Complexity arrives one reasonable decision at a time. Every individual call can be defensible while the whole thing is still wrong.
Chapter 19
A Portfolio, Not a Business
My father could not build the Christmas present.
It was a Big Wheel, the plastic tricycle with the oversized front wheel, and it came in a box in pieces, and he had got as far as opening it and stopped. On Christmas Eve he came and got me, quietly, and took me out to the garage to build it with him.
I was small enough that I could not read the instructions. I looked at the picture on the box and laid the parts out in the order they had to go.
I do not think my mother knows to this day.
I have been building things from pictures since before I could read the instructions. Forty-something years later, that was becoming a problem.
At one point I could have sat you down and shown you dozens of domains I owned, and explained exactly why each one mattered.
It was not a hoard. Every property had a reason. This one owned an idea I did not want somebody else to own. This one was the natural home for a particular kind of reader. This one existed because a term was going to matter in two years and it was cheap now. If you had challenged any single one of them I would have had an answer, and the answer would have been decent.
That was the problem.
The reasoning was fine at the level of each decision and catastrophic at the level of the whole. I had not built a company. I had built a portfolio, and I had done it the way most founders do, which is one entirely defensible choice at a time over several years.
Here is what I eventually noticed, and I noticed it late.
I was organizing the company around things I had built. Websites are outputs. They are the residue of work, not the structure of a business, and when you arrange a company around its outputs you end up with a company whose shape is a record of your own history rather than a description of what it does for anyone.
The question I had been asking was where does this domain fit. It is a reasonable question and it has an answer for everything, which is why it is dangerous. Any asset can be made to fit somewhere if you are motivated.
The question that fixed it was which system does this strengthen.
That one has a lot of no in it. Under that question most of what I owned did not have a job. It had a personality, which is different, and a personality is not something a customer can use.
It simplified fast after that, and the sentence I still use for it is that the websites finally had jobs instead of personalities. One educates. One explains the behavioral model. One assesses where a business actually stands. One takes responsibility for the outcome. One measures and improves. If a property could not answer which of those it was doing, it was not load-bearing, and it went quiet.
It was not painless. Some of those were ideas I liked.
Chapter 20
The CRM Was Not Software
For four years I was the internal point person for the entire HubSpot ecosystem at the company. Not the marketing dashboards. The whole thing, including being the only person in the building who did day to day technical support on it.
It was one of the most useful jobs I have ever had, and it taught me something I could not have learned from the outside.
Here is what I expected. I expected the work to be about the software. Fields, workflows, integrations, the parts you can read about. Some of it was.
What it turned out to be was a job about the company.
A CRM does not create a system. It exposes whether one exists. You can buy excellent software, install it properly, train everybody on it, and still end up with bad data, because the data was always bad and nothing before now required anybody to look at it in one place. You can have inconsistent behavior between two teams that each think they are doing it right, and never discover the inconsistency until both of them are typing into the same box.
Every problem I was asked to fix in that system was described to me as a software problem. Almost none of them were.
Handoffs that had never been agreed. Ownership that had never been assigned, so a record sat with nobody. Definitions that quietly meant two different things depending on who was using them. Adoption that looked complete and was not, because a couple of people had opted out privately and nobody wanted to be the one to say so.
The software did not cause any of that. It just made it visible, all at once, on a screen where the people who could fix it were finally looking.
The uncomfortable part is that being the person who can see it does not make you the person who can change it. I could show anybody the report. I could not make two departments agree on what a qualified lead was, and until they did the report was going to keep being wrong in the same way every month.
This is one of the roots of everything I later built for my own company, and it is why I do not describe what I sell as software. The real system lives between the software and the people. It is made of agreements, ownership, definitions and follow up, and the tool is where you can see whether any of that is actually happening.
When a company tells me their CRM is not working, I now ask what it is showing them. Usually it is working perfectly.
Chapter 21
They Changed the Route and Told Nobody
Three hours into driving two bus loads of students to a winter retreat in the Santa Cruz mountains, I looked out of the window and realized we were going the wrong way.
I asked the driver why we had changed course. Mudslide. The route we had planned was closed, and he had picked a different one.
