What the stack costs

h23 — and what each part has to return to earn its place · built 2026-08-13

“If we add lots of costs the system price goes up and ROI gets harder to sell. A $100k system is a hard sell however good it is.” — Tim, 10 August

$39/moActually spent today on marketing tooling, across the whole network. One tool is switched on.

$410/moWhat the rest would add if every deferred tool were turned on at once. It is not going to be.

Read the “dry” column literally. A tool marked dry has no credential on the machine, which means it cannot run and cannot bill. This table is generated by checking for the key, not from memory — which is how I caught myself telling Tim on the 13 August call that the system finds email addresses. The enrichment adapters are written. They are all dry. Today the pipeline reaches a LinkedIn profile and stops.

Marketing tooling — the decisions

Tool	$/mo	State	Verdict
Apify Scrapers for LinkedIn engagement and Google Maps. One key, many actors.	$39	on	Earns it
Must return: It is the only source of audience and of local-trade knowledge bases. Nothing else in the stack replaces it, and the Maps path is what fixed the wrong-industry bug.
GitLeads First pass of email enrichment — the cheap one.	$49	dry	Not decided
Must return: Cody’s worked example finds ~64% of emails here. If it does not clear half, the waterfall below it gets expensive fast.
Apollo Second pass for whatever GitLeads misses.	$49	dry	Not decided
Must return: Must convert enough of the remainder to beat paying the expensive tool for all of them. Only worth it once volume is real.
Origami Aggregates the whole waterfall behind one call.	metered	dry	Preferred
Must return: Replaces the two above rather than adding to them. Priced per lookup, so it costs nothing until used — which makes it the right first paid enrichment, not the third.
LeadMagic Mobile numbers.	$99	dry	Defer
Must return: Only if someone is actually going to phone. We are not, yet.
Million Verifier Grades an address before you send to it.	$37	dry	Buy with enrichment
Must return: Cheap insurance on a thing that is unrecoverable when it goes wrong: send to dead addresses and the domain stops reaching anyone. Buy this the same day we buy any enrichment, never later.
Instantly Cold-email sending infrastructure.	$97	dry	Defer
Must return: Plus ~$100/mo for burner domains and inboxes. ~$200 all-in for ~10,000 sends. Nothing until we have a list worth sending to and something worth saying.
HeyReach LinkedIn DMs at volume, via API.	$79	dry	Defer
Must return: Second channel, and only after email has a measured cost per lead to beat.
Ordinal Multi-account social scheduling with analytics feeding back.	metered	dry	Defer
Must return: Agent #2 needs it to close the loop between what was posted and what worked. Not until agent #2 is producing publishable drafts.
Model API Judgment where judgment is genuinely needed.	metered	dry	Buy when unblocked
Must return: Metered, not subscription. Needed for ICP stage 2, campaign copy and reply classification — three places, none of them per-lead. The engine and the whole campaign layer run with no model calls at all, deliberately.

Infrastructure — already paid, shared across ~190 sites

Item	$/mo	What it carries
Droplet	$48	Runs the engine, all ~190 sites, the knowledge bases and the pipeline. Already paid, shared across everything.
Postiz droplet	$12	Separate box for publishing. Already paid.

$60 a month carries the engine, every site, the knowledge bases and the pipeline. It does not scale with clients, which is the whole reason the unit economics work.

The answer to Tim’s question

1. The engine itself costs nothing per client to run

No model calls in the routine path — not in the audit, not in topic selection, not in the campaign layer, not in the creative. That is a deliberate design decision and it is the single largest cost control we have. A per-action model call would put a variable cost under every client and make the $140 tier impossible.

2. Buy in one order, and only when the previous one is proven

Origami first, because it is priced per lookup rather than per month — it costs nothing until a lead is worth enriching, and it replaces the two-subscription waterfall rather than adding to it. Million Verifier with it, same day, because sending to dead addresses is the one mistake in this stack that cannot be undone. Everything else waits for a measured cost per lead.

3. The ceiling is the price, not the ambition

At $140 / $280 / $420 a month, a client cannot carry a $200/month tool by themselves. Every paid tool has to be shared across the whole book or it does not go in. That is the actual constraint, and it is why the list above is mostly deferred rather than mostly bought.

It also answers the $100k worry directly: there is no version of this that becomes a $100k system, because there is no component that scales with client count except media spend — and media spend is the client’s, capped by them, and never proposed until the free channels are done.