FY2025 group revenue
$7.9bn
Up 4.1% YoY
Daily shoppers
680k+
Published footprint
Retail space
1.38m m²
GCC network
H1 2026 group revenue
$3.99bn
Down 2.6%
H1 2026 net profit
$88.9m
Down 30%
Q2 gross margin
23.0%
Broadly stable
GCC store estate
283
Six countries · June 2026
Year-one GM added
OMR 112k
2.1× layer cost
01Situation
Where the group's profit moved in the first half.
The group reported a half year in which revenue moved slightly and profit moved a great deal. Food grew modestly while non-food fell. Management cited weaker discretionary sentiment, slower high-value electrical sales, and market-led price reductions in lifestyle. At this margin shape, modest non-food recovery is worth a disproportionate amount at the bottom line.
Measured
Where the decline sits
Food revenue grew modestly in Q2 while non-food fell, especially high-value electrical and lifestyle.
Modelled
Operating leverage
Net margin is close to 2%. Ten million dollars of gross margin is therefore a large proportion of half-year profit.
Measured
Cost route already taken
Q2 operating expense growth was held to 1.2%, and employee costs declined even as stores opened. Demand is the remaining lever.
Measured
Scale already in place
FY2025 revenue reached $7.9bn (AED 29.1bn), up 4.1% year on year, across 1.38 million square metres of retail space and more than 680,000 daily shoppers.
Measured
Expansion remains funded
Lulu has publicly outlined 50 additional GCC store openings, increasing the value of proving a repeatable Oman operating model before wider rollout.
Oman specifically. The group places Oman inside other markets, which grew 8.2% in Q2 2026. Oman is not the problem geography; its stability, 33-store estate, and local decision path make it the right place to prove the method.
02The asset
A reachable member base with a reason to call.
Most loyalty programmes are treated as discount mechanisms. Happiness is structurally more valuable: enrolment requires a name and Oman phone number, points are earned in store, and every balance carries an expiry date. Members are reachable with a service reason to talk—not merely a promotion.
Measured
What is published
Around 67% of 2025 group revenue was linked to loyalty members. UAE registered baskets averaged 1.7× unregistered baskets. Oman launched in November 2024; points expire after one year, vouchers after 90 days, redemption starts at 40 points, and 400 points equal one rial.
Modelled
What follows from it
Voice and WhatsApp need no purchased data or app install. Expiry supplies the service reason; department history supplies the topic. No new customer-side infrastructure is required.
Every member has a phone number
Voice and WhatsApp are native channels for this base.
Expiry creates relevance
A service message opens the conversation without spending margin.
History sets the topic
Department-level behavior makes each call specific.
03The trigger
Four populations inside one member base, each needing a different call.
A single blast to the whole base is what the programme can already do—and why response rates are where they are. The work is splitting the base into groups with a real reason to hear from Lulu, then having a conversation rather than sending a message.
Lapsed on non-food
176,800Trigger: still buying food, but no non-food purchase in 9–24 months. Conversation: their former category, what changed in the range, and today’s member price.
Points expiring
93,600Trigger: above the 40-point redemption floor and expiring within 90 days. Conversation: what they hold, when it goes, and what it covers—without an offer.
Basket narrowed
114,400Trigger: visit frequency held while department count fell by two or more. Conversation: which departments dropped and whether price, range, or availability drove it.
High value, gone quiet
15,600Trigger: top spend quintile with no visit in 60 days. Conversation: a person calls; the agent only identifies and briefs the store manager.
Lead with expiring points. It is the only trigger that costs nothing, carries no discount, and is unambiguously in the member’s interest. It also opens the lapsed-category conversation: after saying what the balance is worth, the natural next sentence is what it covers.
04Mechanism
Six agents, two channels, one human gate.
Detection runs nightly. Voice and WhatsApp carry the conversation in the member's registered language; uncertainty and high value stop at a named person.
