Pod Insights The Agile Daily Standup - AgileDad The Scrum That Actually Worked About this episode The Scrum That Actually Worked
In 1996, Chrysler — a Fortune 500 company with resources to hire the best talent and buy the best tools — had spent two years and millions of dollars building payroll software.
It hadn’t printed a single paycheck.
The project was called C3: Chrysler Comprehensive Compensation. It was supposed to unify payroll for 87,000 employees across multiple divisions. It had executive sponsorship from CIO Susan Unger. It used Smalltalk, an object-oriented programming language that promised to solve exactly the kind of tangled legacy problems Chrysler faced.
How to connect with AgileDad:
- [website] https://www.agiledad.com/ https://www.agiledad.com/
- [instagram] https://www.instagram.com/agile_coach/ https://www.instagram.com/agile_coach/
- [facebook] https://www.facebook.com/RealAgileDad/ https://www.facebook.com/RealAgileDad/
- [Linkedin] https://www.linkedin.com/in/leehenson/ https://www.linkedin.com/in/leehenson/
Transcript 0:16 Speaker 00 Someone asked me to record an episode about an article. So I'm going to read the article now. They asked me to do a blind read. So this will be my first time reading through the article. It's called "The Scrum That Actually Worked." So this should be interesting. Let's get started. In 1996, Chrysler, a Fortune 500 company with resources, hired the best talent and bought the best tools, had spent two years and a million dollars, or millions of dollars, building payroll software. It hadn't printed a single paycheck. The project was called C3: Chrysler Comprehensive Compensation. It was supposed to unify payroll for 87,000 employees across multiple divisions. It had executive sponsorship from CIO Susan Unger.
0:59 Speaker 00 It used Smalltalk, an object-oriented programming language, to promise to solve exactly the kind of tangled legacy problems Chrysler faced. None of that mattered. When consultants arrived in March 1996, no one could answer the most basic question: What day will we be able to print checks? When will it be? What happened next would reshape the software industry, just not in the way anyone expected. Here we go. Once again, we're focusing on software, but I want to emphasize this is the season where agile's not all about software delivery. Let's continue. Three years earlier, Chrysler's Director of Payroll Systems, Tom Hadfield, stood before CIO Susan Unger and demoed proof of concept for a new payroll system.
1:43 Speaker 00 Chrysler already had active payroll software. That wasn't the issue. The problem was that nobody could safely change it. Every pay cycle was a firefight. A tax rule would change. A union contract would add an exception. Engineers would dig through thousands of lines of tangled code, trying to find a problem. Fix one calculation, break another 3,000 lines away.
2:05 Speaker 00 The system had been patched so many times over so many years that no one fully understood how it worked anymore. The culprit was COBOL, the programming language that ran most corporate systems of that era. COBOL was procedural, one long sequence of instructions executed in order, top to bottom. Think of a recipe where every step depends on the step before it. Fine for simple tasks, catastrophic when a recipe is decades old, thousands of pages long, has been modified by hundreds of chefs who each added their own workarounds or secret ingredients. Hadfield's demo used small talk, an object-oriented language. Instead of one long recipe, you create discrete containers, or objects, each responsible for one thing:
2:50 Speaker 00 tax calculations within a tax object, union rules of the union object. Each object knows how to do its job and nothing else. When something breaks, you know exactly where to look. Fix it in one place, the rest of the system doesn't care. For Chrysler paying eighty-seven thousand employees across multiple divisions, each with its legacy roles and special cases, this was the difference between a system that could absorb change and when it collapsed under the pressure of change. Unger saw the potential. She greenlit the project. Then Chrysler hired a vendor to build Hadfield's vision. The vendor decided they didn't need Hadfield's input. He left the team and headed for San Francisco. Over the next two years, the project spiraled uncontrollably. Requirements expanded, deadlines
3:33 Speaker 00 slipped, the vendor's team produced code but not software. Nothing ever actually worked. By early 1996, the project was drowning, and Unger needed answers. Hadfield, still consulting from a distance, made one final contribution. He recommended a consultant named Kent Beck, a prominent figure in the small talk world, who had spent a decade publishing on software practices. Beck arrived in March 1996, hired to tune the system's performance. He looked at the code, he looked at the team. Within days, he had a very different recommendation.
