and by embarrassment, the history of microprocessors and integrated circuits, and the peculiar discipline of the design win, where you sell your component to someone who builds a machine that someone else deploys, so that the customer of your customer decides your fate. A person who has only ever sold software does not really understand how hard that is. I do now. When a deadbolt lock with Bluetooth needs only a sixteen-bit microcontroller that costs less than a coffee, the question of what exactly Intel should sell into that world is not a marketing question. It is an existential one, and living inside it for two years is the closest thing to a semiconductor MBA I will ever get.
Two months in, I stood up at our group offsite with a method rather than a product, because a method was what I had: name every player in the retailer’s ecosystem, define the segments, hypothesize use cases, validate them with real customers, and only then develop a solution. It reads like common sense. Inside a company built on roadmaps and fabrication cycles, going to the market first was closer to a heresy, and it became my whole job. By that first autumn I had discussions moving with Hudson’s Bay and Saks, REI, Nordstrom, and Westfield’s innovation lab, with Sephora, American Eagle Outfitters, and Nike queued behind them. I was not selling anything yet. I was gathering market requirements, and they would end up steering everything that follows.
Taking the experience thesis into the aisle
The thesis I carried in from my Adobe years was that the gap between a brand’s digital experience and its physical one was closing, and that marketers, not IT departments, would drive the closing. At Intel I finally got to argue it on the biggest stages of my career to that point. At IMPACT15 in Las Vegas I told a room of digital marketers that the Internet of Things would make all marketing digital: that the century-old lament, half the money I spend on advertising is wasted and I do not know which half, was about to expire, because when the shelf, the sign, and the store itself become instrumented, for once we will actually know which half is working. At Tealium’s Digital Velocity conference I gave the keynote I published that same week as an essay of the same name, Moving Digital to the Physical Realm: settings and scenarios become more important than channels, physical traffic becomes as measurable as web traffic, and the store can respond to you as personally as a homepage does, sometimes as subtly as an LED strip lighting the aisle in your chosen color. And I meant what I said about the bargain underneath it all: these systems run on personal context, so the winners would be the brands that could promise, credibly, not to abuse it.
Inside the building I pushed the same idea with less poetry. Intel’s retail story had grown up in an embedded-systems world, organized around the devices it already powered: the point of sale, the digital sign, the vending machine. My argument, in customer briefs and in the spring 2015 planning cycle, was that the unit of value was not the device but the experience, and that Intel would win by engaging and supporting the ecosystem that builds experiences rather than by producing more products that faced inward. The phrase experience driven began appearing in the group’s product portfolio planning after months of those arguments. I will not claim a causal chain I cannot draw; the plan itself was a talented cross-functional team’s work. What I know is that I said it early, out loud, with customer evidence behind it, and that watching the frame stick was one of the quiet satisfactions of that year.
The partner motion followed the same logic. If retailers chose their technology with a systems integrator or a digital agency at the table, and every piece of my customer research said they did, then Intel needed those firms inspired to carry Intel with them. I re-engaged the agency world I knew, Sapient and Razorfish, and helped put together a store-of-the-future roadshow with Razorfish and Deloitte Digital. I championed an Android-based reference design that aimed a digital-signage player at a price point under a hundred dollars, because the entry tier of that market was going to ARM and Android with or without us, and I had already learned the first rule of markets: stop fighting them. And with independent software vendors I ran the play that became my template for years afterward: pick the solution categories retailers were actually funding, find the best software players in each, and make Intel the architecture they build on, with our team supplying the enablement, the engineering introductions, and the reasons to bother. Three of them ported their applications to Android on Intel architecture that year. One, a beacon-technology company, went further and converted a device from ARM to Intel architecture outright, the only time in my two years I got to feel what a design win feels like from the winning side.
