Issue 01: The Pipes Matter More Than the Paint

I have watched a lot of lifecycle redesigns ship over the past five years. New copy, new design system, new flow logic, new everything. Sometimes the numbers move. Most of the time they do not. The teams that own the program are good. The new creative is good. So when the numbers do not move, the question becomes uncomfortable: what was actually broken?

In almost every case, what was actually broken was the data underneath.

I want to make a specific argument this week. The lifecycle marketing industry has been mostly arguing about creative for a decade. Better copy, better design, better personalization, better testing. All of that matters. None of it is the constraint. The constraint is upstream: the events that fire, the properties they carry, the identity model that ties them together, and the syncs that move them from the warehouse to the campaign platform. When the data layer is right, mediocre creative outperforms great creative on a broken layer. We see this every week.

This is also why Customer.io's AI Decisioning launch this year is such a meaningful moment for the category. AI Decisioning automates message, channel, and timing decisions at the profile level — but it only works when the profile data is clean. Companies that have been investing in their data layer can switch on AI Decisioning and immediately see lift. Companies that have been decorating over a broken data layer turn it on and get worse outcomes than their rule-based logic produced, because the model is making more decisions on bad data. The split is widening fast. Teams that fixed the pipes are pulling ahead.

Three pieces of evidence from this week's content.

First, event schemas. The pillar post we published Monday makes the case that most schemas in production today were designed by engineers for product analytics. They work fine for dashboards. They fail for orchestration. The event names are inconsistent. The property sets are incomplete. Identity is not stitched. A lifecycle team inherits this schema and tries to build campaigns on top of it, and every campaign is a workaround for a missing property or a missing event. The campaigns ship and underperform, and the post-mortem blames copy. The copy was fine. The schema was the problem.

Second, stale Customer.io data. Tuesday's post walked through the five most common causes of stale data in Customer.io. Four of the five are configuration-level. None of them is a creative problem. A reverse ETL sync that has been broken for three days is sending campaigns to users with month-old plan tiers. A missing identify call at signup is creating two profiles for every user. None of this is visible from inside the campaign platform. The campaigns just underperform, and the team tries new copy.

Third, duplicate users. Thursday's post put rough math on the cost of duplicate profiles in Customer.io. Most PLG companies have 8-15% duplicates. Every campaign sends to inflated volume. Deliverability degrades. Segments are partly fictitious. Attribution is broken. AI Decisioning treats the same person as two decision units. None of this is visible until you run the detection queries and see the numbers. And none of it is fixable in the campaign platform — the fix is in the schema and the CDP.

The pattern across all three is the same. The visible problem is in the campaign. The actual problem is upstream. The campaign platform is the place where you discover the upstream problem, not the place to fix it.

This is uncomfortable for marketing teams because it means the next round of investment is not in better creative or better testing. It is in event schema design, identity resolution, and pipeline reliability — work that does not feel like marketing and that requires sustained collaboration with engineering and data. The work is unsexy. The payoff is large. Most teams skip it because the payoff is invisible until you do it, and the unsexy work is hard to fund when there are obvious creative problems that look more solvable.

I am going to keep arguing this all month. Next week the focus shifts from data foundations to MarTech architecture — the platform layer that sits on top of the schema. Configuration, deliverability, identity, frequency capping. The same theme: most stacks fail not because the platforms are bad but because the configuration is wrong and nobody is owning it.

If you want to test whether your data layer is the constraint, the fastest signal is to run the audit we walked through in Sunday's post. It takes one afternoon. The output is an honest list of what is fixable in the campaign platform and what is not. If the list is mostly schema and identity, that is your answer.

One specific next step: read the event schema design pillar if you have not yet. It is the longest piece we have ever published and the most important. Everything else this month builds on it.

‍