Shell Hong Kong · EV charging · For discussion
Built entirely from public information — Shell’s own site locator and Recharge pages, the EPD charger register, 中電 CLP and 港燈 HK Electric published tariffs, and a listed competitor’s audited accounts. Sourced figures carry their source; modelled figures are labelled as estimates. Nothing is drawn from inside Shell.
Start here
The whole finding
Below roughly 7.5% utilisation, the demand charge alone eats the entire margin on a kWh.
Not a saving I am claiming. A threshold you can test. Twenty-one of your largest charging sites sit in 中電 CLP territory, and 中電 CLP’s Bulk tariff bills on the on-peak maximum demand recorded in the month, plus any off-peak excess, subject to a 100 kVA minimum — a fixed cost, recovered over however many units happen to be sold.
The arithmetic, in full
Demand cost per kWh = rate ÷ (730 hours × load factor)
Load factor here means units sold divided by (billed maximum demand × 730 hours). Not against the equipment installed — against the peak kVA that actually appears on the bill. Those are different numbers, and only the second one is charged for.
That is the entire model. Put in your own demand rate and your own load factor and it answers itself in about sixty seconds. Nothing in it requires anything I could not read off a published tariff table.
Read one row at a time. On my assumptions the spread on a unit is about HK$1.36 — Shell’s published price less electricity at the EV tariff and an assumed site cost. Your real figure will differ. The question is simply whether the demand charge on that site costs more or less than that.
| How busy the site is | Demand charge costs, per unit sold |
Shell earns per unit sold |
Left over |
|---|---|---|---|
| Nearly empty — 2% | HK$5.13 | HK$1.36 | −HK$3.77 |
| 5.5% — my estimate for the network | HK$1.87 | HK$1.36 | −HK$0.51 |
| Quiet — 6% | HK$1.71 | HK$1.36 | −HK$0.35 |
| The turning point — 7.5% | HK$1.36 | HK$1.36 | HK$0.00 |
| Busy — 15% | HK$0.68 | HK$1.36 | +HK$0.68 |
| Very busy — 40% | HK$0.26 | HK$1.36 | +HK$1.11 |
On CLP’s Bulk tariff at HK$74.90 per kVA a month. Above the turning point the demand charge is covered by the spread; below it, it is not — before any other cost is counted. That is a statement about one line on the bill, not about whether a site is profitable overall. Hong Kong Island sites buy from 港燈 HK Electric at HK$48.30 per kVA (low voltage; HK$47.30 high voltage), where the turning point is 4.9% — but only 2 of the 23 large sites are there.
The 7.5% above answers “which tariff should a site I already own be on?” The equipment is already bought, so its cost is sunk and does not belong in that comparison.
A different question — “should we build another bay?” — has to carry the capital too. At HK$317,520 a bay over eight years at 8.3%, plus O&M, that adds the equivalent of about HK$46 per kVA a month on top of the tariff, and the threshold moves to roughly 10.4%.
Both are right. I am only claiming the first. The second is why I would not lead with building more chargers.
Why the cost side, not the price
On a quiet site this one line is bigger than the entire margin
At 5.5% the demand charge costs HK$1.87 a unit against HK$1.36 of margin. It is close, and that is the point — the estate sits right on the line, so which side a given site falls is a question of fact, not opinion. No pricing change reaches a number that size, and raising the price is the one move 中石化 Sinopec has already made impossible at HK$1.68.
What kills it
If those sites are already on the EV tariff
The EV tariff has no demand charge at all. If that is where the estate sits, this idea is worth nothing and I would rather find out over lunch than in week two.
The 5.5% load factor is an outsider’s estimate, not an observation: a public revenue anchor spread evenly across the estate, then set against an assumed 70% coincidence between installed power and billed peak. Real sites will sit either side of it, and the busy ones well above. The 7.5% threshold does not depend on it at all — that comes only from the published tariff and the spread. Everything else here comes from Shell’s own published locator, the EPD charger register and the two utilities’ published tariffs.
