The Marketing Tech Stack I Actually Run in 2026

Google “marketing tech stack” and count the logos. The vendor guides that rank stack dozens of tools into one diagram. Useful as wall art. Useless as a shopping list.

Now look at the most useful result on that page. It is a Reddit thread. Real marketers asking each other what they actually use, because the diagrams never answer the real question.

Here is my answer. My marketing tech stack runs my own sites, my consulting work, and a live Google Ads account on a budget of a few francs a day.

It has six layers. Three carry a bill. The rest are free software doing the heavy lifting.

This is the whole thing. No affiliate links. No sponsored picks. Just what runs, what it costs, and what I deleted.

The stack owner holding an espresso, calmly running a compact six-module console while a colossal tower of grey boxes topples behind him

Key Takeaways

A stack is what runs without you. If a system needs you every week, it is a task wearing a subscription badge.
Data before tools. Four API pipes and one database beat any dashboard license.
Boring infrastructure wins on price. My whole setup costs less per month than one seat of most marketing SaaS.
AI belongs in the maintenance layer, not the strategy layer. It writes the scripts. It never sets the direction.
Every tool pays rent. If you cannot say what it earned or saved last quarter, cut it.

What a marketing tech stack actually is

A marketing tech stack is the set of systems that collect, move, and act on your marketing data without you touching them.

Note the last part. Without you touching them. I organize my whole operation into workflows, agents, and tools. The stack is the plumbing underneath those workflows.

Most teams define their stack by what they pay for. I define mine by what runs on its own. That one filter cuts the list from 40 logos to six layers.

Infographic of the six-layer marketing stack: one server git push deploy, one database four API pipes, measurement loops, static sites, the AI layer, self-hosted plumbing

There is a second filter that matters even more. Ownership. Every system I depend on can export its data to a place I control.

A stack you cannot leave is not a stack. It is a lease with extra steps.

Layer 1: One server and a git push

Everything I own online runs on one small server in a German datacenter. My personal site, my consulting site, the booking system, the invoicing tool, even the git server that deploys them all.

I publish with a git push. The server pulls the change, rebuilds the sites, and swaps the containers. A chat channel gets a notification when the deploy lands, green or red.

If a build fails, the old site keeps serving. Rollback is a git revert. That is more resilience than most managed platforms sell you, and it costs nothing extra.

He presses the single lime button on a small server cabinet while an identical standby unit waits quietly beside it: git push deploys, everything else is redundancy

Cloudflare sits in front of everything. When I test cached pages, they answer from the edge in about 60 milliseconds. My monthly infrastructure bill would not cover a business lunch in Zurich.

Layer 2: The data layer comes first

The heart of the stack is a PostgreSQL database on a home server. Around it, four pipes: the Google Analytics API, the Search Console API, the Google Ads API, and a SEO data provider for keyword volumes and SERP data.

He cradles a glowing lime crystal core hovering above a small home server while four dark cables flow into it from four directions: one database, four API pipes

Every number I report, for myself or for clients, flows through those pipes into that database. Keyword positions over time. Ad performance. Site metrics.

When someone asks where a number came from, I answer in seconds.

That single ability is worth more than any dashboard software. If you cannot answer “where does this number come from” in ten seconds, you do not have a stack yet. You have bookmarks.

I wrote about this in how AI marketing implementations fail: most teams automate production before they automate seeing. The data layer is how you flip that order.

Layer 3: The measurement loops

These are scripts, not platforms. One checks keyword positions and flags what moved. One compares ad spend against conversions every morning. One drops a weekly snapshot into my notes before my first coffee.

They cost cents per run. The keyword research behind this article, volumes and difficulty scores included, cost about one franc of API calls.

The loop that pays for itself fastest is in the ads account. A script pulls the search terms report through the API, flags the irrelevant queries, and queues them as negative keywords. I review, I approve, it executes. The machine does the reading.

First weeks on one small account: 46 negative keywords added, entire junk categories switched off. That account now runs at CHF 6 a day and produces qualified conversations at about CHF 13 each. No agency retainer anywhere in sight.

He sips espresso while a small lime robot sorts cubes on a miniature conveyor belt: grey junk cubes slide into a bin, the few glowing ones stay in a neat tray

The loop also told us the next problem. Quality Score data showed the landing pages needed work, not the ads. The system pointed where to look. That is the entire point of measurement.

Layer 4: The delivery layer

My sites are static, built with Astro, served as plain files. Fast because there is nothing to execute at request time. Secure because there is almost nothing to exploit. Google measures how fast they load, and static files make that conversation easy.

