Methodology
A data product is only worth what its method is worth. Here is exactly how every figure in the guide was produced, and how you can check any of it yourself.
The window
All measurements were taken between 8 and 13 September 2026. Every figure in the guide carries its own date, because platform policies, bid allowances and geo-restrictions change without announcement.
The account
A single real account, resident in India, applying as an Indian engineer would. Not a US VPN, not a recruiter seat, not an enterprise view. This matters: several platforms show a different job set depending on the viewer's location, and the whole point was to see what an Indian developer actually sees.
What counted as "India-eligible"
A role counted only if an Indian resident could plausibly be hired for it. Specifically:
- The posting's own location list included India, "worldwide", or "anywhere".
- The body text did not contradict the metadata. This is the one that catches people — a platform flag saying location is not strictly required, on a posting whose body reads "United States only, W2".
- Engineering roles only. Data-annotation and language work was counted separately, because bundling it in makes a closed platform look open.
Counting, not sampling
Where a platform exposed a full list through its own interface or public API, every row was counted rather than sampled. That is how a claim like "139 open jobs, 27 engineering, 0 India-eligible" can be stated as a count instead of an impression.
The usual method is to browse a few pages, form an impression, and write it up. That produces advice like "Braintrust is worth a try" — which is wrong, and costs the reader a week. Counting produces "27 engineering roles, zero open to you", which is checkable and actionable.
Re-verifying it yourself
Chapter 3 of the guide gives the specific one-minute check for each platform — which list to pull, which field to read, and which field lies. You do not have to take any number on trust, and you should not: by the time you read this, some of them will have moved.
What is not measured
Conversion rates. This report says which platforms will let you apply; it does not claim to know your odds of being hired once you do. Anyone publishing that number from a sample this size would be making it up.
Found a figure that is now wrong? Tell us — corrections are published and buyers get the updated file free.