Your Marketing Stack Was Built by People Who Quit

Go into your marketing automation platform right now and open the list of active workflows. Not the ones you built. All of them.

Read the names.

Lead Routing v3 FINAL. DO NOT DISABLE (ask Brian). Nurture Q2 test COPY COPY. temp fix. [NEW] Scoring Model, last edited four years ago.

Brian does not work here. Brian has not worked here since the office still had a ping pong table. And Brian’s workflow is still running, still writing to a field, still quietly deciding which leads your sales team sees on a Tuesday morning.

That is not a hygiene problem you get to schedule for next quarter. That is your reporting layer.

Every stack is a landfill with a login screen

Marketing systems do not get designed. They accumulate. If you dig into any company past about eighty employees, you can date the layers like rock strata, and each layer is somebody’s tenure.

• The 2019 layer. A demand gen manager who believed in lifecycle stages with a religious intensity. Nine of them. Two are still used.
• The 2021 layer. The ABM pivot. Account tiers, intent fields, a scoring model built in a spreadsheet that was pasted into the CRM by hand.
• The 2023 layer. An agency rebuilt the UTM taxonomy, documented it in a Google Doc, and then the contract ended.
• The 2025 layer. Somebody connected an AI tool that writes back to a custom field nobody named, and now half your contacts have a value in ai_summary_2.

No layer was ever removed. Not one. Every tool in this industry ships a beautiful create flow and a delete button hidden behind three confirmations and a warning that says this action cannot be undone. The software is telling you exactly what it expects you to do, which is add.

Hoarding is free. Deleting is a career risk.

Here is the actual reason your instance looks like this, and it has nothing to do with laziness.

If you turn off an old workflow and it turns out something depended on it, you own an incident. Your name goes in the Slack thread. Sales stops getting routed leads for two days and you get to explain it on a call with people who have no interest in why you were cleaning up. If instead you leave four hundred dead objects running forever, absolutely nothing happens to you today. Or tomorrow. Or ever, at least not in a way anyone traces back to you.

Nobody ever got fired for leaving it on.

So the rational individual move is to leave everything running and add your own layer on top. Repeat that across six people over seven years and you get a system that nobody in the building can fully explain, running the numbers that go in the board deck.

The sediment is what your numbers are made of

This is where it stops being funny.

Your MQL definition is a rule someone wrote to hit a target in a quarter that ended before half the current team was hired. Your lifecycle stages get overwritten nightly by a sync that a former ops contractor set up in one direction on purpose, for a reason he explained once, in a meeting, verbally. Your top-line channel report quietly excludes a region because of a filter added for a single board presentation in 2022 and never removed.

Then pipeline dips, and leadership asks why.

The honest answer is sometimes: a filter written by a person we no longer employ is excluding a market we now sell into. Nobody says that in the meeting. It sounds like an admission that the numbers are not real, which is a much bigger conversation than anyone booked thirty minutes for. So the meeting gets the other answer. Seasonality. Softness in enterprise. A tough comp against last year.

Everyone nods, because the alternative is opening the layers.

The archaeology tax

The expensive part of your stack is not the invoice. It is the gap between someone asking a quick question and anyone producing an answer they will actually stand behind.

Every one of those answers requires a dig. Whose definition of the metric are we using. Which field is authoritative when two disagree. Was that campaign tagged before or after the taxonomy changed. Did the sync run. Is that spike real, or did somebody bulk import a list.

None of that work shows up in a budget line. It shows up as a senior ops person spending Thursday reconstructing history instead of doing anything that grows the business, and as a marketing leader who has learned, correctly, not to trust a number until a human has personally blessed it.

You cannot mandate your way out of this

The industry has one prescribed answer here, and it is governance. Write a naming convention. Run a quarterly audit. Stand up a center of excellence. Hold a workshop about data stewardship where everyone agrees that going forward, we will all be disciplined.

It fails every time, for a boring reason: it asks humans to maintain sustained discipline against a task with no deadline and no visible consequence for skipping it. The convention doc gets written by the one person who genuinely cares about this, and that person is also the one most likely to get a better offer somewhere else. Then the doc becomes another artifact in the landfill, cited by nobody, at version 1.4.

Governance is not wrong. It is just not load bearing. Anything that depends on everyone remembering is already broken.

What actually helps

Three things, all unglamorous.

Put a name and a date on everything you create. Not a naming convention document. Just an owner and a review date in the object name itself, where the next person is forced to see it. Anything with no owner is a candidate for off.

Turn things off loudly. Announce it, wait a week, disable rather than delete, keep it recoverable for a month. The reason nobody cleans up is fear of the silent break. Make breaking loud and cheap and people will clean up.

Stop letting the archaeology decide whether you get an answer. Your systems of record are going to stay messy. That is a permanent condition, not a project. What you can change is whether every question has to be answered by excavating them by hand. Asking what happened last month should be a question you ask in plain language against everything at once, not a research project with four CSV exports and a person who remembers 2022.

Your stack is not going to get simpler. There will be another layer next year, and eventually someone will inherit yours and wonder what on earth you were thinking with the custom properties. The question worth asking is not how to clean up all of it before you are allowed to trust anything. It is whether your ability to get a straight answer should depend on remembering who Brian was.

If you would rather that answer come from one place than from four exports and a seance, that is roughly why we built THE DASHBOARD. And if not, fine. Just put your name on the workflow before you leave.