The GeM mirror on CPPP is missing a quarter of GeM

CPPP describes itself as a single point of access to procurement across the country, and it carries a mirror of GeM to make that true. If the mirror were complete, watching CPPP alone would be the obvious thing to do: one portal, one set of scrapers, one place to look.

We counted both on the same day.

Twelve and a half thousand bids sit on one side of that and not the other. Put the other way round, reading GeM directly gets you about 35% more than the mirror does.

Both framings are the same measurement and people routinely state one while meaning the other, so it is worth being explicit: the mirror carries 26% fewer, and going direct gains you 35% more. We had that backwards in our own copy until recently, which is the kind of thing that happens when a number gets restated one time too many.

Coverage is the obvious argument. The one that matters more in practice is that GeM direct is the only source of the five we read with reliable titles.

That sounds like a detail and is not. Matching a tender against what a firm actually builds means reading the title and the item list, and a title that arrives truncated, generically categorised, or reworded by an intermediate system is a title you cannot match on. Every rule you write against it fires on the wrong things or fails to fire at all.

Honesty about the trade: reading GeM directly is more work. It is a second system with its own conventions, and one of them catches people out. Reverse auctions do not serve their documents from the same endpoint as an ordinary bid, so tooling that assumes one document URL silently gets nothing for that whole class of tender and reports it as a bid with no document rather than as a bid it failed to read.

The general lesson is the one worth taking away. A source returning nothing and a source having nothing look identical from the outside, and every portal gives you at least one way to confuse them. If you build or buy anything that watches these portals, the question to ask is not how many sources it reads. It is what it does on the morning one of them returns an empty list.

Take the ongoing-bids count from GeM's own listing and the count from the GeM section on CPPP on the same day, within an hour or so of each other. The absolute numbers will have moved since we measured. The gap between them is the thing to look at, and it should be substantial.