Act 1
Two estates that never touch — and a promise Shell has already published.
Two estates, no bridge between them. The forecourts carry the retail. The charging sites carry the customers. They are almost entirely different places.
43 forecourts
Shell's own sites. Shell controls the environment.
1 has EV charging.
70 EV sites
Mostly bays rented inside somebody else's car park.
The dwell time happens on someone else's property.
So the opportunity is not “sell them something while they wait” — at a mall site the mall already does that, and keeps it. It is to give the charging customer a reason to visit a Shell forecourt at all.
A reward is only worth something if the driver can reach the place that honours it. So: how far is each EV site from the nearest Shell location that could actually hand over the reward — a Select shop, a car wash, or a service bay? 67 of the 70 sites carry usable coordinates in Shell’s own locator; the other three are excluded rather than guessed:
| Redeem at | Sites | Median | Within 2 km | Worst |
|---|---|---|---|---|
| Select shop | 29 | 1.3 km | 70% | 7.4 km |
| Car wash | 16 | 1.8 km | 52% | 9.0 km |
| Service bay | 13 | 2.7 km | 34% | 12.3 km |
That table decides the design. A reward redeemed at a service bay is an errand — only a third of charging customers have one within 2 km. A reward redeemed at a Select shop or car wash is at least reachable for most. Whether anyone would actually detour for it is the thing distance data cannot tell you, and it is why this needs a pilot rather than a rollout.
Largest fast-charging network in Hong Kong. This is an asset to exploit, not a gap to fix.
Public location data show almost no overlap between the two estates. It cannot show how often a charging customer later visits a forecourt — only that the opportunity is not in front of them.
The shops, washes and bays exist. They are simply not where the EV customers are.
Act 2
A state-owned competitor is selling near cost. Price is not the way out.
中石化 Sinopec’s PIT is the Hong Kong charging arm of a Chinese state-owned oil major. All figures below are from their own listings and public tender records.
中石化 PIT Ultra, 火炭 Fo Tan
HK$1.68/kWh
On 500 kW equipment. The lowest per-kWh price in Hong Kong. Electricity alone costs HK$1.500 on the 中電 CLP EV tariff — so that is an energy-price spread of about HK$0.18/kWh, roughly 11% of revenue, before demand charges, land, equipment and operating cost.
On the published tariff assumed here that sits close to energy cost. Sinopec’s actual unit economics and pricing rationale are not public.
| Network | Sites | Points | Fast | HK$/kWh |
|---|---|---|---|---|
| Shell Recharge | ~70 | 693 | 188 | 3.35–4.20 |
| Sinopec PIT | 21 | 284 | 74 | 1.68–3.70 |
| CrazyCharge | 8 | 31 | 31 | 2.40–3.50 |
| Speedy Base | 1 | 80 | 80 | not published |
Sinopec won a Government land tender at 山尾街, Sha Tin Town Lot 646, Fo Tan — about 370 m² on a 12-year lease, specifically for EV charging. That is a different cost structure, not just a different price.
| Mall or car-park site | Government land tender | |
|---|---|---|
| Landlord share | ~15% of revenue, ongoing | none |
| Tenure | at the landlord's discretion | 12 years |
| Electricity | often shared — may attract a demand charge | dedicated EV tariff |
| Rival on site | the mall can add one | exclusive |
A 15% landlord share on a HK$3.60 sale is HK$0.54 a unit — three times Sinopec's entire gross margin at Fo Tan.
Shell is 114% more expensive per unit than PIT Ultra. That gap cannot be closed, and it should not be attempted — the other side is not trying to make money on the electron. But note the qualification: Sinopec runs 50 petrol stations in Hong Kong to Shell's 43. They are an oil major with a forecourt estate too, not only a charging operator. The advantage to defend is what is ON the forecourt, and how quickly it is put to work.
A price floor above most competitors' ceiling only works if something else is being sold.
If price were rationing demand the premium bays would be the busiest. They are the freest.
