Insights

Coupex tells you what to do next.

A rules-based engine watches campaigns, properties, vendors, and customer segments. Every recommendation carries a confidence score and ties to a real entity in the data. External AI providers slot in later behind the same output shape.

Seven types

Every recommendation slots into one of these.

The taxonomy is locked in the database enum. New surfaces add data, never new types. Operators learn the categories once and they hold.

executive_summary

Executive summary

Top-line portfolio shift. Where revenue moved, what drove it, who is winning.

Coastal portfolio leads Gauteng on redemption value this period.

campaign_warning

Campaign warning

A campaign has a risk flag that finance or approval should review before it scales.

Spring refresh has unusual fee override pending approval.

campaign_improvement

Campaign improvement

Campaign is running but underperforming its peers. Suggest a tactical change.

Winter 15% is plateauing. Try a tighter audience window.

underperforming_campaign

Underperforming campaign

Campaign has zero or minimal redemption. May need a resubmission or cap adjustment.

Holiday play has zero redemptions in week one of the window.

vendor_opportunity

Vendor opportunity

Vendor has unrealised expansion potential in a specific property or region.

Volt Hub redeems well in coastal properties but has no Gauteng campaign.

property_opportunity

Property opportunity

Property has untapped vendor capacity or an operational gap worth closing.

Sandbrook Court has eight open vendor slots in the food court.

segment_suggestion

Segment suggestion

Identifies a high-value customer cohort based on redemption patterns.

Weekday morning coffee shoppers redeem at 3x the evening rate.

Real recommendations

Twelve insights the seeded engine surfaces right now.

These are not promised features. Sign in to the platform and the same engine writes these against your portfolio. Confidence scores are computed, not handwritten.

Executive summary
95%

Coastal portfolio leads Gauteng on redemption value

Coastal certified R412 800. Gauteng certified R298 100. Driver: Pikolo Coffee.

Executive summary
92%

Coffee category is the top fee driver this quarter

Pikolo Coffee and Tableview Brew combined accrue 38% of all fees. Consider category-level reporting at the next portfolio review.

Campaign warning
88%

Hearth & Home spring refresh has unusual fee override

Pending approval. Finance review flagged the discount exceeds standard cap.

Campaign warning
84%

Sandbrook Tap mid-year proposes 60% off

Standard discount cap is 50%. Property finance review required before approval.

Campaign improvement
82%

Cape Threads winter 15% is plateauing

Issued count is healthy but redemption rate is below 8%. Try a tighter audience.

Campaign improvement
71%

Tableview Brew weekday rate flat at 4%

Discount cap may be too low. Consider raising to 25% off across weekday morning slots.

Underperforming
91%

Little Veld holiday play has zero redemptions

The campaign is still in draft. Submit for approval or tighten the discount cap.

Underperforming
83%

Glow Lab dinner mains week-one redemptions zero

Window opens Friday only. Add Thursday evenings to expand visibility.

Vendor opportunity
74%

Volt Hub has no campaign in Northcliff Plaza

Volt Hub redeems well in coastal properties. Propose a cross-group test.

Vendor opportunity
69%

Hearth & Home coverage uneven across properties

Strong performer at Lakehurst Crossing, absent from Wynbrook Centre. Cross-group test recommended.

Property opportunity
79%

Sandbrook Court has unused vendor capacity

Eight slots open in the food court inline strip. Targeted invitations would close the gap inside one quarter.

Segment suggestion
86%

Weekday morning shoppers redeem coffee at 3x evening rate

Tighter audience could lift Pikolo's redemption rate. Try a 06:00 to 10:00 window for the spring campaign.

How it works today

Rules-based. Deterministic. Auditable.

Each recommendation type has its own generator scanning the data for signal. Every row carries a computed confidence score, a clear reason, and an audit trail. The output stays stable when a future provider takes over.

01

Generators scan the database

Each type has a generator. Campaign-improvement looks at redemption rates against issued counts. Vendor-opportunity cross-references property coverage.

02

Confidence is computed

Each generator emits a Decimal confidence score from a deterministic formula over its input signals. Higher signal density gives higher confidence.

03

Rows persist and audit

Recommendations land in the queue with status open. Operators accept or dismiss. Every state change appends to the audit log.

What's next

The recommendation engine is provider-agnostic.

External AI providers slot in behind the same output shape. The operator workflow does not change.

A

Engine toggle

An account-level switch routes recommendations through an external provider when ready. Off by default.

B

Output shape locked

Every recommendation carries the same fields: type, headline, body, confidence, status, supporting metadata. The provider populates them. The dashboard does not change.

C

Provider-agnostic

Swap providers without touching the dashboard. Output stays stable. The audit log stays consistent.

See the insights in your portfolio.

Sign in to the seeded platform and walk through the AI insights page. Every card you see is generated by the engine running against the same data shape your live portfolio will produce.