520,000
Records scanned
601
Flagged today
421
Voice calls
331
24
Human handovers
108
Suppressed
Native stack, not translated output
The group serves more than 695,000 shoppers daily across 130 nationalities. In Oman, Arabic, English, Malayalam, Hindi, Urdu, Bengali and Tagalog all appear in volume. Native intelligence is the difference between a call that completes and one that ends in ten seconds.
A note on scope. India’s 74-language figure applies to text-to-speech. The conversational stack covers about 32 languages, including every Oman language listed here.
Agent roster
Live operating designExpiry watcher
SystemRanks expiring balances by value at risk.
Lapse detector
SystemHolds the lost department, not only the lapse.
Reactivation caller
VoiceCalls in the member's registered language.
Follow-up agent
WhatsAppSends price, nearest store and expiry date.
Store handover
Human gateRoutes value and uncertainty to a named manager.
Consent & suppression
SystemBlocks opt-outs, excess frequency and restricted hours.
Conversation feed
One day of the reactivation layer in Oman
Every line ends in an action, a human handover, or a documented refusal to act.
System
Overnight scan complete
Balances ranked by value at risk; members below the 40-point redemption floor excluded.
Queued for contact · Done
Voice · Malayalam · 2m 48s
Expiring balance; appliance lapse
The member learned the points expire. The agent introduced the changed small-appliance range and current member price.
Price, store and expiry sent · Done
Voice · Arabic · 1m 12s
Stock question at Al Qurum
With no verified stock position, the agent made no claim, offered a checked callback, and logged the gap.
Routed to store · Human
WhatsApp · Bengali
Two-day follow-up
The member asked whether points combine with member pricing; programme terms supported a direct answer.
Answered from terms · Done
Human gate
Top-quintile member, quiet for 63 days
Spend history, dropped departments, and expiring balance went to the store manager. Contact was made by a person.
Store acknowledged · Done
System
Contacts blocked before dialling
Opt-outs, frequency-cap breaches, and calls outside permitted hours were suppressed with auditable reasons.
Suppressed and logged · Done
05Model
The chain from member base to gross margin.
The loyalty linkage and basket multiple are published anchors. Every other number is a DeployOne assumption and remains adjustable.
Oman annual revenueOMR 261m
Non-food revenueOMR 73.2m
Linked to loyalty membersOMR 49.0m
Lapsed on non-food176,800
Addressable in year one150,280
Reached and completed61,615
Converted to purchase8,626
Incremental spend / memberOMR 50.0
Incremental revenueOMR 431k
Gross margin on non-food26%
Gross margin added, year oneOMR 112k
Cost of the layer, year oneOMR 54k
Net contribution, year oneOMR 58k
Year-one return
2.1×
Net contribution OMR 58k
Implied lift on Oman non-food is 0.6%.
OMR 112k gross margin against OMR 54k cost returns 2.1× in year one.
The number to handle carefully is conversion. Everything else is published or observable from the first thousand calls. Conversion genuinely cannot be known before the pilot, which is why the pilot is priced against it—not around it.
Control
What we measure against
A held-out group from the same lapsed population, matched on spend and tenure, never contacted. Incremental revenue is the gap—not a comparison with last year.
Not claimed
What we do not claim
No claim on total loyalty revenue, basket uplift across the whole base, or food. One population and one category group decide the result.
Learning loop
Where it gets cheaper
Reasons for category exit become structured targeting data. The second cycle calls fewer people, converts more, and lowers cost per incremental rial.
06Governance
Contact rules, consent, and what the system will not do.
A programme called Happiness cannot survive calls members experience as pressure. These constraints are what make a base of this size safe to contact.
- Never state a points balance the live extract does not support.
- Never assert stock or delivery availability without verification.
- Never contact outside permitted hours or inside a prayer window.
- Never exceed the cross-trigger frequency cap.
- Never call a member below the redemption floor about expiring points.
- Never write to the Happiness points ledger—read only, without exception.