He was right. It was the correct call, made by the person best placed to make it, using information the rest of us did not have.
It was also a failure, and I have thought about it for years.
Because the decision was made without telling anybody else, the band took the original route. So did the musicians. So did the luggage truck, which is the one that really mattered, since it had everybody's bags in it. Several separate groups of people spent the afternoon driving toward a road that was shut, and then spent the evening trying to work out where everyone else had gone.
Nothing about the driver's judgment was wrong. The cost landed entirely on the people who were not in the conversation.
There is a smaller and more uncomfortable half to this, which is that I was responsible for two buses full of other people's children and I found out about the change three hours in, by looking out of a window. Nobody hid it from me. It simply did not occur to anybody that I needed to know, and it did not occur to me to ask, because from where I sat everything was fine.
I have watched the same shape play out in businesses more times than I can count. Somebody makes a good decision quickly. The decision is correct. It gets communicated to the people it directly affects and to nobody else, and three weeks later a team is still building against a plan that was abandoned.
The instinct when this happens is to add process. More meetings, more sign off, more people in the loop, which slows down the exact fast decision making you wanted in the first place.
I do not think that is the answer. What I have come to think is that the obligation is not to consult everybody. It is to broadcast afterward, quickly, to a wider circle than feels necessary. The driver did not need my permission. He needed to say, on the radio, that the route had changed.
A correct decision that nobody hears about is still a failure, and it fails in a way that is particularly hard to spot, because everybody involved in making it remembers doing the right thing.
Chapter 22
When It Is Not Your Fault
Most fragmentation is not one person's collection. It accumulates across an organization, and it is much harder to see because everybody involved is doing their job correctly.
The website says one thing. The sales team, who talk to actual customers all day, have evolved a slightly better version and use that instead. The printed material is on last year's positioning because reprinting is expensive. The social account has a different personality because a different person runs it, and they are good at it. One location has professional signage and the one two hours away has a sign somebody made themselves, which is the most honest thing in the whole system.
Nobody did anything wrong. Every one of those choices had a reason a manager could defend.
But a customer does not experience your company one department at a time. They assemble it, from whatever fragments they happen to encounter, into a single impression, and a set of signals that do not agree with each other assembles into a business that is hard to be confident about.
Trust requires predictability. Predictability requires consistency. A fragmented brand is by definition unpredictable, which means it is asking every customer to take more on faith than a coherent competitor is asking. Most of them will not do it, and none of them will tell you that is why.
For a franchise system this gets worse rather than staying the same, because the inconsistency is not damaging one market, it is damaging the brand everywhere the inconsistency is visible. A customer who has a good experience in one city and a mediocre one in another does not conclude that one operator is weaker. They conclude that the brand does not mean anything reliable, and they start judging location by location, which is work, and which no brand should be making its customers do.
Why franchises get hit hardest
Franchise systems are unusually exposed to this, and the reason is arithmetic rather than culture.
A single-location business that is inconsistent damages trust in one market. A franchise that is inconsistent damages trust in the brand, in every market where a customer can see the inconsistency, which in practice is everywhere, because reviews and listings are not geographically bounded any more.
Then it becomes a loop. Inconsistent delivery produces weaker reviews. Weaker reviews make customers work harder to decide, so fewer of them decide. Harder acquisition means less money, and less money makes it harder to invest in the standards and training that produce consistency in the first place.
I have watched franchisors try to break that loop with marketing spend, and it does not break with marketing spend. It breaks with operational discipline, applied unevenly at first, defended for years, which is slow and boring and works.
The uncomfortable part for a franchisor is that the operators who most need the standards are usually the ones with the least margin to absorb them, and telling somebody with a thin month that their signage is costing the network is a conversation that goes badly. That is a real problem and I do not have a clean answer to it. What I know is that the systems which got through it are the ones where the standard was demonstrably making individual operators money, not just making the brand tidier.
Chapter 23
Coherence Costs Less Than Repair
There is a quieter kind of fragmentation that has almost nothing to do with design.
