Summary
With over a million+ orders per day and more than 2 million daily users, Talabat, Delivery Hero's MENA subsidiary, is the region's leading food and q-commerce platform. I joined as Senior Product Designer within the Fintech tribe, shaping talabat Pay (the in-app wallet) from a static MVP into a functional payments product with top-ups, transaction history and the groundwork for pay-later (BNPL) features.
Shortly after, leadership moved me to the organisation's most complex challenge: the Growth tribe, squad Rewards. The loyalty program was a strategic priority at C-level, yet adoption flatlined. Nobody was saving points, nobody was redeeming them, and the program wasn't retaining anyone. My mandate: figure out why, redefine the proposition, and prove its value with evidence instead of belief.
Talabat in the Appstore
The business challenge
Rewards suffered from a rare kind of problem: top-down conviction versus bottom-up doubt. Leadership saw loyalty as a flagship initiative; teams on the ground saw a costly program with low adoption, competing internal offers (subscriptions, other discounts) and a Flutter migration eating into every quarterly OKR.
The customer data confirmed the doubt. Qualitative research showed users weren't incentivised, had low awareness of the program, and were overwhelmed by terminology (points, missions, raffles, streaks) all for a single proposition. Quantitatively: high bounce rates and structurally low adoption.
Objectives
Two metrics defined success:
Increase adoption (users burning points monthly);
Increase GMV through the program.
However, underneath both sat a harder objective: shifting the organisation's thinking from a conversion funnel ("please push users to a catalog") to more behavioural loops ("make rewards part of the natural ordering journey").
Research & Discovery
I deliberately started with an unbiased exercise: a greenfield value proposition review, before groupthink or existing constraints could color the direction. Through seven cognitive walkthroughs I mapped five key friction points, from the expensive point economy (500 points for free delivery) to the total absence of rewards visibility during vendor selection. Together with the data team, the squad quantified the problem: over 95% of order GMV came from areas in the app where customers couldn't identify a rewards vendor at all.
The consequence: 74% of vouchers were redeemed passively at the basket, meaning: rewards couldn't claim it changed any behaviour, nor that it drove incremental orders for vendors.
To involve the wider product organisation, I facilitated structured ideation: brainwriting sessions with nine cross-functional participants across four journey areas (vendor list, search, menu, post-order), grounded in the Octalysis gamification framework. This diverged thinking across 40+ product designers and their PM counterparts, while keeping ownership within the squad.
Redefining the value proposition based on experiments
The existing proposition positioned talabat as a curator: earn points on every order, exchange them for benefits curated by talabat. Research pointed to a fundamentally different model: talabat as a framework: a platform that empowers vendors to reward their retaining customers, increasing GP for vendors and talabat alike.
From this, three strategic directions emerged: doubling down on gamified points, vendor-level rewards, or missions as the core mechanic. Rather than betting the roadmap on opinion, we translated all three into testable experiments.
Design experiments
Three experiments, three bets:
Experiment A: gamified loyalty mechanics (social proof, raffles, custom app icons). Session duration in Rewards rose 12%, but there was no significant order uplift. Clear signal: users value tangible monetary rewards over gamified incentives.
Experiment B: platform-wide monetary discounts (e.g. €2,50 off) within the core ordering journey. Order frequency rose 15% for users applying discounts, but the incentives were largely talabat-funded, making the model costly to scale and impossible to attribute to any vendor relationship. Effective, but not sustainable.
Experiment C: customer-vendor stamps, living on the vendor page with pushed nudges ("you're one stamp away!"). The winner. Drop-off in the redemption flow halved from 22% to 11%, notification CTR hit 24% against an 18% benchmark, average basket rose from €18.2 to €20.1, and vendor opt-in for loyalty campaigns grew from 25% to 45%, proving vendors were willing to fund loyalty once it demonstrably retained their customers.
Delivering through complexity
Experiment C made loyalty vendor-anchored but journey-wide with nudges across menu, notifications and basket. Which made it alignment-heavy across squads. I introduced a weekly design jam cadence with peer designers of affected squads, keeping 38 designers and their PMs aligned without slowing delivery.
Alongside the design work, I took on managerial responsibility within the tribe: weekly 1:1s and mentoring of two mid-level designers from neighbouring squads, plus interviewing new design hires.
Lessons learned
Staying close to the data team proved essential, framing and data biases can quietly turn discovery into top-down validation, especially without a unified C-level vision. And when belief and evidence collide, the designer's strongest tool isn't a better argument; it's a well-designed experiment.
Key results
The redefined proposition shifted rewards from a cost centre into a retention mechanism. Users engaging with the vendor-based loyalty program placed repeat orders 15% more often than non-loyalty users. Orders driven by stamp progress saw average basket value climb 10%, from €18.2 to €20.1. Customers added items to reach their next reward. On the B2B side, vendor participation grew from 25% to 45%, proving vendors were willing to fund loyalty once it demonstrably retained their own customers. Most importantly: the organisation stopped debating whether loyalty worked, and started scaling what demonstrably did.
+15% repeat orders
+10% avg basket value
+20% participating vendors
−50% redemption drop-off
