Nobody Knows How It Works Anymore

The bus factor in your marketing stack — and why the bill always arrives at the worst

possible moment.

There is one person at your company who can fix the campaign that breaks on Black Friday. Have you met them? Do you know if they are happy?

Most marketing leaders cannot answer that question. The ones who can usually answer it too fast, with a name and a nervous laugh.

Here is a better test. Ask your team what happens to your abandoned-cart journey if the person who built it takes a two-week holiday with no laptop. If the answer involves the phrase "we would probably just pause it," you do not have a technology problem. You have a single point of failure with a LinkedIn profile.

Engineering teams have a name for this. They call it the bus factor: the number of people who would have to get hit by a bus before the system becomes unmaintainable. In software, a bus factor of one is treated as an incident. In marketing, it is treated as Tuesday.

How marketing stacks quietly become undocumented

Nobody decides to build a system only one person understands. It happens the way debt always happens — one reasonable decision at a time.

A workflow gets built under deadline pressure, with a note to clean it up later. Later never arrives. Someone adds a temporary condition to handle a 2019 promotion, and it is still firing. An agency builds the welcome series and hands over templates but not logic. A contractor writes a data import job that works beautifully, then leaves. Three reorgs happen. The person who knew why the segmentation excludes customer type 7 moves to a different company.

What is left is a platform that runs perfectly, right up until the moment somebody needs to change it.

The Tells

You already know if this is you. But in case it is comfortable to look away:

•     Every change request routes through the same person, regardless of size

•     New hires take six months to become useful, and nobody can explain why

•     There is a workflow everyone is scared to touch

•     Your team can tell you what a journey does, but not why it was built that way

•     Documentation exists, and it was last accurate two platform versions ago

•     Someone recently said the words "let us not change that, it works"

That last one is the loudest signal of all. "It works" is what people say about systems they no longer understand.

This is not a people problem

The instinct is to blame the specialist. Hoarding knowledge. Building job security. Refusing to document.

Almost always wrong. That person is usually the most overworked human in the building — the one who never takes real time off because they know exactly what happens when they do. They did not build a moat. They kept the lights on for four years while nobody funded the maintenance.

Blaming them also misses the mechanism. Knowledge does not disappear because someone hides it. It disappears because there was never a moment in the process where writing it down was anyone’s job.

What it costs, in three currencies

Speed. Every change gets queued behind one calendar. Your campaign velocity is not limited by your platform’s capability — it is limited by one person’s availability. That is why "simple" requests take three weeks.

Risk. Undocumented systems degrade silently. Integrations break and nobody notices until deliverability drops. Compliance asks who has access to customer data, and the honest answer takes a week to assemble.

Money — usually at the worst possible time. The bill for years of undocumented drift almost always arrives during a migration. You go to move platforms and discover you cannot inventory what you have. Now you are paying to reverse-engineer your own marketing operation before anyone can move a single campaign. That is the surcharge nobody budgets for.

Fixing it is less work than you think

It is not a documentation project. Documentation projects fail, because writing 200 pages nobody reads is a way of feeling productive without reducing risk.

What actually works is narrower:

Inventory first, and be ruthless. Most stacks are carrying hundreds of assets, workflows and templates that have not run in years. You do not need to document those. You need to delete them. The surface area you have to understand shrinks dramatically once you stop protecting things nobody uses.

Capture decisions, not clicks. Screenshots of a UI go stale in one release. The durable knowledge is why — why this suppression rule, why this exclusion, why this data model. That fits on one page per journey.

Write runbooks for the ten things that actually break. Not everything. The ten. Failed data loads, stuck workflows, the integration that times out. That is most of your operational risk, handled.

Put a second pair of hands on the critical path. The most reliable de-risking move is structural: make sure at least two people — internal, partner, or both — can execute anything that matters.

What this looks like in practice

We spend a lot of our time inside marketing platforms that outlived the people who built them. Two examples.

When an Entertainment Industry Giant Partnered with Celerity

They ran their Adobe Campaign recipient imports through a single-threaded file exchange process. Twenty to fifty files a day, some carrying tens of millions of records, all queued behind whichever file happened to arrive first. A large file could hold the pipeline for hours.

The marketing team had adapted. They knew not to bother submitting a small campaign in the morning. That knowledge lived in people’s heads, not in a ticket — which is precisely why the underlying problem went unaddressed for so long. A workaround that everyone knows stops looking like a defect.

We rebuilt it: queue-based, multi-threaded recipient validation, and file routing by size so small and large files run in their own lanes. The result was roughly 10x faster processing, and small campaigns launching the same day instead of waiting most of it out.

A digital banking client: turning tribal knowledge into an operating system

A fast-growing consumer fintech had the platform, the data and the ambition — but execution depended on a small number of people who held the whole picture. Every send needed legal review, eligibility checks and suppression logic, and all of that context sat with individuals rather than in the operation.

We embedded a team and scaled it to six specialists, then did the unglamorous part: a 30-page playbook covering naming conventions, reusable content blocks and deployment standards, alongside documented QA and production-launch processes and training materials so new team members could get productive without leaning on the veterans.

The outcome was not just tidier. Campaign velocity rose roughly 5x against baseline capacity, engaged users grew 30–35% over the year, and 20+ personalized next-best-product journeys went live — because capacity stopped being rationed by who was available.

Our Platform Optimization and Managed Platform Operations services exist for exactly this. We inventory what you actually have, retire what you do not need, write the runbooks for what breaks, and then run the day-to-day alongside your team — so the knowledge lives in the operation rather than in one person’s head.

The goal is not to replace your specialist. It is to let them take a holiday.

If one person left your marketing team tomorrow, what would stop working?

If you have to think about it, that is the conversation worth having.