Labeeb Mirza — Performance Marketer and Indie App Builder

Every marketing team eventually pays someone a few thousand a month to move numbers out of ad platforms and into a place where a human can read them. I got tired of paying that bill and built the thing instead.

It has turned into the tool the team actually works in. Campaign dashboards that drill down to a single campaign and its impression share. A performance tracker that sorts the week into winners, losers, and the ones holding steady. A blocking chart for planning flights across channels, with Excel import and export, because that is the format media plans arrive in whether you like it or not. A QR generator that issues tracked codes and reports the scans back with location, so print and out of home stop being a black hole. An ad copy generator that produces Google Ads Editor ready CSVs and learns from what the team marks good or bad. Scheduled data pushes into Snowflake or Supabase for anyone who would rather query it raw. Experiments, weekly reports, a logbook, and permissions so people only see the parts that are theirs.

There is also a chat that investigates a question against live account data. The thing I insisted on is that deterministic checks run between the model steps, so it cannot quietly invent a number and hand it to me with confidence.

It connects to roughly two dozen sources. The ad platforms I buy on, analytics, search console, and the warehouses on the far end. Data lands in Postgres, gets cached so nobody waits on a slow API twice, and syncs on a schedule.

The part I actually care about is the diagnostic layer. A dashboard tells you spend went up. This works through the likely causes in order: budget, targeting, landing page, creative, fatigue. When the answer really is the copy, it drafts new copy, and it snapshots the campaign first so I can come back in three weeks and see whether the change did anything.

Aeroly started because I wanted to watch a specific plane move. Every tracker I tried showed me a row in a table. I wanted the globe.

So that is what it is. A dark 3D globe with real aircraft on it, flight paths drawn as great circle arcs that behave correctly when they cross the antimeridian, a night line sweeping across the map, and airports detailed down to runways and gates. Positions arrive every few seconds and get smoothed with dead reckoning in between, so nothing on screen ever jumps.

My favourite part is the delay prediction. Your flight is rarely late because of your flight. It is late because the aircraft is still sitting on a runway two cities away. Aeroly looks up the registration, finds the inbound leg, does the turnaround math, and tells you before the airline does.

Splitting bills with friends is a social problem wearing a math costume. Divide-It handles both halves.

Groups, expenses, settlements, and a simplifier that collapses everybody owing everybody into the smallest number of payments. iOS also has a personal tracker with a calendar, cashflow, recurring items, and a receipt scanner that reads a photo and assigns the lines to people.

Two codebases, on purpose. iOS is native SwiftUI with a glass material system I built and then spent a long time making not drop frames. Android is the opposite look. Black and white, flat cards, no glass anywhere. The money engine underneath is the same on both, kept in parity down to the rounding.

I lift, I log it, and every app I tried wanted a monthly fee for the privilege of storing my own sets. So I started writing my own.

Routine builder, set logging with rest timers, supersets, warm-ups, RIR. The stats screen is the fun one: an activity heatmap, a body map that shades whatever you have been working, and a recovery model that tells you what is still cooked.

The exercise catalogue is 1,324 movements sitting in a read-only database. Nothing about you goes to it. Your training history stays on the phone. It is early and I will say so plainly: most of the screens exist, not all of them work yet.

I run paid acquisition for Clutch.ca, Canada's largest online used-car marketplace. Search, Performance Max, YouTube, Meta, Reddit, TikTok, programmatic display, Bing, and connected TV.

The interesting wrinkle is that the business has two sides that behave nothing alike. People selling a car want a number and they want it now. People buying one take weeks, read reviews, and come back through four different channels before they convert. Same budget, two completely different jobs, and a lot of arguing with attribution about which one deserves the credit.

Most of my week is diagnosis rather than button-pushing. Something moved, and the job is to find out whether it was the auction, the creative, the landing page, the seasonality, or a change somebody made eleven days ago and forgot to write down.

Six years of this now, in four fairly different rooms. Agency side first, planning media for Kia and Hyundai at Innocean, then for Samsung Canada at Starcom, where a launch either hits its pre-order number or it does not. After that, in-house at Freedom Mobile running $5M a year across digital and traditional, television and radio and out of home included. Now all of it is performance, on a marketplace with two sides to buy for.

PERFORMANCE

Google Ads, DV360, CM360, YouTube, Meta, TikTok, X, The Trade Desk, StackAdapt

BRAND AND STRATEGY

Go to market, campaign planning, integrated launches

TRADITIONAL

Television, radio, out of home, print

MEASUREMENT

GA4, Mixpanel, Rudderstack, Google Tag Manager, SQL, Python, geo lift and brand lift studies

A friend was studying for the G1 on a site that buried every question under three ads and a popup. I built the version I would have wanted: questions written from the official handbook, a full exam simulator, and nothing to sign up for.

Over 300 questions across road signs, rules of the road, and fines and demerit points, each one citing the handbook passage it comes from. A readiness meter works out your pass probability from your recent attempts, and a challenge bank holds on to the ones you got wrong until you stop getting them wrong.

It runs in the browser, keeps progress on the device, and will read questions aloud if you ask it to. No account, no paywall, nothing to install.

Every clipboard manager I tried wanted me to learn a new shortcut. ClipIt uses the one I already know. Tap Cmd+V and it pastes, exactly as it always did. Hold it and a panel appears with everything you copied recently, arrow keys to move through it, release to paste the one you want.

It sits in the menu bar with a searchable history, and you can pin the handful of things you paste constantly. There is an exclusion list as well, so it stops recording while your password manager is in front. A lot of sensitive things pass through a clipboard and none of them need to end up in a database.

Native Swift and SwiftUI, stored on device, no account, and it updates itself. macOS 14 and later.