I first heard the term “technical debt” about eight months ago. I heard it more in a mechanical engineering context, rather than software, but since then have seen it pop up more and more there too. As a concept, it’s the kind of thing you hear and start nodding your head – yeah, I get it, technical debt – but I’ve found that locating a precise and practical definition of it is more difficult.
For my purposes, let’s call technical debt the opportunity costs of current enterprise technology architectures.
To expand on that, you first have to think through what “opportunity cost” means. Every technical decision you make locks you out of other choices by default. Those other choices – the roads not taken – are the opportunity costs of the choice you made. If you opt to implement one specific ERP for the majority of your enterprise capabilities, you have, by default, locked yourself out of, for example, a best of breed strategy: rating strong internal integration as a priority automatically shuts out a strategy that focuses on having the richest capabilities to meet every business need.
Technical debt isn’t just about the solutions a company has purchased and implemented. Debt also accumulates over time through the use of hard-coded integration points between solutions, as well as a piling up of custom code. In retail, one place where this is especially egregious is in store systems. Back before Montgomery Ward went under, I was part of a project to evaluate how to replace their point of sale (POS) solution with something new. But over time, the company had written custom functionality, which was hard-coded, into their POS to such a degree that the original software (which had once been a packaged solution maintained and enhanced by a software company) was unrecognizable. The technical debt of what they’d done was so large that in the end, it played its own role in bringing down the company, by requiring such an investment in technology to overcome it that the company couldn’t afford to pay the price (there were other reasons why the company couldn’t afford it, but the technical pressure was still overwhelming).
Retail has never been good at measuring, let alone managing, their technical debt. There are a lot of reasons for this. One, there are often competing interests inside the organization that are never resolved, even within executive leadership. It often comes down to an IT organization that is trying to minimize technical debt, versus a business organization that doesn’t want to give up “perfect fit” in order to avoid creating technical debt in the first place. Some of that comes from no one on the business side taking a holistic view of the enterprise – when the CMO wants a solution for marketing that she swears they need, and can provide a compelling business case to prove it, and she’s paying for it out of her budget, she’s not really going to care too much if it makes IT’s job harder. That’s their job to work through technology issues, after all, and they should get around to doing it.
That attitude comes from a budgeting process that does not calculate or charge lines of business for the technical debt their technology decisions incur. I’ve seen it plenty of times – a CIO who says “We’re going to standardize on this database, and if it’s not on this database we’re not going to buy it.” And that lasts for about 10 minutes, when some LOB leader comes in and says “I don’t care that this solution doesn’t run on your selected database, I need it.” Does that business leader pay for the additional costs to the IT organization of supporting a non-standard database – another set of skills to bring in-house, another set of licensing and maintenance contracts to negotiate (without much leverage), etc? Nope.
Sometimes the IT organization itself drives that behavior. The company stance is “IT delivers whatever drives business value.” That works great, as long as “business value” is calculated to include both the direct IT costs of supporting a capability, alongside a recognition of the opportunity costs incurred by whatever technology decisions are made to enable the capability. When it does not include those costs, then a company ends up incurring an enormous pile of technical debt. No wonder, in those instances, the IT department is viewed as a cost to be minimized, rather than an asset.
Technical debt is like heart disease: it can be a silent killer. It piles up with no real symptoms, until, like for Montgomery Ward, it crushes the business. The only signs of trouble come from IT – a growing backlog of enhancements, enhancements that increasingly cost more and take longer, and even failed new implementations where the integration requirements smother any new capabilities. If a business faces a big, wrenching change, technical debt becomes a sudden albatross when the capabilities underpinning the business must also change, and change rapidly.
When you look at retail today, that wrenching change in expectations is exactly what has happened. Consumers have substantially changed their behavior and expectations, and it’s causing a lot of technical debt to come due in retail, right now. But consumer expectations are only the catalyst. The real issue is that up until now, no one was really aware of just how expensive the technical debt was getting. Everyone had the same amount of technical debt as everyone else and it was just as expensive no matter which company you looked at.
