← Selected work
Synthetic reconstruction

Rider Growth Control Plane

A public, interactive version of the planning logic I built at Deliveroo: connect marketing spend to applications, hires, activation, retained supply, hours and orders, then make the trade-offs visible by city.

Public data boundary

Every city, budget and performance number below is fictional. The workflow and metric relationships reflect my experience. Nothing here is Deliveroo data.

Are we creating enough demand?

Applications

7,626

Demand created across six fictional cities

Cost per applicant

£21

Blended acquisition efficiency

Cities on target

3 / 6

Application target only, before quality

Applications

7,626

Hires

2,910

38% conversion

Activated

2,102

72% conversion

Retained

1,707

81% of active

Decision signal

Larchester hit the application target but still create weak active supply.

The response is not automatically more media. Check screening, onboarding capacity, proposition fit and early rider experience before adding budget.

City decision table

Where should attention go next?

CityApplicationsCPATargetSignal
Northport1,890£2299%Review
Bellwick1,716£18107%Healthy
Eastmere1,115£2593%Review
Larchester1,109£20106%Review
Calder1,093£17109%Healthy
Westhaven704£2498%Review

1. Inputs

Spend, applications, hiring capacity, activation, retention, hours and orders.

2. Decision layer

Pacing, CPA, CPH, CPAR, cost per hour and cost per order, viewed by city.

3. Actions

Move budget, improve onboarding, change the proposition or stop low-return activity.

What this demonstrates

A dashboard is useful only when it changes a decision.

The original challenge was not simply reporting acquisition. It was connecting a fragmented growth funnel to the operational outcome the marketplace actually needed.

This reconstruction makes the key lesson tangible: an application target can look healthy while hire, activation or retention rates quietly undermine supply. The system has to follow the whole chain and point to the next intervention.

See more selected work