Case Study: Bayt.com

What They Do

Bayt.com is the largest job platform in the Middle East, with over 30 million users across the MENA region. At peak, they were sending 220–260 million emails per month — two to three sends per day — across a complex ecosystem of products serving both job seekers and employers.

My Role

Brought in to fix a declining email channel. Ended up rebuilding the entire email infrastructure from the ground up and eventually took on a broader role leading the full job seeker product experience.

The Challenge

When I arrived, Bayt was operating in what I can only describe as the dark ages of email infrastructure.

They were running Postfix — an open-source MTA with zero visibility. No metrics. No bounce processing whatsoever. Email bounces were being ignored entirely, meaning they were sending to dead addresses indefinitely with no cleanup mechanism. They couldn't control send rate, couldn't stratify by message type, couldn't see domain-level performance. The only signal they had was a month-over-month decline in email referral traffic in Google Analytics. That's when they called me in.

The compounding problems:

• No bounce management: Hundreds of thousands of bad addresses in the system, actively damaging sender reputation
• No message streams: Transactional emails, job recommendations, marketing campaigns, and third-party ad sends all went through the same process with equal priority
• No sending controls: No rate limits, no quiet hours, no throttling — just blast everything at once
• Microsoft/Outlook deliverability crisis: The steepest traffic drops were coming from Hotmail referring domains in GA. SRD vote policy changes at Microsoft had accumulated negative reputation that was suppressing delivery to the entire Hotmail family
• No double opt-in: Years of sending without confirmed email addresses, compounding list quality problems
• Organizational fragmentation: Multiple internal teams sending to the same database with no coordination, no standards, no approval process

And this was a 220M email per month operation. Not a startup.

What We Did

Emergency Triage — Microsoft Fix

Before anything else, I needed to stop the bleeding. Without domain-level send data (because Postfix couldn't produce it), I couldn't quantify the Microsoft problem precisely — but GA showed me where the traffic drop was concentrated. I hard-coded a temporary stop on all non-critical sends to Microsoft-hosted addresses. Only transactional emails and job recommendation emails (the most important email type in the product) continued to Hotmail, Outlook, and Live addresses. Everything else — marketing, ad sends — was paused for that domain family entirely.

Not elegant. But it stopped the reputation damage while we built the proper fix.

Bounce Management — Before PowerMTA

Before the infrastructure migration, we had no bounce processing at all. I manually coded approximately 500 bounce codes into soft or hard classifications and established a threshold: 5 soft bounces equals a hard bounce, triggering suppression. The dev team was not thrilled. But it worked — and it was the first time Bayt had any bounce management in its history. The fact that they'd sent for years without it and hadn't been blocked by major ISPs still surprises me. I'm not sure what saved them, but it wasn't good practice.

I also introduced a JavaScript email syntax validator at registration and integrated Kickbox for deliverability checking prior to send — catching bad addresses before they ever entered the system.

Infrastructure Migration — Postfix to PowerMTA

Migrating from Postfix to PowerMTA required buy-in from the infrastructure team, the dev team, and non-technical stakeholders. The commercial cost was approximately $25,000 per year — a significant argument to make. I made it on the basis of what was impossible without it and what it was costing them in traffic and reputation to stay on the existing system.

What PowerMTA enabled:

Multi-stream sending: Defined four distinct message streams by priority — transactional (highest), operationally critical (job recommendations), strategic marketing (CV completion nudges etc.), and lowest priority (third-party ad sends). Each stream had its own sending rules, rate limits, and reputation management. For the first time, a job recommendation email couldn't be delayed because an ad campaign was saturating the queue.

Proper bounce management: PowerMTA understands thousands of bounce codes natively. A 4xx code isn't always a soft bounce and a 5xx isn't always hard — every major ISP has its own taxonomy. Managing this manually was always going to be approximate. PowerMTA handled it properly.

Quiet hours: No sends overnight. Queue builds up, delivers in the morning when users are actually at their inboxes.

Send rate control: Capped at roughly 1.2 million emails per hour — no more bursting the entire queue at once. Counterintuitively, sending less more carefully meant critical emails reached inboxes faster and more reliably.

The net effect: we sent significantly fewer emails overall. And our key metrics went up. Because "upload a photo to your CV" is worthless compared to "here are your recommended jobs." Prioritisation made the important emails actually land.

Organisational Standards

My direct manager was the General Manager of the company — not the CEO, but close enough. That reporting line was the only reason I could walk into the ad sales team and tell them their campaigns were paused. Email departments typically sit low on the org chart, which is exactly why they don't get the authority they need to do this work properly. I had an unusual structural advantage and I used it.

Beyond enforcement, I ran training and workshops for every team that touched email. Not just rules — education. What good sending practice looks like, why frequency matters, how reputation works, what HTML and deliverability mean in practice. I wasn't just a cop. I was also a teacher.

Content, Segmentation and Lifecycle

Rebuilt the job seeker lifecycle — onboarding, retention, re-engagement — from scratch. Stopped campaigns that weren't earning their place. Introduced new ones calibrated to actual user behaviour and lifecycle stage.

The MENA market added real complexity. Dubai: high-tech, email-native, app-friendly. Egypt at the time: low-tech, SMS-preferred, email confirmation rates low. Arabic and French alongside English. RTL email coding alongside LTR. Three languages, two text directions, wildly different engagement patterns across the same region. The company wanted to be truly pan-MENA — not just Dubai — and making the email programme work across that spectrum was genuinely difficult.

The Ramadan factor also deserves mention. The dev team was based in Jordan. During Ramadan — effectively six weeks when you account for the wind-down — development velocity drops significantly. Heavy infrastructure refactoring and Ramadan don't mix well. You learn to plan around it.

Results

• 90%+ inbox placement rate — achieved without leasing expensive third-party infrastructure or certifications
• 140% increase in open rate
• 150% increase in email-driven site traffic year over year
• Microsoft deliverability fully restored after a sustained programme of technical and operational fixes
• First functioning bounce management system in Bayt's history
• Multi-stream sending infrastructure replacing an undifferentiated single queue
• Promoted to lead the full job seeker product experience following the email turnaround

What This Project Taught Me

Scale changes everything. The same principles apply, but the blast radius of every decision is enormous — and the organisational complexity is a separate challenge from the technical one. Getting a dev team in Jordan, an infrastructure team, an ad sales team, a content team, and a GM all aligned on email standards required a different kind of persuasion than writing good automation flows.

Bayt pushed my skills to the limit across every dimension simultaneously: technical debt, language barriers, cultural complexity, RTL coding, stakeholder management, and sole accountability for 220 million emails per month. I'm proud of what we built there.

Stack

PowerMTA · Postfix (inherited) · ReturnPath Inbox Monitor · Litmus · Kickbox · Google Analytics · Custom bounce classification system · Multi-stream send architecture · RTL email coding