No engagement went deeper than American Eagle Outfitters. What started as introductions through my retail relationships grew into a standing series of store-of-the-future workshops, many of them in Pittsburgh, and eventually into a continuing advisory role in the work. The moment I am proudest of from that whole engagement is a quiet one. In July 2015 their executive team gathered to hear a major agency pitch the future of their stores, and they asked me to be in the room, not to present, but to help them evaluate what they were hearing. I sat on the retailer’s side of the table, taking notes on someone else’s pitch, flagging where the concepts would and would not survive contact with a real store fleet. Intel’s name got me the meeting; being useful with no product to sell got me invited back. I did not know it yet, but I had just discovered my favorite seat in the industry. Those workshops earned us the chance to test ideas together rather than trade slideware, and I learned more about how retail executives actually make technology decisions than any analyst report could have taught me.
“It doesn’t pencil out”
The education that stayed with me longest came from the customers who told me no.
The product I most wanted to win with was an RFID sensing platform a colleague in our retail division was building: ceiling-mounted readers, powered over Ethernet, designed to see the tagged inventory in a store and place it on a map of the floor. Inventory accuracy is one of retail’s oldest and most expensive problems, and this attacked it with real engineering elegance. I believed in it enough to carry it personally to retailers where I had relationships, Gap and American Eagle among them, and to sit in the meetings where they took it apart.
A very senior executive at Target gave me the sentence I have repeated in every product role since. Vendors come to her constantly with clever technology, she told me, and almost none of them do the simple math of multiplying their idea by the number of stores she operates. It just doesn’t pencil out. Retailer teams made the same point with their own numbers in working sessions that spring: even tags at four to six cents felt expensive at chain volumes, before a single reader went up in a single ceiling, and the whole system had to beat the humble alternative of a store manager who already knows what is on the shelf. Run the same arithmetic across any large specialty fleet and tag costs alone reach seven figures a year, with the capital cost of readers amortizing over years on top, all to chase an inventory-accuracy gain retailers kept telling me they could partly capture with process instead. One retail operations lead summed up the whole category with a joke I still quote: more hardware, more problems.
Here is the part I am proud of. I kept believing in the product and kept doing the math on it anyway, and I refused to let those two be in tension. In September 2015, the same month the platform’s pilot with Levi Strauss reached the trade press, I sat down with the business plan and wrote over eighty margin comments in a single afternoon, and the ones that mattered were pure Target-executive math: price times units times stores, versus the software and integration costs nobody had penciled yet, next to blunt questions like whether we had a documented market requirement from a single retailer or just our own hope. I do not know how much the comments changed; the program rolled on to its showcase at the industry’s big January show, and its champion deserves that credit. What they changed was me. My own conclusion by the end of that year was that the platform was ahead of its economics: right idea, wrong tag price, and only time would fix the arithmetic. Enthusiasm and arithmetic, both, in writing, at the same time, became my standard. Every pricing decision I have made since, at IBM and at Yottaa, has that unnamed Target executive somewhere in the margins.
Reading the stack
While the retail work taught me unit economics, the product work was teaching me something harder: the pieces themselves were not right, and the market kept saying so in plain language.
I ended up directing much of our market-requirements work, which meant I heard the evidence firsthand and in volume. In 2015 I found a process gap, built a market requirements document template, and mentored a rising star on my team through the first one, a clear-eyed study of why the explosion of tablets and mobile devices in retail stores was going almost entirely to Apple and Android on ARM. Its opening question was the honest one, is Intel positioned to win, and its opening answer was, not currently. The template and the discipline around it spread; by the following spring the group had ratified a standard version, my team was fielding requirement documents built on dozens of named customer and partner interviews, and I had personally interviewed everyone from cloud platform giants to gateway makers, systems integrators, and a federal agency at industry shows. The method was the point. Requirements traced to a named source, debated hard, decided once. One line from an operating-principles slide I wrote that year still summarizes my product philosophy: rigorously debate everything, then make a decision.
