Proxies for McDonald's: Menu Prices Across 100 Countries, and the Big Mac Index Problem

The Economist has priced a Big Mac in dozens of countries since 1986 as a rough yardstick for currencies, and its published methodology says the prices come "from McDonald's directly and from reporting around the world" (The Economist, big-mac-data). Anyone trying to build the same dataset without The Economist's contacts runs into the structure of the company. At the end of 2025 there were 45,356 McDonald's restaurants in more than 100 countries, and about 95 percent of them were franchised (McDonald's 2025 Form 10-K), which means franchisees set their own prices, every country runs its own website and its own app, and the corporate US site shows a menu with no prices on it at all. Reading McDonald's prices at scale is therefore a geography problem before it is a scraping problem, and geography problems are what proxies are for.

We run country-pinned proxy traffic for jobs like this, and on 24 August 2026 we also tested mcdonalds.com from a datacenter server and through 20 free proxies so the front-door section below is observed rather than assumed.

The short answer for McDonald's

Country-pinned residential proxies, driving a real browser, one country per session. For store-level prices in the United States, where the corporate site publishes none, the readable surface is the delivery marketplaces, which means metro-pinned US residential exits with one sticky session per address, exactly as in proxies for DoorDash. Datacenter exits were stalled in our test, and the app is out of reach for any proxy because it locates by GPS.

What McDonald's is as a target

McDonald's is not one website. Each country has its own site under its own path or domain, with its own menu, its own promotions and its own legal text, and each country publishes its own app. In the United States the website lists the menu without prices, with a note that prices and participation vary, because the franchisee owns the number. A price appears in the app once a specific restaurant has been chosen by location, and it appears per store on the delivery marketplaces that list McDonald's, where the franchisee's delivery menu carries the franchisee's delivery price.

For a data job that produces three layers. The country layer: what is on the menu and what it costs where the site prints prices. The store layer: what a specific restaurant charges today, readable through the app for one store at a time or through the marketplaces for many. And the location layer: 45,356 restaurants, with hours and services, exposed through the restaurant locator on each country site.

Jobs a proxy is the right tool for

• Cross-country price and menu comparison. The Big Mac index shape: the same items across markets, over time, with currency and date attached. Researchers, analysts and journalists build it, and it needs an exit inside each country so the site serves that country's menu.
• Regional item and promotion tracking. Items that exist in one country and not another, seasonal returns, country-specific promotions on the country sites. Marketing teams at competitors and food media track these per market.
• Store-level price monitoring in the United States. Because the corporate site prints no prices, store-level monitoring runs through the delivery marketplaces, one address per session, metro by metro. The method is the DoorDash and Uber Eats method, and those pages have the setup.
• Location data. The restaurant locator on each country site is the public record of where the restaurants are, with hours, drive-through, delivery and other services. Site-selection analysts, competitors and mapping projects read it, and at 45,356 restaurants that is many paged locator queries.
• Localization testing. The German site in German with euro prices where printed, the right promotions, the right legal footer, read from a German exit. The method is in geo-testing with proxies.

Jobs a proxy is not the tool for

App deals in another country. Apps are published per country and choose a restaurant by the phone's location. Neither the app store region nor the GPS reads the IP address. This is the most common thing people ask us about McDonald's and the answer is a flat no.

Multiple accounts for deals. Deals attach to an account with a phone number or email and to the app on a device, and creating accounts to reuse offers runs against McDonald's terms. Nothing but the network address changes behind a proxy.

Ordering from abroad. Ordering runs through the country's app with a local payment method and the phone's location. None of that is a network property.

The front door, observed

The request we sent to mcdonalds.com on 24 August 2026, from an ordinary datacenter server with a normal browser user-agent, completed its TLS handshake with an Akamai edge, opened the HTTP stream, and then received nothing at all until the timeout. We repeated the request through 20 elite HTTP proxies from our free proxy list filtered to the United States; every one of them timed out the same way. The site did not refuse the datacenter clients. It stalled them, which costs the client a full timeout per attempt and tells it nothing.

