Your tender alerts are either noise or a blind spot

Almost every supplier who has bought a tender alert service ends up in one of two states: forty emails a day nobody opens, or a quiet inbox and the recurring discovery that a contract they would have won went out without them seeing it.

These look like opposite problems. They are usually the same one.

The three failure modes

1. Keyword-only targeting

A feed built on the words you use rather than the classification buyers use. It misses everything titled in the buyer's vocabulary instead of yours, and floods you with unrelated notices that happen to contain 'support' or 'services'.

Symptom: high volume, low relevance, and a nagging sense that things are getting through.

2. Over-narrow CPV

Subscribing to five-digit codes when buyers tag at two. The feed goes quiet and you conclude there is no work in your sector, when there is — filed one level up by someone in a hurry.

Symptom: very low volume, everything that does arrive is relevant, and competitors keep winning work you never saw.

3. Source gaps

The targeting is fine and the source list is not. A UK-wide feed will not show you Scottish below-threshold work, and no aggregator replaces registration on the portal your biggest buyer actually uses.

Symptom: relevance is good, but the misses cluster — always the same buyer, always the same nation.

Diagnosing yours in an afternoon

1. Take the last twenty contracts you would have wanted — won, lost, or found too late. Any source will do: your CRM, your competitors' award notices, memory.
2. For each, find the notice. Record the CPV codes the buyer used, the service it published on, and the exact title.
3. Count how many your current alerts would have surfaced. That is your recall, and it is usually worse than people expect.
4. Look at the misses together. If they share CPV codes you do not monitor, that is failure mode 2. If they share a source, that is failure mode 3. If they share nothing except unfamiliar wording, that is failure mode 1.
5. Then measure the other direction: of last week's alerts, how many were worth opening? Below one in five and your set is too broad.

What good looks like

• CPV at division or group level for breadth, with keywords layered for the work that CPV describes badly.
• A value floor set to what you will actually bid for, so small work is filtered rather than skimmed.
• Every relevant source, including the national portals for the nations you sell into.
• A digest at a fixed time rather than notifications on arrival — a batch you read is better than alerts you dismiss.
• Closing-soon separated from newly-published. They are different decisions and they need different urgency.
• Something that tells you when a monitor returns nothing for a fortnight. Silence should be a signal, not a default.

The part most tools skip

An alert that tells you a notice exists has done a third of the job. What you need before deciding whether to open it is whether it fits: your codes, your regions, your value band, your accreditations — and who the incumbent is.

TenderGraph sends a daily digest of what is new against your profile and what closes within the week, scored against your saved targeting, with the evidence for each score and the unknowns named. The same data is available to an agent you connect yourself, if you would rather ask than read.