He slides a thin dark slab with a hairline lime edge onto a neat stack of identical slabs: static files, nothing to execute, nothing to exploit

Multilingual by design, too. My consulting site speaks four languages, my personal blog three. One article becomes three localized versions with correct hreflang, because the workflow handles the multiplication. Keyword data decides the slug for each language, not a literal translation.

I stopped paying for managed WordPress hosting on my own sites and never looked back. WordPress still pays my bills through client work. My own properties never return to it.

Layer 5: The AI layer

Here is the uncomfortable part for the martech industry. The most productive tool in my stack is not a marketing tool at all. It is a coding agent running on an open source model.

It writes the scripts. It wires the APIs together. It builds pages, debugs tracking, and drafts the boring 70 percent of anything code-shaped. The model costs between $1.40 and $4.40 per million tokens, and I wrote about how it got to the top by changing nothing.

What it does not do is strategy. The model is trained on the average, so it suggests the average. Positioning, pricing, the decision of what to build next: those stay human. Automate execution, never judgment.

A small lime robot wires tiny machines at the workbench while he stands behind holding an unrolled blueprint: the machine executes, the human decides

If you are curious what that division of labor looks like in practice, the AI marketing playbook has the full workflow.

Layer 6: The plumbing people forget

Bookings run on a self-hosted Cal.com. The scheduling page, the confirmation emails, the reminder timing: all under my domain and my brand.

Lead notifications land in my chat system the second a form is submitted, with the source attribution attached. Email goes through my own mail server. Invoices go out from a self-hosted invoicing tool with a Swiss QR code on them.

None of this is glamorous. All of it removes a monthly subscription and a third party from the path between a lead and me. When something breaks, I fix it the same day, because it is mine.

He leans back with hands behind his head while dark wall pipes with flowing lime light connect a calendar, a chat bubble, an envelope, and a receipt into his tidy desk

What the whole stack costs

Vendor stack diagrams never show prices. Mine does, in honest ranges, because the exact numbers move.

Server, mail, and self-hosted tools: a two-digit euro amount per month, total.
Keyword and SERP data: pay per request, cents per check. Bursts, not subscriptions.
The AI layer: $1.40 to $4.40 per million tokens on an open model.
Domains: a few coffees each per year.

All of it together costs less each month than a single seat of the average marketing platform. When a demo shows you 40 logos, remember that each logo carries a per-seat price and an export fee.

The subscription I did not buy deserves its own line: an all-in-one SEO suite. Pay-per-request data costs me cents per check. A suite subscription costs three figures every month whether I check one keyword or ten thousand.

If you query data all day, subscribe. I query in bursts, so I pay per drink.

The rules I build by

If you take nothing else from my setup, take the rules. They work at any budget.

If it does not run without you, it is not a layer. It is a task. Automate it or delete it.
Data before tools. Pipes and one database first. Dashboards maybe never.
Own your data. Every system must export to something you control. Lock-in is a strategy failure, not an IT detail.
Every tool pays rent. Quarterly, ask what each piece earned or saved. No answer means it goes.
Boring beats shiny. The stack’s job is to disappear so the work can happen.

How to build yours in 30 days

You do not need my exact components. You need the order.

Week 1: connect your data. Analytics, Search Console, ad accounts, into one queryable place. Nothing else.

Week 2: automate one report you actually read. If nobody reads it, stop and pick a better report.

Week 3: automate one decision. Spend pacing, position alerts, negative keywords. The machine prepares, you approve.

Week 4: cut one subscription. You will not miss it. That is the point.

Thirty days, four moves, zero new logos on a diagram.

Frequently Asked Questions

What is a marketing tech stack?

The systems that collect, move, and act on your marketing data automatically: analytics, your site, your automations, and the pipes between them. The test is simple. If it could break during your holiday and nobody notices for a week, it is decoration, not stack.

How much should a marketing tech stack cost?

Far less than the industry implies. A solo operator can run sites, tracking, ads management, bookings, and reporting on a two-digit infrastructure bill plus usage-based API and AI costs. The expensive habit is per-seat software for jobs a script does better.

What is the difference between a martech stack and a marketing data stack?

The terms overlap, so keep them apart on purpose. The data stack is the pipes and the database, where numbers live and move. The tech stack is everything, delivery included.

Build the data stack first. The rest hangs off it.

The takeaway

The most useful page on Google for “marketing tech stack” is a Reddit thread, because the diagrams answer a different question than the one people have. The real question is never “which 40 tools”. It is “which six systems run my marketing without me”.

Mine: one server, one database, four pipes, static sites, a coding agent on a cheap open model, and self-hosted plumbing. It costs less than one SaaS seat and runs while I sleep.

If you want a second pair of eyes on yours, book a call with me. Bring your subscription list. We will find the six layers hiding underneath it.