Returning customers
Know how many vehicles come back, and how often. Today that’s invisible unless someone uses a loyalty card.
Oil Vision
Forecourt Intelligence
An idea for Jio-bp Kerala
Every vehicle that visits tells you something. Today we don’t write it down.
A simple idea: at each station, quietly count the vehicles that come in — type, time, and whether they’ve been here before. No cameras on people. No customer records. Just an honest picture of how a forecourt is actually used.
Concept only · Synthetic data · Not connected to any live camera, customer or government system
From individual visit
To network intelligenceHow it works
Four steps. Nothing complicated.
01
A vehicle arrives
A car or two-wheeler pulls into the station.
02
It’s counted, not identified
Time, vehicle type, and whether this vehicle has visited before. Nothing personal.
03
It becomes a number
The station sees patterns: busy hours, vehicle mix, how many customers return.
04
Someone acts on it
Better staffing, better stock, better offers, better planning.
What it could do
Four things become measurable.
This isn’t about watching customers. It’s about knowing how a station is actually used — something we currently guess at.
Day-to-day operations
See the real busy hours and vehicle mix. Staff and stock the forecourt to match.
Planning across stations
Compare stations properly. Spot which ones are stretched and which have room.
Non-fuel sales
Understand who returns regularly, and test offers on fuel, convenience and services against real patterns.
Rough numbers
A rough sense of scale. Move the sliders and argue with it.
None of this is a forecast. It’s a small model built on public fuel-retail figures, so the scale of the idea can be sanity-checked rather than taken on trust. Change the assumptions — the numbers change with them.
For context: one good station today
A station selling around 330 kL of fuel a month sees roughly 650 vehicles a day. Most of its margin comes from volume, not per-litre. Which is why knowing the busy hours — and who comes back — matters more than it sounds.
Visits a month
19.5 lakhIllustrativeacross the stations selectedReturning vehicles spotted
6.8 lakhIllustrativeanonymous, no identity attachedOpted-in customers reachable
2.9 lakhIllustrativeonly where a customer has agreedReturning customers
Repeat visits seen · network / month
6.8 lakhIllustrative
A measurable base of regulars, where today there’s only a guess.
Station operations
Busiest-hour vehicles · site
78Illustrative
Peak pressure becomes visible, so staffing and lane flow can follow it.
Planning across stations
Sites running above comfortable peak
No station is stretched at this settingIllustrative
Which stations are stretched, and which have room.
Non-fuel sales
Extra non-fuel transactions · site / month
293Hypothetical
If regulars were engaged better. Arithmetic, not a forecast — the rupee version sits in the disclosure.
How these numbers are calculated — and where they came from
- Fuel volume per station = vehicles/day × 30 days × ~17 L blended fill — the bridge that makes the 330 kL/month anchor hold at ~650 vehicles/day
- Fuel revenue = volume × blended ~₹106/L (Kerala petrol ₹115.49/L · diesel ~₹98/L, Sept 2026 price trackers) · dealer margin = volume × blended ~₹2.8/L (OMC commission ≈ ₹3,100/kL petrol · ₹2,300/kL diesel, plus a price-linked component)
- At the default settings that gives one station ≈ ₹3.51 crore/month fuel revenue (pass-through), ≈ ₹9.3 lakh/month dealer margin, and ≈ ₹1.9 lakh/month existing non-fuel sales
- Visits a month = stations × vehicles per station × 30 days
- Repeat visits seen = monthly visits × returning-vehicles slider (anonymous, no identity attached)
- Opted-in customers reachable = monthly visits × loyalty opt-in share
- Busiest-hour vehicles per station = vehicles/day × 12% assumed busiest-hour share (twin morning/evening peaks)
- Sites above comfortable peak = stations whose busiest-hour vehicles exceed ~90 — roughly 6–8 dispensing positions serving ~12–15 vehicles/hour each; the improvement slider eases the peak before the same test
- Campaign response rate = 5% base + operational-improvement slider
- Extra non-fuel transactions per station = monthly visits × loyalty opt-in × campaign response
- Indicative rupee uplift per station = extra transactions × ₹200 assumed basket (typical ₹150–350)
Grounding ballparks · approximate public ranges, illustrative. National average outlet throughput works out to roughly 100–150 kL a month (PPAC fuel-sales data ÷ ~87,000–90,000 retail outlets), so 330 kL marks a well-performing site · dealer commission runs ≈ ₹3,100/kL on petrol and ₹2,300/kL on diesel plus a price-linked component — about ₹3–4/L and ₹2–3/L (Business Today, Mar 2026) · Kerala petrol ≈ ₹115.49/L, diesel ≈ ₹98/L (Goodreturns tracker, 12 Sep 2026) · non-fuel conversion is single-digit in India versus 20–50% at mature international forecourts (convenience-retail industry reporting) · mature fuel-loyalty programmes link 25–40% of transactions to a loyalty identifier, Indian programmes often 10–20% (industry reporting) · typical non-fuel basket ₹150–350.
Sources
- PPAC, Ministry of Petroleum & Natural Gas — monthly fuel-sales and retail-outlet data (ppac.gov.in); per-outlet throughput derived from public sales ÷ outlet counts
- Business Today — “Pump commission rules explained: how much petrol pump owners earn per litre in India”, 18 Mar 2026
- Goodreturns — state fuel-price tracker, Kerala, 12 Sep 2026
- Convenience-retail and loyalty participation benchmarks — industry reporting; to be replaced by Jio-bp internal data during the pilot
Assumed constants: blended fill ~17 L · blended fuel price ~₹106/L · dealer margin ~₹2.8/L · busiest-hour share 12% · comfort level ~90 vehicles per busiest hour · baseline non-fuel conversion 5% · response base 5% · basket ₹200. Benchmarks are public ranges — the pilot would replace them with measured values.
What we’re asking for
Start with one station. Three if you want the full picture.
No funding, no procurement, no commitment beyond the pilot. Just permission to set this up at a site or two and see whether the numbers hold up. If they don’t, that’s a useful answer too.
Three different kinds of site would tell us the most:
Worst case, we learn what doesn’t work, at one station, at no cost.
A busy urban station
Peak demand, queues, regular customers
A highway site
Passing traffic, longer stops, different mix
A two-wheeler-heavy site
The hardest case, and the most common one
Built to be defensible. Vehicles are counted, not identified. Anything linked to a person requires that person to opt in, separately. No customer data is sold, shared or resold — that’s a design decision, not a policy statement.