On the seven-engagement cap.
We never run more than seven concurrent engagements. It is the single hardest rule we operate, and most of what we do well comes from it. Here is why the number stays.
Every agency we've ever worked alongside grew by saying yes. It is the natural physics of the business. A good month becomes a hiring round becomes a bigger office becomes a sales target becomes a head of growth becomes a roadshow becomes, eventually, an account on your books that none of your senior people can pick out of a lineup. We say no, instead. The cap stays at seven.
01The rule.
No more than seven concurrent engagements. Reviewed quarterly. No exceptions for short projects, no exceptions for a brand we'd love to have on the homepage, no exceptions for a discount. If we are at seven and another good fit comes in, we put them on a waitlist and tell them the truth: we will be free in approximately N weeks. Some wait. Most do not. Both outcomes are fine.
02Why seven.
The number is not arbitrary, but it is also not magic. It is the upper bound at which the team we have can hold three things simultaneously.
Context. Every engineer in the room can name every client, what they sell, what runs on their stack, what's currently shipping, and where the last incident came from. Past seven, that ledger starts to slip — people begin specialising into accounts rather than into disciplines, and "I'll have to check with whoever's on that one" creeps into conversations. Once it creeps in, it never leaves.
On-call. A meaningful incident-response commitment requires a primary, a secondary, and a fallback, with overlapping fluency across every account on the contract. At seven clients on a team of seven engineers, the math works. At twelve, it does not, and the work to make it look like it does becomes its own job.
Quality of refusal. The single most underrated skill in this category is the ability to say "no, you don't need that yet" to a paying client. When you have too many accounts, you stop saying it. The client always wins the argument, because saying yes is the path of least resistance, and the cost lands a quarter later. A small book of clients is what makes that "no" survivable.
03The math.
The economics most agencies operate look approximately like this: more clients, lower per-client price, hire to absorb the load, repeat. The unit of growth is a head. The hidden cost is that the senior engineer on the bid is rarely the engineer on the work. By contract three, you are an account manager.
The economics we run look approximately like this: fewer clients, materially higher per-client price, no hire unless the existing team can describe what the new person will do for at least three of the seven accounts. The unit of growth is depth, not breadth. The senior engineer on the bid is the engineer on the work.
The first model scales further. The second model produces work the first model cannot produce at any price. We are honest about which kind of work this is, and we charge accordingly.
Bandwidth is a feature. It is the thing we are actually selling. The disciplines underneath are interchangeable; the attention is not.
04The eighth client.
People want to know what happens to them.
We tell them three things, in order. First, we are not the right team if their work needs to start this month — we will not free up that quickly, and a project rushed into a full plate is the worst version of us. Second, the waitlist is real; the order is preserved; we will give them an honest window. Third, we will introduce them to two or three teams who we trust to do the work to a standard close enough to ours that we will not feel embarrassed by the referral.
About a third of waitlisted prospects wait. The rest go elsewhere or build something internally. Both outcomes are fine. We are not trying to be the only option. We are trying to be the right option for a specific kind of buyer.
05What this gets us.
One thing, primarily: the ability to operate every client's stack like an internal team, rather than as a vendor in someone else's process. Internal teams have the luxury of context — they know the history, they know the constraints, they know which decisions were made on a Tuesday at 3am and shouldn't be relitigated. We get to that posture inside three months on every engagement, because the engineers carrying the work were the engineers on the kickoff call, and there were not eleven of those calls that quarter.
The work feels different to a client who has worked with both kinds of team. We do not need to explain the difference. They tell us.
The cap stays at seven. The number itself may change someday — if we hire a particular kind of senior engineer, eight is conceivable. The principle does not. There is a number above which the work we do becomes the work everyone else does, and we will not cross it for money.