Case · Uber Eats EMEA · Apr 2024 – Apr 2025

One in six emails never landed.

Aug 7, 2026 · 5 min read · verified in raw delivery logs

250,000 merchants, 17 countries, two million messages a month — and global deliverability stuck at ~84%. Everyone had learned to live with it. The defect turned out to be one nobody was looking for, in the one place nobody owned.

The number everyone had learned to live with

Merchant communications at Uber Eats EMEA: 250,000 restaurants, 17 countries, over two million messages a month — onboarding, payouts, menu changes, incident notices. Global deliverability: ~84%.

84% reads like a solid B. In email it means one message in six does not exist — at this volume, three hundred thousand messages a month landing nowhere. A payout notice that never arrives is a support ticket; an onboarding email that never arrives is churn.

Underneath: ~2,000 automations accumulated across teams and regions, no intake process, no cross-region visibility, no single owner. Every dashboard reported sends and opens. No dashboard reported “never arrived.”

The question that reopened a settled number

My audits start with one question: what does the CFO see that the team doesn’t believe is true? The CFO saw “two million messages, 84% delivered.” The team saw list hygiene, IP reputation — normal at our scale. The honest version of the question was simpler: why are we okay with one in six not landing?

Nobody was un-okay with it, because the number had no owner. That is how deliverability decays everywhere: invisibly, politely, one percentage point at a time — and by the time bounce-rate creep shows up in a vanity dashboard, months of sends are already gone.

The diagnosis: the platform, not the list

First move: visibility. I built the first EMEA merchant CRM dashboard — SQL over the SFMC warehouse, one source of truth instead of per-team screenshots. With the actual delivery logs in one place, the failure pattern stopped looking like reputation: it did not follow list age, segment, or sending domain.

The trail ended at a platform-level SFMC defect — configuration, not content. No amount of list cleanup would have touched it. Hygiene work would have polished the list forever while the platform quietly dropped the mail.

The fix — and what it compounded into

The defect was escalated and fixed — and because it was platform-level, the fix shipped worldwide: EMEA, LATAM, US&CA, ANZ. Every market benefited, not just mine.

With mail actually arriving and the system finally visible, the rest was disciplined lifecycle work: an intake process that ended shadow campaigns, and A/B + AMPscript repairs on journeys that could now be measured honestly.

The numbers

ⓘ Verified in raw delivery logs, not platform vanity metrics. The one forecast below is labeled as a forecast.

~84% → 95%+deliverability · global
43.9% → 52.7%open rate · +20%
1.53% → 2.28%CTR · +49%
>80×WhatsApp beta ROI vs email
€25MFY26 forecast · co-authored strategy
12 moApr 2024 – Apr 2025

The €25M is the incremental-impact forecast in the Merchant Education CRM strategy I co-authored — a forecast, not a booked result. Every other number above is a measured delta.

What this case is actually about

Deliverability is the quietest metric in the stack: it degrades invisibly, and teams reach for the explanation their tools can see — reputation, list quality, subject lines. When the platform itself is broken, all of that is cosmetics. The leak is always bigger than anything built on top of it.

“Oleks quickly became a valuable part of Uber, initiating scalable improvements that benefit the broader business well beyond a single market. He combines deep CRM systems expertise with a very pragmatic, business-first mindset.”

Ricardo Zertuche · Director of Marketing Technology, Uber

Run your own numbers — the Deliverability Cost Calculator →All cases, numbers and provenance →

This is the thinking. The Diagnostic is where I point it at your stack.