Running the roadmap on my own agency

Ok, first one.

The plan for Sparkfile has always been that I write the roadmap by living it, not by imagining it. So I pointed it at the most real thing I have: my own agency. If the roadmap is any good it should work on my business. If it isn’t, running it is how I find out.

Here’s what the first pass actually looked like, what held up, and the place it broke.

Phase 1: the memory files, plus two the roadmap never mentioned.

I wrote the four files the roadmap tells you to write: PRODUCT, USERS, GOALS, VOICE. That part worked. The most useful hour was USERS.md, because it made me say out loud what customers actually complain about before they find me. For a services business it’s the same shape every time: they keep getting asked a question their business can’t answer yet, or the thing they’ve got doesn’t do the one job they need it to.

But an agency isn’t a product. The four files weren’t enough. The center of a services business is a roster of clients and how you price the work, and the roadmap had no home for either one. So I added two files it never told me to write: a roster, one file per client, and a pricing file. For a services business those aren’t optional homework. They’re the equivalent of PRODUCT.md.

That was the first crack. It turned out to be the only crack, showing up three times.

The pricing shift.

Filling in the pricing file forced a decision I’d been dodging. The work had always been a one-off fee, which means you’re always hunting for the next client. A treadmill. So I wrote down a different model as the standard: a lower build fee plus a monthly care plan, instead of one big number and goodbye. Recommend the recurring plan, keep the flat fee as the fallback.

The real shift underneath it is how you think about value versus effort. A one-off fee prices the build and ignores everything that keeps the thing working after launch. The recurring model prices the relationship, which is where most of the actual value and most of the ongoing effort both live. Writing that down as the standard offer, not just a thing I did once, is the whole point.

The workflows.

The roadmap says your first workflow should be a weekly review, so that’s the one I built first. It reads the memory files and the roster and gives me one straight read on the whole business once a week: what moved, what’s stuck, what’s owed, and the three things that actually matter next week. It held up. It’s the right first automation because it’s the one that keeps everything else honest.

The second workflow is where the agency shape reasserted itself. I had live prospects and kept writing the same kind of document by hand, in my own voice, using the new pricing. So I did the thing the roadmap is actually about: I turned the pattern into a reusable workflow, so the next one starts from a draft instead of a blank page. Research the prospect, open on the real pain, price from the offer file, flag every assumption, save the draft.

That second one is the tell. A product roadmap’s getting-customers week is cold email. An agency’s is proposals. Same slot in the plan, completely different motion, and the roadmap only had the product version. The weekly review was universal; the proposal generator was the piece the plan didn’t have a place for.

The big one: the roadmap assumed I was a startup.

Every crack above is the same crack. The roadmap was written for a founder building a product. Specs, users, features, an investor update. I’m not that. I’m an agency owner with clients, proposals, and recurring billing, and none of that had a place in the plan.

So I broadened it. The 90-day spine is the same for anybody: memory, workflows, judgment. What changed is the roadmap now runs in tracks. Product, services and agency, local business. It has the two files I had to invent, a client roster and a pricing/offer file, as real modules instead of things I bolted on. And the getting-customers week now includes the proposal motion, not just cold email, with the workflow I built as the worked example.

One small dumb thing for the record: I flip-flopped the newsletter name and day, standardized on the wrong one, and had to walk it back. It’s The Spark, Sundays. Writing that here so I don’t relitigate it in a month.

That’s the run so far. The test case is already paying rent. It’s writing the roadmap for me.