All 18 districts are labelled. All five fuel networks sit within ±1.5% of the Hong Kong average — that is noise, not a pattern. Do not use it to argue siting.
A 13-page deck, EV Charging Analysis of Hong Kong, was prepared for this market on roughly 16 days of February 2025 session data. Read closely, three of its conclusions do not survive its own charts:
Different data cuts may explain it. Either way I treat those conclusions as unverified and have not relied on them here.
Raised as a question, never as an accusation — the deck may have been drafted against a different cut of the data.
Act 3
Stop discounting. Start paying selectively, for behaviour worth more than it costs.
Speedy Base gives 15% back on top-up. CrazyCharge gives 30%. On our estimates Shell breaks even at about 7%. Matching either is arithmetic Shell loses.
| Same generosity, as the customer feels it | Cost / yr | Net / yr |
|---|---|---|
| Cash bonus, 15% (matching Speedy Base) | HK$6.51M | −HK$4.84M |
| Partner services, same 15% felt | HK$2.01M | −HK$0.01M |
A dollar of perceived reward costs Shell about 31 cents when it is redeemed against a partner service, and 100 cents when it is cash.
Speedy Base and CrazyCharge have no forecourts, no service bays and no car washes. Their only instrument is a deeper discount, which costs full face value and tells them nothing about the vehicle. Shell has 43 forecourts with 28 shops, 16 car washes and 13 service bays — a redemption catalogue that already exists.
One honest qualification. 中石化 Sinopec operates 50 petrol stations in Hong Kong to Shell's 43 (Consumer Council registry). What retail sits on them is not published, so this cannot be presented as an advantage no oil major has — only as one the charging startups do not have. Shell will know the answer on Sinopec better than any outside analysis can.
This is the difference between a promotion and a mechanism. A price cut is paid to everybody for behaviour they were going to exhibit anyway. A points scheme lets Shell pay selectively, for the specific behaviour that makes the bay more valuable.
The whole scheme in one line
Earn 1 point per HK$1 charged. A point is worth about 10 cents to the customer and costs Shell about 3.1 cents — because it is redeemed against a service, not paid as cash.
That 3.2× gap between felt value and delivered cost is the entire economic engine. Everything below is built on it.
The base rate is uninteresting. The multipliers are the product: they buy the two behaviours that are worth more to Shell than the points cost.
Standard session
The baseline. Costs nothing extra and buys nothing.
Ends at or below 80% charge
Buys the bay back. The last 20% of a battery takes longer than the first 70%.
Starts between 00:00 and 07:00
Buys a session in a bay that was empty, at no extra fixed cost.
The overnight multiplier costs 6.2 cents in the dollar. An overnight session that would not otherwise have happened contributes its full margin, because the demand charge and the site licence are already paid whatever the bay does.
Shell is buying a marginal contribution for six cents. That is the trade a flat discount cannot express, because a flat discount pays the peak customer the same as the 3am customer.
| Reward | Points | Feels like | Costs Shell | Returns data |
|---|---|---|---|---|
| Car wash — 16 Shell sites | 1,200 | — | ~HK$36 | HK$1,200 |
| Vehicle health check (途虎養車) | 4,000 | ✓ | ~HK$125* | HK$4,000 |
| 四輪定位 wheel alignment | 6,000 | ✓ | ~HK$225* | HK$6,000 |
| NTI insurance excess waiver | 10,000 | ✓ | ~HK$145 | HK$10,000 |
| Charging credit | 1,000 | — | HK$100 | HK$1,000 |
Charging credit is deliberately the worst item on the list. It costs full face value and
returns nothing. It exists so nobody can say the scheme has no cash-equivalent exit
— not because Shell wants it chosen.
* Not quoted prices. No approach has been made to 途虎養車 or
NTI, and a delivered cost would have to carry staffing, booking and no-shows, site
capacity, systems integration, indemnity and procurement overhead — plus whatever
the partner wants in return. An order of magnitude to test, not a price. The car wash is
the exception: Shell already owns and operates those 16 sites.