4:08 Speaker 00 What nobody mentioned was C3's failure was typical. The Standish Group's Chaos Report, published in 1994, I quote that one a lot. It quantified the catastrophe across the software industry. Thirty-one percent of projects canceled outright. Another 52% delivered late, over budget, or with fewer failures than specified. Only 16.2 succeeded. That's roughly 1/6. I always use 16.9. I didn't realize it was only 16.2. The dominant methodology was called waterfall. The promise. We will tell you what you're getting and when you're getting it. The plan was a contract. Management job was to monitor execution against the contract, catch deviations, and hold people accountable.
4:51 Speaker 00 The problem was that this promise was almost always a lie. It was just a straight-up lie. It wasn't deliberate. It wasn't deliberate at all. It was just a structural promise. Frederick Books had diagnosed it two decades earlier. The waterfall model assumes one builds a whole system all at once, combining all the pieces end to end, and system-tested after all the implementation, design, and everything is done. Most of the coding, much of the component testing, after everything's done. By then, it's often too late. The team had already built the wrong thing. C3 followed the waterfall playbook faithfully. Two years of requirements, design, and development, and nothing to show for it. Beck saw what any reasonable consultant should have seen ages ago.
5:35 Speaker 00 A project that needed to be shut down entirely before any more money was wasted. Instead, he proposed something radical. "I had a mental image of knobs on a control board," Beck later wrote. "Each knob was a practice that, from experience, I knew worked well. I would turn all the knobs up to 10 and see what happened. If code reviews were good, review code all the time. Two programmers at one keyboard. Testing is good, write tests before you write the code. If integration is important, integrate continuously, multiple times a day. If short iterations are good, make them a week, not months." Beck walked this proposed luck to Susan Under. Ron Jeffries, who was present, later recounted the exchange. Beck explained how they'd work in iterations of a few weeks, building whatever the customer asked for.
6:20 Speaker 00 Under asked the obvious question: "How will I know whether you're on track or not?" Beck held up a stack of story cards. "Here are all the things we have to do. We'll be adding things as we discover them and removing any that we don't need anymore to be done." We ask you to visit us once a month and we'll show you these cards and show you how many are done, how many are left to do. If you're not satisfied to progress, you can cancel the project. Unger's promise, I can do that? Absolutely. Beck spent two weeks at a whiteboard designing what he called extreme programming. The name was deliberate. Each practice was a known good idea cranked to the extreme. Together, they formed mutually reinforcing systems. Four practices defined a core. Pair programming where two programmers, one keyboard, not tutoring, not watching,
7:05 Speaker 00 a continuous dialogue about how to solve a problem. When Martin Fowler was asked whether this reduced productivity, he answered that would be true if it is the most consuming, uh, most time consuming part of programming was typing. Test-driven development. Write the test before you write the code. The test define what done means. No ambiguity, no drift. No, I hope I think it works. Uh, onsite customer. A real user embedded with the development team. Uh, available to answer questions, clarify requirements and make decisions, not a proxy, not a business analyst, actual person who's going to be using the software. Sustainable pace. Tired programmers make bad choices. Overtime, overtime is a sign of planning failure, not dedication.
7:50 Speaker 00 The methodology explicitly called for 40-hour weeks, a radical stance in an industry that's celebrated to having all-nighters. The remaining practices, continuous integration, small releases, planning game, collective code ownership, refactoring, simple design, coding standards, and a system metaphor supported these four. Remove one and the others became harder to sustain. Continuous integration enabled small releases. The planning game enabled sustainable pace. Collective code ownership required coding standards. None of these ideas were new. Pair programming had been independently invented multiple times. Larry Constantine documented it at PJ Plogers Group in the 1970s, calling it dynamic duo development.
