Data Analysis
Analyzing network performance to identify the most efficient growth levers for Savyu's table booking product.
My Role
Data Analyst
Tools Used
SQL, DBeaver
Key Finding
Active partners drive 4x the volume
The Context
Savyu's revenue model is based on the number of customers (pax) that come through each restaurant, not the number of bookings. A restaurant with many small reservations is worth less than one driving large parties.
With a growing network of partner restaurants, the question was: which cuisine types, which locations, and which partners are actually driving growth?
254
Bookings / week
876
Guests / week
3.4
Avg. party size
Sample week: Jan 12–18
Finding 1
The first discovery was a massive gap between all outlets and active outlets (those with at least 1 booking). The data showed that activation is the key lever.
Insight: Active partners drive 4x the median pax volume. The product works; the challenge is activation, not the product itself.
Finding 2
Not all food categories perform equally. I ranked every cuisine by daily pax per active outlet, the metric that directly ties to Savyu's revenue.
Top Performing Categories
Underperforming Categories
Finding 3
Beyond total volume, I looked at pax per outlet to find where each partner restaurant delivers the most value.
HCMC District 3
~28
pax / outlet
Fewer outlets, high density
Da Nang
~11
pax / outlet
Growing market, strong potential
HCMC District 1
~10
pax / outlet
Largest network, most volume
Insight: Some districts showed significantly higher efficiency per outlet despite having fewer partners. Expanding in high-efficiency areas offers better ROI than adding more partners in saturated locations.
The Decision
Savyu earns per guest (pax), not per booking. A restaurant with ten small bookings is worth less than one with three large parties. Optimizing for booking count would have pointed us at the wrong restaurants in the wrong locations.
Averaging across all partners — including dormant ones — would have buried the signal. Isolating active outlets (at least one booking) made the performance gap visible: the product works well when used. That reframes the problem from "grow the network" to "activate what's already there."
Total pax by district favors large markets by default. Normalizing to pax per outlet removes the size advantage and surfaces where each individual partner over-delivers — which is where acquisition and activation effort has the best return.
Recommendations
Key takeaway: In a per-customer revenue model, optimizing for booking volume is misleading. The real metric is pax per outlet, and that's driven by cuisine category and location, not just total reservation count.