Promo codes for Al Fursan, before Saudi National Day arrives

02 · the problem

Two problems. The second is the one people miss.

The first is straightforward. The second changes how every campaign has to be built, priced, and reported.

Problem one: no promo engine exists

There is no code field, no condition engine, no redemption record. It gets built. That part is not interesting.

Problem two: a redemption is not final at checkout

On Al Fursan, loyalty miles are credited 30 days after an order is accepted, and they are cancelled if the order is cancelled or returned inside that window. A promo redemption inherits the same clock.

Thirty days, then a fork

Every campaign number is provisional until this fork resolves.

Every hard open question in this document hangs off that one fact. Four consequences follow, and none of them are optional to handle.

• A campaign cannot be reported as successful on the day it ends. It has not settled.
• "How many redemptions did we get" has two different correct answers at any given moment, Provisional and Confirmed.
• An influencer paid on provisional numbers can be paid for revenue that later reverses.
• Whether miles are earned on the pre-discount basket or on the amount actually paid changes the true cost of every campaign, and it is still open. See decision 3 in the next tab.

Two correct answers, mid-window

Figures are illustrative, not from a live campaign.

The latency budget

Third-party hop inside the budget. This hop is now the settled design, not an option: tier is always resolved live. Phase 1 never makes this call at all.

Why a nightly batch was not the cheaper answer

The obvious way to avoid a call inside the checkout path is to pull the membership base overnight and read it locally. That option was examined on 3 August 2026 and it fails for a reason that has nothing to do with privacy, which is worth stating because the privacy argument tends to crowd it out.

Merit only learns that a member exists when that member signs in to the store. If the programme holds a million members and only a fraction have ever signed in, Merit knows that fraction and nothing about the rest. A batch built on what Merit already holds would therefore answer for a minority of the base while appearing to answer for all of it. A batch built on the whole programme would require Al Fursan to hand over its full member list, which is a larger data-sharing decision than the one this initiative needs. The live lookup sidesteps the choice: the question is asked about exactly one member, at the moment that member is at checkout.

What the decision costs at campaign creation

Promo codes are a creation feature as much as a redemption one, and the two halves are not affected equally. Redemption gets better, because the tier read is always current and an upgrade applies to the very next basket. Creation gives something up, and it should be visible here rather than discovered during Phase 2.

Because Merit holds no copy of the membership base, the campaign creation screen cannot size an audience. Building a Gold tier offer, marketing will not see how many Gold members exist, what share of the base they represent, or how far a given set of restrictions would reach. Only Al Fursan holds the data that answers those questions. The accepted approach is the simpler one: define the code and its restrictions without member data in view, and let eligibility resolve per member at redemption. Audience sizing, if Al Fursan wants it, is a separate conversation about what data Al Fursan is willing to share, and it is not a dependency of Phase 2.

03 · the open decisions

Five business decisions, not an architecture choice

For this initiative the architecture is not contested. The business definitions are. These five items block Phase 1 from being scoped.

D5, D1, D2, and D3 all feed the Phase 1 scope freeze. D4 depends on D3 but is needed before contract, not before scoping.

Decision 1: what does "new customer" mean

Gating item 6. Owner: Al Fursan. Blocks CS-01.

Decision 2: acquisition offer, or one-per-member cap

Gating item 7. Owner: Al Fursan. Blocks CS-02.

Does "first purchase discount" mean a genuinely new customer, or the first time that member uses that code? These are different products. An acquisition offer is not the same thing as a one-per-member cap. The workshop transcript supports both readings, which is why it must be settled explicitly.

Decision 3: miles on the pre-discount basket, or the amount paid

Gating item 8. Owner: Al Fursan and Merit commercial. Blocks ML-01.

Decision 4: who absorbs the discount

Gating item 3. Owner: Thamer Bamieh and Al Fursan.

The PRD currently states Merit absorbs the discount from its own share. That was written for Merit-owned marketplaces and has not been agreed for a tenant.

This cannot be priced until decision 3 lands, because the true cost of the discount depends on which basket value earns miles. Not a Phase 1 scoping blocker, but needed before contract.

Decision 5: how a household counts

Gating item 2. Owner: Al Fursan. Blocks RW-01.

Where a household holds one membership with family members underneath it, the per member limit can apply once to the household, or once to each family member. The rolling per member limit is Phase 1 scope, so this cannot be deferred. It has a second effect: the same answer decides whether a family member still counts as a new customer once the household has already purchased, which feeds straight back into decision 1.

Non-blocking, listed so nothing looks hidden

Mirrors BRD section 8. These do not block Phase 1 scoping.

04 · the scope

Six epics, following the PRD's own structure

What Al Fursan is actually buying, grouped the way the PRD groups it: a user story, the requirement IDs it carries, which phase it ships in, who owns it, and the trade-off where one exists.

The condition ladder

Every code drops through these gates from the top. The first gate it fails decides the message the shopper sees, and the gates below it never run. For Saudi National Day only the Phase 1 gates are switched on. The greyed gates arrive in a later phase.

1. Code exists and is active Fails when the code is unknown, or it has been paused or ended. Phase 1
2. Inside its validity window Fails when the code has expired or has not started yet. Phase 1
3. Belongs to this store Fails when the code was made for a different storefront. TN-01 Phase 1
4. Reserved member matches, where the code is member-bound Fails when a bound code is presented by anyone else. VC-02 Phase 3
5. Total usage cap not reached Fails when the code has been fully used up across all shoppers. Phase 1
6. Within the shopper's per-member limit Fails when the shopper already used this code inside the rolling window. RW-01RW-02 Phase 1
7. The shopper's tier has a discount on this code Fails when the code sets no discount for the shopper's loyalty tier. TR-03 Phase 2
8. Matches the new or existing customer setting Fails when a new-only or existing-only code meets the wrong shopper. This gate is decisions 1 and 2. CS-01CS-02CS-03 Phase 1
9. Marketplace scope matches Fails when the code is not valid in this marketplace. A baseline gate, not exercised by a single-store tenant. Phase 1
10. Cart meets the minimum value Fails when the basket is below the code's minimum. Phase 1
11. At least one item is in scope Fails when nothing in the cart qualifies for the code's category or product scope. Phase 2

The phase map

The eight systems, for engineering readers

05 · the build

One member, one code, thirty days to know

A single Al Fursan member applies SNDAY25 to a SAR 400 basket. These are illustrative round figures, roughly a median store basket, and the member ID shown is a placeholder in the real format. It is not a real member.

When something breaks

Pick a scenario. The walkthrough replays on the same diagram with that component in the broken state, and the narration changes to the degraded path.

The three phases

Capability accumulates. Each phase keeps everything the phase before it shipped and adds the next layer. Phase 1 must land before Saudi National Day. Step through to see what each phase unlocks and what it still cannot do.