8:35 Speaker 00 And that was the thing. The first had advocates, iterative development, and had been done for decades. Beck's innovation was putting them together in a package where each practice reinforced to others. Back, uh, back to the story here, Beck recruited Ron Jeffries to help coach the team. They started over from scratch. Two entire years of work, code and documentation were tossed straight into the bin, never to be seen again. Eleven months later, C3 went live. It started paying 10,000 employees. The team had roughly the same number of test assertions as they did lines of code. Defect rates were, in Jeffrey's words, "very low." The difference showed up in small moments. A change request would come in, a new tax rule, an edge case, and nobody flinched.
9:18 Speaker 00 Write a test, change the object, run the suite, deploy. What used to take weeks now took hours. The team had stopped being afraid of their own code. This was not supposed to be possible. A project that had failed for two years was not supposed to ship in 11 months. A methodology designed on a whiteboard was not supposed to work the first time it was tried. The industry's 1 in 6 success rate was not supposed to be beaten by a team that had already failed before. Word spread. Beck, Jeffries and their collaborators, Ward Cunningham, Martin Fowler and others, began writing and speaking about what they'd done. They called it extreme programming, or XP for short.
9:56 Speaker 00 To forgotten success, what most people don't know is Daimler Chrysler would eventually cancel further C3 development in February 2000. Critics have used this to dismiss XP as a failure, but the critics were wrong. Chet Hendrickson, a C3 team member, later provided context. Daimler Chrysler still wants to replace all their payroll systems with a single solution and it gearing up to spend about four times C3's total cost to do it. This will be Daimler Chrysler's fifth attempt to do this. C3 is the only one to ever go live. Five attempts, only XP shipped. The system kept paying employees for four years after the development stopped.
10:37 Speaker 00 But there were real problems worth examining.
10:41 Speaker 00 Marie de Arminant, the payroll supervisor who served on C3's onsite customer, had been extraordinary, writing user stories, assigning priorities, approving acceptance tests. The team took her for granted. "We don't know how Marie learned to do all this," Hendrickson later wrote. "Marie didn't know how she knew it. Eventually, it became too much. Marie developed a facial tic from the stress and transferred to a less demanding role. Her replacement was capable, but the team had never learned how Marie did what she did. All we did know," Hendrickson admitted, "was that the job had been too much for only one person we had ever seen actually do that job. The customer role as design might be
11:22 Speaker 00 unsustainable." Kent Beck identified another fracture. The gold owner and gold donor weren't the same. Let me say that again. The gold owner and the goal donor weren't the same. The embedded customer wanted incremental improvements. IT management wanted to eliminate mainframe entirely. If the people who tell you what to do don't agree with the people who evaluate what you're doing, you're stuffed. You're done. And these were real vulnerabilities. But still, they didn't kill us. It wasn't until the manifesto came through in 2001 to help clarify some things and to help gather all these lightweight things together to help align and figure out what we're doing. Then in 2002, Ken Schwaber and Mike Combe
12:06 Speaker 00 founded Scrum Alliance and put together a program for certified Scrum Masters and certified product owners. The scoop is, this is a history lesson over time to tell you that it all started with one idea: improve patterns, improve systems, improve software. And if we can get to the point where we can acknowledge that it's not one right and all wrong, and we can learn about the Agile landscape, that's when we're going to be better at creating better products and services over time. I encourage you to visit the Agile website and download the latest February, or the latest 2026 Agile Landscape Solution. It's absolutely awesome and you'll love it. Until next time, I encourage you to stay healthy, stay well, and stay agile, my friends. Do take care. Bye now.
Previous The Concept of Scrum Fluid Teams Next What Curling Can Teach Us About Agile - Mike Cohn Page updated November 6, 2026 · transcript and summary are machine-generated