Breakage
3.5%
Points issued and never redeemed, taken to income. Deliberately conservative — published airline and retail schemes run far higher. Over-claiming breakage is the classic way a loyalty P&L flatters itself.
Expiry
24 months
Long enough to reach the health-check threshold on ordinary usage, short enough to bound the liability. Every unexpired point is a balance-sheet obligation.
Liability
3.1c
Carried per unredeemed point — the delivered cost, not the HK$0.10 the customer perceives. Booking the perceived value would overstate the provision threefold.
Sinopec sells at HK$1.68 when electricity alone costs HK$1.50. Shell cannot follow that price and should not try. But a points scheme moves the fight to a field where the cheapest electron does not win: the reward is worth HK$0.10 to the driver and costs Shell HK$0.031, and it is redeemable only against services a pure charging operator does not have.
Matching Sinopec on price costs a dollar to give a dollar. This gives away ten cents of felt value for three.
Act 4
The physics of a charge curve, and the one thing a charger can never tell you.
Above roughly 80% the battery management system holds voltage and cuts current to avoid lithium plating on the anode. Power falls with current. The charger cannot override it — it is the car's decision.
| Hyundai Kona 64 kWh on a 120 kW bay | kWh | Minutes | Bay-min per kWh |
|---|---|---|---|
| 10 → 80% | 44.8 | 42.0 | 0.94 |
| 80 → 90% | 6.4 | 15.1 | 2.36 |
| 90 → 100% | 6.4 | 33.2 | 5.19 |
The last 20% of the battery takes 48 minutes — longer than the first 70% takes. Charging every kWh at one price means the driver who stops at 80% subsidises the driver who fills to 100%.
| Vehicle | Usable | 80–90% | at 22 kW | at 120 kW | at 350 kW | tail cost |
|---|---|---|---|---|---|---|
| Hyundai Kona EV | 64.0 | 6.4 kWh | 17.5m | 15.1m | 15.1m | 5.5× |
| Tesla Model 3 LR | 75.0 | 7.5 kWh | 20.5m | 5.5m | 5.5m | 3.2× |
| Hyundai Ioniq 5 | 77.4 | 7.7 kWh | 21.1m | 6.0m | 6.0m | 3.4× |
| BYD Seal | 82.5 | 8.2 kWh | 22.5m | 10.0m | 10.0m | 4.9× |
| BYD Atto 3 | 60.0 | 6.0 kWh | 16.4m | 12.4m | 12.4m | 5.5× |
| MG ZS EV | 49.0 | 4.9 kWh | 13.4m | 9.7m | 9.7m | 5.5× |
| Nissan Leaf 40 | 39.0 | 3.9 kWh | 15.4m | 15.4m | 15.4m | 5.5× |
Look at the 120 kW and 350 kW columns: identical for every vehicle. Above about 30 kW the car limits the rate, not the charger. The 80–90% band is 6–8 kWh whatever the car, and it takes 5–22 minutes. Paying for 350 kW hardware to serve that band buys nothing.
The new app bills by actual kWh — but the same price for every kWh, regardless of how much bay-time it consumes. .
0 – 80% charge
HK$2.90
Below the market median of HK$3.20
80 – 90%
HK$4.20
2.4× the bay-time per kWh
90 – 100%
HK$6.00
5.5× the bay-time per kWh
A driver charging 10→80% pays an effective HK$2.90/kWh — less than today. The considerate driver is rewarded; the bay-blocker pays for the bay. Apply it only at congested sites: at a quiet mall the tail costs nothing.
Moving the 6.4 kWh in the 80–90% band takes 17.5 minutes at 22 kW and 15.1 minutes at 120 kW. Only 2.4 minutes apart, because above about 30 kW the car is the constraint, not the charger. Buying 350 kW hardware to serve tapering cars buys almost nothing.
Physics, not policy. The bay is occupied longest exactly when it is earning least.
A pricing ladder is a throughput scheme. Judged per bay-hour, the cheap floor gives away more than it fills.
Whatever single rate is chosen, it is losing one of them.
Every CCS2 session exchanges data that identifies the vehicle and describes its behaviour. Three levels of confidence, which should never be presented as one.
Direct
Derived
Inferred
One technical caution. The “time remaining” field is time to the vehicle's own full-charge target, which is optional and not always 100%. A car may report a target of 90%. Treating it as time-to-100% produces wrong estimates for some vehicles.
CCS2 never reports the exact model or the true battery capacity. A 3-minute OBD read on a phone does. The trick is not to ask for it.
The reward the customer wants — a free vehicle health check — is the data collection.
Driver redeems points for a free health check
↓
3-minute OBD scan at a Shell forecourt or partner
↓
Customer receives battery health, fault codes, service due
Shell receives exact model, true battery capacity, odometer
On our estimates roughly 55% of redemptions return a read — around 8,900 vehicles a year identified. Each one improves the pricing and siting decisions for every session after it.
Act 5
The largest lever is not the loyalty scheme. It is which electricity tariff a site sits on.
The Island is served by 港燈 HK Electric, the rest of Hong Kong by 中電 CLP. Their EV tariffs are not the same. Both are published.
| Per kWh | CLP · Kowloon & NT | HK Electric · Island |
|---|---|---|
| EV tariff energy charge | 110.6c | 142.4c |
| Fuel adjustment | 39.4c | 57.3c |
| All-in cost | HK$1.500 | HK$1.997 |
On these published rates Island electricity costs up to 33% more, while Shell's Island price premium appears to be around 12%. If that holds, the Island price sits roughly 16% below the premium needed to preserve the same markup, across 15 Island sites.
Caveat, stated up front. The 33% uses HK Electric's top consumption block against CLP's flat EV rate. Blended across blocks the real gap may be nearer 18%. And Shell's 12% premium comes from the tariff published before the December change. Both need checking against actual site bills — which is exactly the kind of thing that is invisible from outside.
HK Electric's published fuel clause charge, most recent months, in cents per unit:
35.4 → 35.4 → 34.0 → 30.4 → 26.0 → 31.3 → 41.9 → 57.3
26%
of the CLP EV tariff is the fuel adjustment — revised monthly, not fixed.
HK$2.8M
a year of contribution lost if that swing repeats on CLP — more than any improvement discussed here.
2026
CrazyCharge has guaranteed its prices for the whole year while its fuel cost floats. That is risk they are carrying, not strength.
479 points sit at 7 kW. At Shell's own price that bay tops out at HK$23 an hour, fully occupied.
The car limits what it can absorb, while CLP’s Bulk demand charge is billed on recorded maximum kVA — so higher-rated equipment raises exposure only insofar as it raises measured peak load.
Where the two curves cross the dashed line is where the demand charge has consumed the entire margin on a unit sold.
Act 6
The ask
Two weeks. HK$60,000 fixed.
No contingency.
And it goes through legal and whatever process it lands in. If it dies there, it deserved to.
What I need
Your electricity bills and interval demand for the 21 large sites in 中電 CLP territory. No customer data. No identifiers, nothing touching the PDPO. Your meter, your money.
What you get
One page per site: recorded peak against installed, what the coincidence cost last year, what sequencing would save with no session finishing later than it does today, and a flag for whether the saving is yours or the landlord’s.
Sites where the answer is zero go on the list too — a clean negative is a result, and it is the one I would report first.
Or, honestly
I would rather have the job than write about it.
Either a two-week HK$60,000 fixed engagement through Shell legal and procurement, or a conversation about a salaried role. I am not proposing unpaid work.
This question deserves a straight answer, because the obvious one is that you don’t need me.
I am the customer
Daily EV driver since 2023. I run out to the SENT landfill regularly and drive across Hong Kong for meetings, so I charge on the public network in the same conditions your customers do. Every assumption on this site is one I have tested with my own money at a real bay.
I have built the other side
Eight solar-powered sensor poles across the haul roads of the SENT landfill for Veolia — off-grid batteries sized for seven days, a UPS on the compliance-critical circuit, IP65 enclosures, delivered to the operator’s access and PPE rules alongside live waste operations. I also built g63m, my own Hong Kong vehicle-data system. A charging site is the same class of problem: power, enclosures, connectivity, uptime, on somebody else’s operating site.
And shipped it fast
Purchase order came in late November. Everything was specified and ordered inside a month. The client’s own programme then slipped into February, so we went on site mid-February and had all eight poles installed and handed over before the end of March 2025.
About six weeks on a live site, and roughly four months from order to handover including a client-side delay. It is still under paid annual maintenance.
And it changed something
The client reported an 11.6% fall in speeding violations after deployment. Instrument a site, change a behaviour, prove it with a number. That is the same shape as what I am proposing here, on a different asset.
Before that, four years as an HKT field engineer on live sites — the airport, the Exchange, the Police Force. A PolyU data science degree, an HKU MBA finishing now, and in 2025 I built the Hong Kong charging-pattern analysis for EV-BOY, who run the territory’s real-time charging map. So I have seen roughly what this data looks like at scale, from the outside.
| An IT analyst | Can query anything you ask for. The gap here isn’t querying — it’s knowing that a charge curve flattening at 80% is a commercial fact, and that a demand charge is fixed. That comes from standing on sites and driving the cars. |
| Someone from CLP | Would know the tariff far better than I ever will. But CLP sells the electricity, and this finding is about buying less of it more carefully. That is structurally the other side of the table. |
| A charging operator | Knows charging, and has no forecourt, no shop, no wash and no service bay — so the only instrument they have is price. The whole argument on this site is about the things a pure charging business doesn’t own. |
| A consultant | Delivers a snapshot that is true the day it lands and stale a quarter later. This data drifts every month. |
Shell’s Recharge page publishes a support number, +852 5506 2265. The same number appears on Halo Energy’s site. And the Shell Recharge user guide is hosted at halo-e.co, footed “Copyright © 2023 Halo Energy Limited”.
Read plainly, that suggests the app, the wallet, the payment flow, the transaction records and the existing occupancy fee sit on a partner platform, not Shell’s own. I may be reading too much into two public pages, and you will know.
| Idea | Who would have to agree |
|---|---|
| Loyalty points, redemption catalogue | The platform partner would have to build it |
| Session-level measurement | Likely the platform partner holds the data |
| Idle / occupancy fees | Already live, already administered there |
| The electricity tariff on the site supply | Not the app. Plausibly Shell’s alone. |
That is the real reason to start with the bill. Every other idea on this page needs somebody else to say yes first.
In January you opened a Shell Recharge station at CUHK — two 120 kW units for the campus bus fleet, after eight years supplying that fleet with fuel. A depot runs at high, predictable utilisation, so it sits well above the turning point and the Bulk tariff is the right choice there.
My question is only about the quiet bays, where the same tariff behaves in the opposite way.
Source: Shell Hong Kong press release, 20 January 2026.
Not a resourcing proposal, and deliberately not costed here. The tariff question alone would not justify standing capacity, and I wouldn’t pretend otherwise. What a second stage would test is whether there is a continuing pipeline of decisions worth measuring — utilisation, pricing, retail attachment, site selection — or whether this was one good question and then nothing. That is only worth discussing if the first answer pays for itself.
It would need a Shell owner carrying the number, not a team sitting beside the business. If nobody inside wants to own it, that is a good reason not to do it.
And a stated failure condition: if a redemption pulls less than about HK$220 of extra forecourt spend, the reward scheme doesn’t pay, and I’d say so in writing.
All public sources: Shell’s own station locator and Recharge pages; EPD register of public chargers; CLP 2026 tariff tables and 港燈 HK Electric published tariffs; Cornerstone Technologies (SEHK GEM:8391) audited accounts; LegCo Environmental Affairs Panel, Sept 2025. Nothing from inside Shell.