What the interviews said, over and over, was that Intel’s IoT stack had been assembled for a market that had not arrived. Our small IoT processor could not compete with ARM on cost or capability, a verdict so consistent that one requirement in our own gateway study was bluntly titled to say it was not a viable option; the public post-mortems, when the product line was eventually discontinued, read like transcripts of those interviews. On operating systems I will claim a little standing of my own, having administered systems and designed QA tests in an earlier life. Wind River was, and remains, a serious embedded software business in the worlds where hard real-time and safety certification rule. Our problem was narrower and harder: in the thin-margin device tier we were testing, a commercial embedded OS was built for an era when device makers cared which operating system they ran, and that era was ending underneath us. The market wanted free Ubuntu, and the flagship IoT gateway of one of our biggest OEM partners, which carried our own embedded OS on its option list, still put Ubuntu in the launch spotlight. Our security acquisition, whatever its strengths up-market, had almost nothing for the smallest footprints of embedded devices. Our API and services acquisitions were good products that did not add up to a platform a device maker could build on. One partner told us our gateway reference design was a great way to start a conversation and not something anyone took to production, and then asked the question that hung over everything: where is the revenue? Measured against the scale of the established silicon business, the answer was not one anyone enjoyed saying out loud.
And I struck out a lot. That sentence stays, because it is true and because everything I learned came through it. I spent two years trying to fit square pegs into round holes with real conviction and real effort, and I heard our kit was too expensive or underwhelming from more prospects than I can count. Some of my own reads were wrong too: I was sure our sub-hundred-dollar Android signage play would take back the entry tier of that market, and it did not. I had read the price pressure right and underestimated how much of the product was ecosystem, certified media software, device management, distribution, developer momentum; cheap hardware was necessary and nowhere near sufficient. The low point that taught me the most was a day-long strategy review with a major partner in the summer of 2015, where my own top takeaway, written in my own meeting notes that night, was that we presented poorly and looked disconnected from the industry. They were focused; they saw us as scattered. They had given up waiting on parts of our stack and moved on pragmatically, and they were right to. I have sat through many hard meetings since. That one recalibrated my sense of how fast a market moves when you are busy admiring your own strategy.
The case that was too big for the room
All of that evidence pointed one direction, and by the fall of 2015 I had followed it to an uncomfortable conclusion: the layer of the stack that was winning the Internet of Things was the one Intel did not own, the low-cost, low-power silicon that Texas Instruments and its peers built on cores licensed from ARM. So I did what a product person does with a conviction. I wrote it down. In October 2015 I drafted a strategy, revision zero, that proposed going big in IoT silicon: sell 32-bit processors at 16-bit prices, offer the industry’s most consistent software and tools across every platform, and form a strategic alliance with Texas Instruments, with the explicit ambition of migrating TI’s processors from ARM to x86 and covering the gaps Intel could not. The background section stated the problem without cosmetics: Intel had no footprint in the microcontroller space where the actual volume of things would live. In the hallway versions, the idea was blunter still: if the silicon layer was the prize, buy a company that owned it.
Two things are true at once here, and the story only works if I hold on to both. I was not in the room where decisions about billions of dollars of M&A get made, and I will not pretend I was. I was a director with a thesis, a deck, and a growing file of customer evidence, sharing the idea with anyone in the flow of strategy and corporate development who would engage, and trying to figure out how to build a case for something I suspected was simply too large for the company to swallow. The idea did not carry. Most ideas that size do not, and plenty of mine have deserved their fate; if I had it to do again, I would have spent less time perfecting the deck and more time recruiting believers two levels up, because evidence does not climb an org chart by itself. Then the outside world weighed in. In July 2016, my last month at Intel, SoftBank agreed to pay roughly thirty-two billion dollars for ARM, explicitly as a bet on the Internet of Things. The day the news broke, a former teammate messaged me: we had a timely chat about an ARM acquisition a couple of weeks ago, and it just happened today. While writing this chapter I went back into that old LinkedIn thread, partly to make sure I had not rewritten my own memory, and there it was, dated the day of the deal. I hold the two halves of that story apart deliberately. SoftBank’s check did not prove my prescription, and it certainly did not prove Intel should have written one like it; whether my version would have worked is a question nobody will ever answer, and I can even laugh at the irony that dragging TI’s chips to x86 probably broke my own first rule of markets. TI’s moat was never just an instruction set; it was analog breadth, tools, channels, and a developer base, which is exactly why the alliance half of my idea was always stronger than the migration half. What that thirty-two billion dollars corroborated, loudly, was the diagnosis: the layer Intel did not own was the strategically central one. The diagnosis is the part that aged well. And years later came a coda I can only smile at: the executive who had run our strategy office, one of the people I respected most there, joined ARM’s board.