That behavior shapes the plan. A stalled request looks like a network fault, so a scraper that retries on timeout will retry forever against a wall. Use a real browser on a residential exit, set a firm timeout, and treat a stall as a block signal rather than a transient error. Our guide to scraping past Akamai covers the sensor-cookie side of that stack, and every proxy error explained covers how to tell a stall from a dead proxy.

Which proxy type fits, by job

Where the two residential flavors differ, rotating vs static residential proxies has the tradeoffs.

Free versus paid for McDonald's

The 20-for-20 timeout is the free-proxy story here. Free proxies are datacenter addresses, the site stalls datacenter clients, and every attempt costs you the full timeout for no data. That is worse than a clean refusal, because it burns time as well as the request. For learning and for testing your code against a site with a softer door, our free proxy list is there and refreshes every few minutes; verify anything in the proxy checker before you rely on it. For McDonald's itself, our residential proxies start at $0.44/GB, pay as you go, no KYC, and pin to any of the countries the sites serve.

How to set it up

1. One country, one exit, one session. Pin the residential exit to the country whose site you are reading, then open that country's site path. Reading the Japanese site from a US exit may still render, but the promotions, the language defaults and the risk checks all key off where you appear to be.
2. Drive a real browser with a firm timeout. Playwright or Puppeteer, current user-agent, cookies kept, and a timeout you enforce, because a stalled request never fails on its own. The Playwright proxy guide shows the launch flags.
3. Treat a stall as a block. Retire the exit, do not retry it in a loop, and log it. A rising stall rate means slow down or widen the pool.
4. For store-level US prices, go through the marketplaces. Set an address, read the McDonald's storefront for that address, rotate for the next address. Proxies for DoorDash and proxies for Uber Eats have the full method, including their own front doors.
5. Page the locator slowly. Locator queries are coordinate-based and there are tens of thousands of restaurants; sweep a grid at a human pace and cache what does not change. More on pacing and back-off in avoiding IP bans while scraping.

How many IPs

Size by countries and stores, not by menu items.

Country price index, 30 countries, monthly:
  countries                    30
  pages per country            ~20 (menu, promos, a few locator pages)
  sessions per exit per day    3-4, slow pacing
  exits needed                 1-3 residential per country

Store-level US monitoring via marketplaces, 10 metros:
  addresses per metro          40
  sessions per exit per day    4-6
  exits in rotation            ~8-10 US residential per metro

Residential is billed by traffic; both plans are pacing budgets,
and text-only reads keep the bill small.

A monthly index run costs a month of traffic and nothing between runs; see pricing, which is pay as you go with no expiring balance.

Where a proxy stops

An address inside the country or metro is what the proxy supplies, and the list of what it leaves alone is longer: a site that publishes no prices still publishes none, an app that locates by GPS stays where the phone is, a stalled datacenter request stays stalled until a real browser on a residential exit replaces it, and McDonald's terms, which prohibit scraping, read the same through any exit. Within that, residential exits let a legitimate data job read the public country sites and marketplace storefronts the way local customers do, one country or one address at a time, at a human pace, with a block rate you watch.

For the two other chains where the same layers apply with different front doors, see proxies for Starbucks, where the locator is the main dataset, and proxies for Taco Bell, where test markets make city-level reads the point.

Sources

• The Economist, big-mac-data repository (methodology: prices "from McDonald's directly and from reporting around the world"): github.com/theeconomist/big-mac-data
• McDonald's Corporation, Form 10-K for the year ended 31 December 2025 (45,356 restaurants, more than 100 countries, about 95 percent franchised): sec.gov
• HProxy observation, 24 August 2026: one direct request to mcdonalds.com from a datacenter server (TLS completed, no response before timeout) and 20 US elite free proxies from our free proxy list (all timed out).