- Never route a high-value member away from the agreed human treatment.
Consent and data
Contact runs on consent captured at enrolment and the programme’s live opt-out state—not a separate list. Suppression decisions are logged with reasons. Personal data handling follows Oman’s data protection framework and Happiness privacy terms, with Lulu legal sign-off before the first call.
The extract is deliberately narrow: member identifier, registered language, consent state, points balance and expiry, and department-level purchase history. No payment data and no identity documents.
Consent verified
Offer caps set
ROI guardrails
Legal sign-off
On LuLu 2.0. The Digital and Omnichannel Transformation Office and Mastercard work sit on the inbound journey, where a customer is already shopping. This layer handles outbound conversation with a member who is not shopping. They do not overlap or compete for the same integration, and this can run at country level while the group programme continues.
07Pilot
Ninety days, one country, one control group.
WEEKS 01–02
Extract & sign-off
Size populations, lock the control group, approve rules and success metrics.
WEEKS 03–04
Language calibration
Build against the registered languages and calibrate with the store team.
WEEKS 05–11
Live campaign
Expiry first, lapse second; weekly readout against control.
WEEKS 12–13
Result & decision
Read control lift, cost per incremental rial, and reasons for category exit.
Success 01
Revenue
Incremental non-food revenue against the held-out control group, per contacted member.
Success 02
Efficiency
Cost per rial of incremental gross margin, compared with acquiring the same demand through promotion.
Success 03
Programme health
Opt-out and complaint rates, with an agreed ceiling that stops the pilot automatically.
Commercial shape. Fixed price for 90 days, billed monthly in advance, with programme health as a stop condition. If the three measures do not land, the pilot ends and nothing further is owed. Rollout pricing is agreed before the pilot starts.
08Scale
How this moves from Oman to the rest of the group.
Oman is not the biggest opportunity. It offers a local decision, a workable 33-store estate, and nearly two years of Happiness purchase history against which to detect a lapse.
UAE
Then the same method, larger base
The largest member base and longest programme history. The method transfers without redesign; only language mix and contact rules change.
Saudi Arabia
Then the market that needs it most
Revenue fell 11.3% year on year in Q2 2026. It is where a demand instrument matters most—and where arriving without proof would be a mistake.
Separate scope
Then the adjacent triggers
The same detection layer can support supplier communication, fresh availability, and e-commerce order care. Those are separate conversations and are not proposed here.
Group gross-margin potential
OMR 1.29m
Same assumptions · 6m members
09Questions
Nine facts Lulu can use to replace the assumptions.
On the base
How many Happiness members are registered in Oman, and how many transacted in the last 90 days? What proportion registers a language other than English or Arabic? What share holds a balance above the 40-point floor?
On the data
Is purchase history available at department level against the member identifier, and for how far back? Is consent and opt-out held inside the programme or separately? Who owns the extract—Oman or group?
On the decision
Can Oman authorise a 90-day pilot, or does member contact require group sign-off? Must the Dubai transformation office review outbound contact or only inbound journeys? Who owns rollout if it works?
Prepared by DeployOne. Published figures are drawn from Lulu Retail Holdings PLC FY2025 and H1 2026 disclosures, Lulu corporate materials, and public reporting. Modelled figures are DeployOne assumptions and remain adjustable. Nothing in this canvas is a forecast of Lulu’s results.
Growth pilot · operating layer
Lulu Oman growth command center.
The decision canvas becomes a daily workspace for conversations, pipeline, media, insight, and governed execution.
Agent team & evaluations
Operating quality across people and AI.
Manage role coverage, evaluation sets, rubric performance, escalation behavior, and release readiness.
Active agents
8
2 voice · 6 workflow
Eval pass rate
96.1%
500 test cases
Open regressions
3
None critical
Priority view
What needs attention
Arabic code-switch
98% pass
Unsupported stock claim
100% blocked
Opt-out recognition
99.6%