That’s not true any longer, thanks to two big changes: technology evolution, and the rise of the tech-driven retailer.
Technology Evolution Finally Comes To Retail
Retailers have a persistent belief that their industry is unique. While I wouldn’t necessarily discount it, I would maybe reframe that position. The individual challenges that the industry faces are not unique, but retail faces a unique collection of challenges:
A highly distributed enterprise. When you get into the realm of fast food and dollar stores, you are talking about supporting tens of thousands of locations. That is hard.
Online, in a weird way, actually makes this easier, because it aggregates demand in ways that retailers might not see when it’s distributed across a bunch of independent retail locations. For retailers paying attention to how the data coming out of their online operations relates to both stores and the customers who live around each store, there are suddenly a lot of new things to learn.
Poor network connectivity. I would not be surprised at all if turned out that retail is the industry that invented trickle polling – the act of dialing up the enterprise to send short bursts of data, usually sales transactions. Always-on, high-speed connectivity across even a thousand stores has, in the past, been desperately expensive. That’s why stores have often been referred to as “islands of automation” – they have had to be able to exist as autonomous centers of technology, capable of sending or receiving only short, small bursts of data at a time.
Between urban stores that might get clogged access and rural stores that didn’t have good connectivity choices to begin with, making sure every store is connected all the time has significant costs. It has only been in this decade that Moore’s Law has kicked in to the point where retailers can seriously consider the cost of a store server (and all of the support that requires) vs. high-speed bandwidth to that store. And even then, very few retailers have been willing to cut the store server completely, out of fear of losing that network connection.
Retailers will never give up on making sure that stores can run independently of the network. But the opportunity cost of having to choose between hardware in stores vs. investing in the network is changing significantly, to where it is increasingly possible to have a very small hardware footprint in stores, fed by a much larger amount of bandwidth.
An enormous amount of data. When the term “big data” started getting used about a decade ago, retailers laughed. Retail has always had a big data problem. Take forecasting demand. In a B2B environment, maybe you’re talking 1,000 items across roughly the same number of customers. That’s a million combinations of possibilities. Back in the early ‘00s, that wasn’t a trivial exercise – forecasting demand could easily have been an overnight operation, depending on how much data you wanted to use as the basis for your forecast.
A typical grocery retailer can carry 30,000 SKUs, and have millions of customers – you and me, complicated by having to consider both households, as well as individuals within a household. Then add in the complication of having to locate that inventory close to demand – on the shelf of the store where it’s most likely to purchased. Walmart has over 5,000 stores. Kroger has nearly 3,000 stores. If even 10% of American consumers shopped at Kroger (and it’s probably a lot more than that), we’re talking 4.5 trillion possible combinations.
Running a forecast of that size could easily take longer than the typical shopper frequency – in grocery, easily 1x per week, and more frequent the more urban the location. In other words, it would take you longer to figure out what to sell to consumers (let alone get it to the right place) than it takes consumers to decide to buy it. Retailers have worked around this problem through aggregation – looking at higher-level summaries, of either products or customers. Stores, historically, were a good proxy for all of the customers that shopped at a store. Sometimes retailers would look at demand only for a few key items, or would look at demand at a category level – for salad dressings in general, for example, rather than, say, demand for Ranch vs. Italian, let alone at the level of Kraft Ranch vs. Newman’s Own Ranch.
This much data has made them highly skeptical of data-driven solutions, not because they don’t trust those solutions – they just don’t trust them to scale to meet retail needs at a level that could be break-through for retail operations. But Moore’s Law, combined with cloud computing, have chipped away at this issue for retailers. And when you layer in the break-through pattern-recognition capabilities that come from artificial intelligence, now you have a significant shift in what’s possible – if retailers are not so burdened by their technical debt that they can actually access these capabilities.
