Operations Analysis: Fuel Delivery

Client: Filld | Bay Area Start-up

Role: Independent Research Consultant

Research Question

Filld, a San Francisco Bay Area fuel delivery startup, wanted to credibly address city regulator questions about whether their service would help or hurt urban sustainability goals. Specifically, would fuel delivery service produces extra vehicle miles travelled (VMT), CO2 emissions, and traffic congestion? No prior research existed on this question.

My Role

I designed and conducted this study independently, with sole responsibility for study conception, data collection and analysis, and findings. The client provided an anonymized operational dataset; all analytical decisions were mine.

Approach

With no prior literature on mobile fuel delivery to draw from, I took a three-part approach. First, I synthesized existing behavioral research on driver refueling decisions to establish a baseline -- how much VMT do gas station trips actually generate, and when? Second, I analyzed the company's fleet telematics and billing data (covering 843 delivery stops and 998 customer fills across the Bay Area) to calculate operational metrics and compare them to gas station equivalents. Third, I built 2030 future scenarios using Census trend data and energy forecasts to model what happens to both systems at scale.

A key methodological challenge was that standard ITE trip generation methods produced VMT estimates that expert reviewers judged too high. I developed an alternative methodology using the California Household Travel Survey's activity variable, filtering to single-activity refueling trips under ten minutes. This produced more defensible estimates and is the approach I would use again.

What We Learned About User Behavior

Gas station trips are remarkably efficient. Between 84% and 93% are pass-by trips that generate no additional VMT because drivers simply stop on the way to somewhere else. Estimated VMT per fill at a gas station was just 1.4 miles, a tight target for any delivery service to match.

Delivery operations are off-peak and during the night, which is a traffic benefit. This finding about the timing of fueling trips was the sharpest contrast between personal trips and home delivery: 48% to 58% of gas station trips happen during peak congestion hours, while only 9% of fuel delivery departures did.

Fleet and residential customers are completely different problems. Residential customers averaged 2.9 miles VMT per fill. Fleet customers averaged 0.3 miles -- an order of magnitude better -- because a single stop served 12.7 vehicles on average.

Decisions This Work Informed

The analysis gave the client a credible answer to a question that regulators and city partners were starting to ask.

Fuel delivery reduced CO2 emissions, on net. Under current operations, total annual CO2 emissions from the service (76 metric tons) were lower than a single typical gas station (97 metric tons). At scale, with optimized routing, the service could capture up to 5% of Bay Area fuel market share without adding net VMT. The path to that outcome required roughly halving VMT per residential fill -- a specific operational target, not just a directional finding.

It also identified strategic priorities to improve operational efficiency and ensure equitable access.

Grow fleet customers first. Each fleet delivery removed roughly 26 gas station round-trips from the road network. At current Bay Area volumes that translated to 15,000 VMT and 6.1 metric tons of CO2 reduced annually.

Inclusive payment and scheduling is a design requirement. As gas stations disappear from urban cores, mobile delivery could fill the gap. However, customers need smartphones, bank accounts, and the means to subscribe. Those without access could face both higher prices and longer trips to fewer stations.