A business can have an excellent website and a Google listing that has not been touched in three years. Consistent social accounts and directory entries all over the web with three different phone numbers. A strong story on the homepage and reviews underneath it that describe a different company.
Search and AI systems read all of that at once. When the signals agree, they reinforce each other and you get placed confidently. When they disagree, you get averaged into something vague, or skipped in favor of a competitor who is easier to be sure about.
In an environment where a machine is doing the first round of recommending, consistency stopped being an aesthetic standard and became a discoverability requirement.
Only one of these costs shows up as a line item
The investment required to unify a brand is real and nobody should pretend otherwise. It needs standards, training, enforcement, and somebody whose job it is to notice drift. In a franchise system it needs alignment across independent operators who have their own priorities and their own money at risk, which is a negotiation rather than an instruction.
It is not easy. I watched it take four years.
But the alternative is not free either, it is just invisible. Fragmentation erodes trust slowly and quietly, and rebuilding trust that has already eroded takes years, while maintaining consistency takes discipline. Those are both costs. Only one of them shows up as a line item, which is exactly why the wrong one usually gets chosen.
The most consistent brands in any category are rarely the most creative or the best funded. They are the ones that said the same true thing, the same way, in the same voice, for long enough that the market stopped being able to imagine the category without them.
Chapter 24
Running It From Bed
The business kept running. The operator could not sit up.
For a stretch of years I could not sit upright, and I ran the company from my bed.
Don Kick built me a wooden stand that held a laptop above my face while I lay flat. Wireless keyboard on my lap.
He is the third man in the water in the shark chapter. In 1987 I was the one trying to save him from a pod of dolphins. Thirty-something years later he built the thing that let me keep working. That is the whole picture and I do not need to give you more of the medical detail than that: multiple surgeries, a fusion, and a back that has been in pain every day since a truck hit me.
In the next room the office was still an office. Servers in the closet, a backup connection to a building downtown, a generator, and four employees who came in and did their jobs.
There is an inspirational version of this chapter and it is useless, so I am not going to write it.
This is not a story about overcoming anything. I did not overcome it. I am still in pain, and I was in pain the entire time.
It is a story about what a constraint does to a system.
Everything that could be made not to depend on me physically had been, one problem at a time, over months. Not because I planned it. Because I could not stand up, and each thing that broke had to be solved in a way that did not require me to stand up.
That is a strange way to discover something I now believe: a business that depends on one person being physically available has not built enough system, and most owners never find out which parts those are, because nothing forces the test.
I got the test whether I wanted it or not.
And what I learned from it is not really about redundancy or infrastructure. It is that the pieces which survived were the ones where somebody else knew what to do next, and the pieces which stopped were the ones where the knowledge was in my head and only in my head.
That is the same problem as the next chapter, arrived at from a different direction.
Chapter 25
The Company Had No Memory
For a long time I thought what I was missing was visibility into all of this. More dashboards. Better reporting. Something that would show me the state of everything at once.
That was wrong, and working out why took me a great deal of building.
The pattern that kept happening was this. Something would break. Somebody would notice, eventually. It would get fixed, usually well. And then a few weeks later something structurally similar would happen somewhere else in the business, and we would work it out again from scratch, as though it were new.
I did not have a reporting problem. The company had no memory.
A dashboard tells you the current state of things. It does not remember that this exact thing happened in March, what we decided, whether the decision worked, and what we said we would do differently. If the only place that knowledge lives is in the founder's head, then the business has not learned anything. It has just been near somebody who learned it, and that is not the same thing at all, especially on the days that person is unavailable or tired or wrong.
So the thing I ended up building was not automation. Automation without memory just repeats the same mistake more efficiently. What was actually needed was for the company to notice, remember, and be responsible for the follow-up: a failed process raises an alert, a customer struggling early gets help before they give up, a leak in the path to purchase becomes work assigned to somebody, ground we were losing gets flagged while it is still recoverable, and a question researched once is available the next time it comes up instead of being researched again.
The test I use now is a single question. How will the company notice this next time without me.
If the honest answer is that it will not, then I have not built a system. I have just been paying attention, and attention is not transferable and does not scale.