No control over customer engagement channels. Finally, retailers have little control over how consumers choose to engage with them. This is part of what has disrupted the retail business model in the last fifteen years in the first place. eCommerce was additive at first, and then began to cannibalize store traffic, which had an impact on store sales. Retailers have been “right-sizing” their store count ever since. But then other digital channels popped up, and retailers have had to additionally invest to serve those channels, or risk losing customer loyalty – mobile commerce, Twitter and Facebook for customer service. The cost of supporting the old channels – stores, call centers, etc. – has not gone down. The new channels are additive, with additive costs that are not completely offset by additive sales.
Retailers can opt to not take on a channel, but if they don’t, then someone else will. And there are very few retailers that can impose a level of switching costs to keep customers despite not serving them the way they want to be served. Back to grocery – if you don’t offer some kind of curb-side pickup option in the next couple of years, odds are you’re going to lose business to retailers who do. Consumers love the food brands that are meaningful to them, not the retailers who sell those brands.
There are many industries who face one or other of these challenges, but not so many that face them all, all at once – while every one of these challenges is encountering so much disruption. No wonder there are cries of “retail apocalypse.”
The Rise Of The Tech-Enabled Retailer
While traditional retailers have struggled with these challenges, some upstart retailers – with an eCommerce-only heritage – have upended everything about retail. Walmart’s recent acquisition list is a neat summary of pretty much every kind of “new tech retailer” out there, from Jet.com to Bonobos, to Modcloth. And of course, all of these companies were purchased in order to bulk up against the ultimate disruptor, Amazon.
I’ve talked to quite a few of the upstarts, from CEOs to CMOs to CIOs. One thing that strikes me across all of the executives from all of these companies is that while they are all passionate about their individual brands and the products they sell, every single one of them is acutely aware that their success is completely dependent upon technology. Their business model, their cost structure, the data and insights that are the fuel of their strategy – none of these things are possible without major investments in technology.
Today, these retailers have no technical debt. That’s not to say they won’t lose their way at some point and take it on through the accumulation of bad IT decisions over time. But if you want to see what technical debt is costing retailers, just look at what these upstarts are capable of. And I say that knowing that some people are right now thinking “Yeah, it’s easy for them to be innovative and fast and responsive to swiftly evolving consumers – they got to start from scratch.” My answer to that is simple: Exactly. That is exactly the price of technical debt.
I’ve read the arguments that companies like Toys R Us or Claire’s or the dozens that came before were loaded up with debt and so over-leveraged they could not keep up with changing demand – that it was a finance thing that did them in. But where would they have spent the money if they had it? On technology. The retail apocalypse victims were all facing a load of technical debt that was so high, it has been more than these companies can muster to overcome it. Their hearts gave out.
And the pressure is not going to stop. Former founders and executives from some of these acquired upstart firms are already out in the marketplace looking to fund or build the next round of disruptors – Natalie Massenet of Net-A-Porter has a $75M VC fund looking for the next generation of disruptors, and she’s but one example.
The Bottom Line
The maturity date on retailers’ technical debt has come due. As many have already said, there is no “retail apocalypse” in the sense that the value retail as an industry provides is over. But we have seen a Darwinian wave of survival of the fittest. The financial crisis and subsequent Great Recession have shaken out the most vulnerable retailers – the CompUSAs and Linens N Things. Those were retailers who had no brand value of their own to speak of. Of the retailers who survived that shake-down, the ones who have no money to invest in their business came next. Toys R Us and Claire’s are near the end of that wave.
My prediction: the next wave will take out the retailers who cannot overcome their technical debt, not because of finance issues, but because of cultural ones – because CEOs and CIOs can’t convince the rest of the business that technical debt is a disease, and it’s killing them. Failed projects, revolving door CIOs, lawsuits against system integrators or software vendors – those will be the symptoms. And it’s only after that wave is cleared out that a retail renaissance will truly begin.