The narrowing
Through the spring of 2016 the ambition I had been hired into was narrowed away around me, and I watched it happen at close range. Intel’s company-wide restructuring that April refocused everything; inside our world, program after program that I had helped scope was stopped, resources ramped off, gateway plans cut back hard, and my team’s own support structures reduced. None of this was villainy. It was a large company moving, unevenly, in the direction the evidence pointed, concentrating on the layers where it could actually win. But it also meant the path I had joined to build, software-defined solutions for the connected store, no longer ran through Intel. Our division’s sponsoring executive had moved on, the team was on a course to be disbanded, and when they offered me ownership of the gateway product line to stay, I understood the offer as the compliment it was and declined it anyway. I had read that market’s requirements personally, forty interviews deep. I knew what I would be signing up to manage.
The full story of the phone call that pulled me out belongs to the next chapter. What belongs here is why I was ready to take it: I had done the thing I came to do. I had tested the best collection of IoT parts in the industry against the real market and reported back exactly what it said, and I was leaving with more useful scar tissue than most jobs hand out in a decade. The epilogue began writing itself while I was still in the building, when the API and services businesses were sold, and finished over the next few years in public: the security business was spun back out, the small-silicon dev boards were discontinued, the embedded OS company was sold (and, credit where due, went on to thrive under new owners), the processor line itself was retired, and the IoT group as I knew it was eventually folded into a larger edge-computing organization. Every piece I had flagged as not the right part was eventually sold or discontinued. Those exits had their own causes, portfolio logic as much as product verdicts, and I do not mistake them for proof of my influence. What told me the read had been right was narrower: the reasons stated publicly matched what the interviews had said years earlier, cost against ARM, free operating systems, footprints too small for the security stack. A divestiture alone proves little; the reasons come closer. I take no joy in any of it. It is just what the evidence said in 2015, and I happened to be one of the people listening.
What I take from it
Intel is the chapter where I earned the right to talk with hardware people. I walked in a software and marketing guy who had never seen a design win up close, and I walked out able to read a connected-device stack from silicon to cloud, argue unit economics with a straight face, and hold my own in a room full of electrical engineers, at least until the third whiteboard. I would not trade those two years for a smoother-looking line on the resume. Getting to work inside the company that built the modern computer, while it wrestled with what computing would become next, was a front-row seat to one of the great strategy problems of the decade, and I got to poke at it with real customers and real silicon.
Three habits came home with me, and I still use all three weekly. The pencil-out test: multiply every beautiful idea by the number of stores, machines, or customers, add the costs nobody penciled, and make it beat the humble alternative, before falling in love with it, the way a Target executive once taught me. The stack read: for each layer of a system, price it against the market’s alternative and ask who is actually paying for it, then say what you find while it is still useful, in writing. And the partner instinct: markets are won by the ecosystem that carries you, not by fighting the direction the market has already picked. I also learned where I most like to sit, in that advisor’s chair on the buyer’s side of the table, and I keep finding my way back to it.
And the thesis that brought me to Intel held its direction even where it missed its timetable. The physical world really is becoming one more surface for digital experience; it just arrived on a different schedule, and on different silicon, than any of us drew on our roadmaps. What kept my conviction alive through two years of hearing no was the retailers themselves: every workshop, every pencil-out session, every executive who said not yet was also saying, keep going, this is coming. I left Intel more energized than I arrived, impatient to keep building toward it. All I needed was a company whose pieces could actually deliver the experience layer I kept sketching. Whether one existed, and what it would take to wake it up, is the next chapter.
This chapter is part of My Work, my career told one company at a time.
← Adobe · All chapters · IBM →