Principal Consultant, Kugler Maag Cie North America Inc.
Therein, one might expect that the macroeconomics of a large demand would inspire an associated large supply. However, this has not been the case. “Over the past ten to fifteen years, we have actually seen a drop in talented architects,” says Peter Abowd, CEO of Kugler Maag Cie North America. “For many of our assessments – both at manufacturers and suppliers — Systems and Software Architecture have been sore points. There’s generally a shortage of understanding of the purpose of architecture in product development as well as engineers that do it effectively.”
Maybe that’s because in the Age of Misinformation many executives don’t understand the Return on Investment (ROI) for software architects or, more accurately, the Opportunity Cost of a missing one. In 2018, an article in Median wrongly suggested that “In Agile, the development team needs to own the architecture collectively.” In 2019, an Uber developer titled his blog “Software Architecture is Overrated; Clear and Simple Design is Underrated” and then described a fantastic process for collaborative architecture creation. In a 2010 article entitled “What is the Value of a Software Architect”, the author suggests it should be a temporary position where the architect should sell his ideas to developers. So an ignorant executive doing quick research might find the Internet less than affirming.
In the end, there appears to be four reasons for the dearth of talented software architects which we shall delve into below: missing training, confusing accounting, dysfunctional evaluations and unmeasured velocities.
Anyone familiar with a search engine can find 1,001 different software architecture courses with several of them labelled “The #1 Architecture Course” (*NOTE: This makes me wonder if some international judge evaluated each course and declared a tie). And, yes, for some aspects of automotive, this would be sufficient. However, embedded, automotive architecture has its own quirks, and frequently these classes don’t provide sufficient understanding.
For starters, the partitioning of embedded functional safety and cybersecurity are very specific to industries, so generic classes are less helpful. For example, if a vehicle intends to have a lot of connectivity and downloads to a non-safety portion of the vehicle or module (e.g. “infotainment”), then the architecture should consider isolation (e.g. “gateway”) from the safety-related portions (e.g. anti-lock brakes). Such architectures should essentially communicate to the team the interfaces that provide valuable information between systems (e.g. wheel speed from the anti-lock brakes to the infotainment system for speed-dependent volume), while protecting against hackers and critical failures. Therein, the training must be more specialized, which means a lower supply.
Then comes the increased demand for training. Due to cost-competitive pressures and exponential growth in software development needs, developers have been increasingly hired in the past 15-20 years from lower-cost regions. In 2006, The Association of Computing Machinery published a 288 page study entitled “Globalization and Offshoring of Software” and correctly predicted that “Globalization of, and offshoring within, the software industry will continue and, in fact, increase … [with] the future, however, [being] one in which the individual will be situated in a more global competition. The brightness of the future for individuals, companies, or countries is centered on their ability to invest in building the foundations.” However, that foundational development required investment for strategic areas akin to architecture, which hasn’t happened.
The 1998 romantic comedy “Sliding Doors” starred Gwyneth Paltrow and showed two parallel realities for the same woman: one who successfully caught the life-altering subway after being fired, and one who missed the fateful subway, thereby living out an entirely alternate reality. In a similar vein, the 2002 Tom Cruise thriller, Minority Report, also weighed two possible paths. There was the likely future predicted by the soothsaying “precog’s” — who foresaw individual acts of potential violence – or the crimeless future where these upcoming felonies were thwarted by the Precrime Unit, an elite police force whose timely arrival prevented the impending doom.
Both stories are fiction. We rarely understand the costs and benefits of alternate realities.
That sad truth haunts value-add positions like architects. Accountants are unable to quantify the avoidance of future costs via efficiencies introduced today. For instance, what does it cost global corporations to have unclear, inefficient communications with major organizational integration issues because there hasn’t been a clear picture created by an architect? What strategic directions have been missed like platform extensibility, reuse of assets, and horizontal growth because the architecture has not been well-conceived or overseen? How poorly have change requests been quoted because of it is not possible to determine what existing architectural pieces could satisfy the change?
To justify the costs of hiring or training an architect, the beancounters need to understand the Opportunity Costs, which is incredibly confusing without Tom Cruise or Gwyneth Paltrow.
In one of two 2018 Harvard Business Review (HBR) studies, the researchers found the best performing salespersons were 15% more likely to be promoted and, therein, ended being the worst managers with a strong correlation to a 7.5% decline in overall performance of the team. Interestingly, another HBR study of 400,000 U.S. workers within top corporations measured fifty-eight (58) traits and found “Promotions go to those who best deserve them” was in the bottom three.
This same mistake is made with software architects.
“Frequently the technically-proficient, hard-working and extremely confident software programmers are promoted to architects,” says Abowd. “These traits don’t translate. Gathering the best ideas of the team, providing clear abstractions of the intended solution and shepherding the team’s collaboration with rigor are not typically part of the annual performance review of an individual contributor.”
And the study of 360-degree assessments from 70,000 leaders in 2018 supports this assertion. The “two key skills that [got] young individual contributors promoted to leadership positions” and then fail were technical/professional expertise in their old position and a “drive for individual results” as opposed to the team. Those mediocre leaders were overly focused on their technical prowess and work ethic, but ignored collaboration with the team.
“Some people are not are great fit for architects,” states Abowd. “Corporations should really spend the time to have personalities evaluated since an ineffective architect can undermine a successful product line strategy and cost the business millions in redundant development and needless rework.”
For difficult to find, train and retain resources, a very typical strategy is to label them “shared”. These gurus will be utilized across multiple projects so each, individual effort does not require 100% of their time.
The frequent issue, though, is a lack of understanding the true x% required. Ideally, organizations would estimate the overall effort, measure how quickly they finish that effort (a.k.a. “velocity”) and then plan future endeavors based upon capacity. This especially becomes complex when projects are simultaneous but need that guru at specific moments within the development lifecycle, thereby needing precise scheduling between moving targets.
Imagine the Detroit Lions winning the Superbowl. That’s about how frequently management understands the capactity of its architects and the coordinated project management of shared resources. Instead, assumptions are made “based upon what we’ve always done” or Optimism Bias without considering data that measured those choices and, thereafter, success.
Some executives will read this and think, “We have plenty of architects. That Tengler guy is talking to someone else.” However, a few key questions are worth pondering: