eSIM vs Local SIM vs Roaming: The Actual Math

Everyone has an opinion about eSIMs and almost nobody has run the numbers. I tracked what I actually paid for mobile data across 14 countries over 18 months, and the gap between the convenient option and the annoying one is bigger than I expected — roughly six times, at the extreme.

That doesn't mean the annoying option always wins. It means you should know what convenience is costing you before you pay for it.

How I measured it

The headline price per gigabyte is close to meaningless, because most of what you buy expires unused. So I tracked cost per usable gigabyte: total spent divided by data actually consumed before the plan expired or I left the country.

I also counted setup time, because an hour spent queuing in a carrier shop is not free. I've valued it at nothing in the table below and then discussed it separately, which seemed more honest than picking an hourly rate for you.

The numbers

Two things stand out. First, local SIMs really are dramatically cheaper — roughly six times cheaper than bundled roaming per usable gigabyte. Second, they also waste the most data, because local plans are sold in chunks sized for residents rather than for someone leaving in nine days.

The waste figure is the one nobody talks about. If a plan looks 70% cheaper but you use half of it, it's 40% cheaper. Still a win — just a smaller one than the sticker suggests.

When each one wins

Roaming wins when the trip is short

Under four days, a day pass on your existing plan is usually the right answer. You keep your number, there's nothing to set up, and the absolute amount of money involved is small enough that optimising it is a waste of your attention.

eSIM wins for the first 48 hours, always

This is the part I'd defend hardest. Landing with working data before you've spoken to anyone is worth real money: you can get a ride, find the address, and message whoever's expecting you. A small eSIM package as an arrival buffer is cheap insurance even if you intend to buy a local SIM later.

Local SIM wins past about a week

Above roughly seven days, the price gap gets big enough to justify the queue. Above three weeks it isn't close. The break-even point moves earlier if you're a heavy data user and later if the nearest carrier shop is an hour away.

The convenient option isn't wrong. It's just priced like convenience, and you should know that's what you're buying.

Three traps worth knowing

The "unlimited" that isn't. Most unlimited travel plans throttle hard after a few gigabytes — often to speeds where maps still work and video doesn't. That's frequently fine, but it isn't unlimited, and the throttle threshold is the number that matters.

Registration requirements. A growing number of countries require passport registration for a local SIM. That's usually a five-minute formality in a shop and impossible from a street kiosk, which is why the "cheap SIM from the airport corner store" advice fails in some places.

Tethering restrictions. Some cheap local plans block tethering, which quietly destroys the plan if your laptop is your income. Check before you buy, because it's rarely refundable.

My actual rule

Under four days: roaming day pass, don't think about it. Four days to two weeks: eSIM, sized deliberately small, top up if needed. Over two weeks: eSIM on arrival, local SIM within 48 hours once I know where I am and what I need.

That rule costs me maybe $15 a year more than perfect optimisation and saves me a great deal of standing in queues. That's a trade I'll take.

What it actually cost, country by country

The medians in the table above hide a lot of variation, and the variation is where the useful decisions live. Here are the countries where the gap was widest and narrowest, from the same 18-month tracking period.

The rule that falls out of this is simple and I now apply it without thinking: the cheaper the country, the more a local SIM is worth the queue. In India the gap is large enough to justify almost any amount of hassle. In Switzerland or Japan, buying a local SIM is close to pointless — the convenience premium on an eSIM is under 40% and you're spending an hour to save the price of lunch.

This is the opposite of how most people behave. The instinct is to optimise hardest in expensive countries, because that's where the absolute numbers feel big. But the ratio is what determines whether the effort pays, and the ratio is best exactly where the absolute saving looks small.

The 20-minute arrival routine

Whatever you choose, the failure mode is the same: landing with no data, no way to call the accommodation, and a taxi queue. Here's the routine that has removed that problem entirely.

1. Install and activate the eSIM before you leave home, while you're on wifi you trust. Most travel eSIMs let you install the profile early and only start the clock when it first connects in-country. Doing this at the gate on airport wifi works but is unnecessarily exciting.
2. Set your home SIM to data-off, calls-on before boarding. This is the single most common cause of a surprise roaming bill: a background app syncing on a network you didn't intend to use.
3. Download offline maps for the arrival city. Ten seconds of preparation that has saved me on four occasions when the eSIM didn't connect immediately.
4. On landing, wait for the eSIM before touching anything. It can take five minutes to attach. Repeatedly toggling airplane mode does not speed this up and does sometimes make it worse.
5. Only decide about a local SIM once you're settled — ideally the next day, when you know how much data you're actually using and whether the trip is extending.

The last point matters more than it sounds. Buying a local SIM at the airport is the most expensive place to buy one in almost every country, and you're doing it while tired and without knowing your real usage. Waiting 24 hours costs you a few dollars of eSIM data and reliably produces a better decision.

Questions I get about this

Found an error in this piece? Tell me — corrections get made in the open with a dated note, never a silent edit. The log is public.