Everything that went wrong at my agency was a scope problem

I started Let'sGo in March 2022, when I was fourteen. It is now six or seven people serving UK and international clients, and between the agency and the freelance work before it I have shipped somewhere close to a hundred websites.

Enough of them went badly that a pattern is visible, and the pattern is embarrassingly consistent. Not one of the bad projects failed because the work was too hard. Every single one failed because I agreed to something I had not defined.

What it looks like from the inside

It never announces itself. It sounds like agreement.

The client says they want the site to feel modern. You say yes, because you know what modern means. Six weeks later they mean something else, and you are on revision nine, and neither of you is wrong. Nobody lied. You simply both signed a sentence that had no fixed meaning, and then discovered that at the most expensive possible moment.

Or: the scope says five pages. Halfway through, they ask whether the team page could pull from a database so they can update it themselves. It is a small request. You say yes, because it is small and you want them to like working with you. That yes was a CMS, an admin interface, an auth decision and a hosting change, and none of it was priced.

Or the worst version, which is entirely mine: a client asks for something outside what we do, and I say yes because I am fifteen and turning down money feels insane. Now I am managing a subcontractor on work I cannot evaluate, for a client who thinks it is my work, and it will be my reputation when it is late.

The lesson I actually learned

For a long time I thought the fix was better contracts. It is not, though better contracts help. A contract is a description of an agreement, and the problem is upstream: the agreement itself was never specific enough to describe.

The fix is that defining the work is the work.

Before the estimate, before the timeline, before anything, there is a job to be done that consists of turning a client's sentence into something with edges. It is unglamorous, unbillable-feeling, and it is the highest-leverage hour in the project. I used to treat it as the friction before the real work started. It is the real work. Everything after it is execution.

Three rules that came out of it

Anything qualitative gets an artefact. "Modern" is not a specification. Three reference sites the client has picked, with a sentence each on what they like about it, is a specification. If the word cannot be pointed at, it does not belong in an agreement.

Small requests get priced at zero, not skipped. The instinct is to absorb a small ask to keep goodwill. What that actually does is teach the client that scope is free, and by the twentieth small ask the relationship is worse than if you had charged for the first. Now I quote everything, and I sometimes quote it at zero. The number can be nothing. The line item cannot be missing.

Say no to work outside what the team does. This one cost me the most to learn, because early on every no felt like it might be the last client. It is not. Subcontracting work I could not personally evaluate has never once gone well, and each time it went wrong it damaged a relationship worth more than the project.

Why the hybrid model exists

People ask why Let'sGo keeps some work in-house and outsources the rest, as though it were an efficiency decision. It is a scope decision.

Strategy, client ownership and the core build stay in-house, because those are the things where a fuzzy agreement is fatal and I need someone who can push back in the room. Specialist one-offs - an unusual video edit, a niche integration - go out, because those have naturally hard edges. A video is a video. It is deliverable, reviewable and unambiguous in a way that "make the brand feel more premium" is not.

A small agency that hires for every discipline it sells becomes a payroll problem in the first slow quarter. One that outsources everything cannot hold quality or a relationship. The split is not about cost. It is about which parts of the work you can afford to have someone else define.

The part nobody tells you at fourteen

Running this at fourteen and fifteen taught me scope discipline faster than anything else has, and the reason is that I had no cushion. An agency with runway can absorb a badly scoped project. I could not. A month of unpaid revisions was a month of not paying anyone, and consequences that immediate teach quickly.

I would not recommend the pressure. I would recommend the lesson, which generalises well past agencies: most of the failures I have watched in other people's companies were not execution failures either. They were somebody agreeing to something nobody had defined, and finding out in month six.