For sales operations and bid teams, a useful tender monitoring system produces a review queue that someone can act on. Start with what you can deliver, where you can deliver it, and which buyers matter. Then test whether the search retrieves known relevant notices.
This guide produces a monitoring specification and a weekly quality check. It is tool-independent: translate the logic into the fields your chosen portal actually supports.
Write the market scope before setting filters
Record your offer in buyer language, acceptable delivery locations, buyer types, mandatory capabilities and exclusions. Separate preferences from hard constraints. “Prefer universities” is different from “cannot perform outside Belgium.”
Use official CPV definitions to build a code shortlist. Keep a keyword search alongside it: a relevant purchase may sit under a broad main code or another code in a mixed lot. TED search help explains its search fields and browsing by business sector and place of performance.
Use three complementary searches
| Search | Logic to configure | What it protects against |
|---|---|---|
| Core offer | Relevant codes OR specific offer phrases; then delivery region | Notices described differently from your sales material |
| Buyer watch | Named buyer identifiers and aliases; inspect relevant stages | Broadly classified purchases from priority accounts |
| Discovery sample | Wider category or keyword search, reviewed separately | Relevant notices excluded by the core rule |
Do not require every notice to match a code, keyword and buyer simultaneously unless that narrow intersection is deliberate. Do not assume a portal's parent-code selection includes every child code: test it against known notices. Buyer location can differ from the place of performance.
Worked example: a university network supplier
Illustrative configuration, not a claim about current contracts or Duke coverage:
- Offer phrases: campus network, wireless access, network maintenance, and local-language equivalents.
- Candidate classifications: the relevant equipment and service branches identified from previously suitable notices.
- Geography: places where installation and support are deliverable; review remote-service lots separately.
- Buyer watch: named universities plus their purchasing consortia, using identifiers where available.
- Stages: planned procurement in a preparation queue; open competitions in an active review queue; awards in a buyer-history queue.
- Exclusions: remove clearly unrelated construction only after checking whether network work is included in a mixed lot.
Take ten known relevant notices as a starter test set. Record whether each is returned and why. Ten is a practical starting sample, not proof of market completeness.
Route every relevant result to a decision
Keep notice ID, lot ID, official URL, latest version checked, publication date, deadline with timezone, reviewer and next action. Merge repeated publications into a procedure history without losing amendments. A cancellation or changed deadline should reach the person already reviewing the pursuit.
Assign one reviewer and a backup. Set review frequency from actual deadlines and clarification windows; use a slower cadence for historical research. Official document access and portal registration deserve an early check.
Tune noise without hiding unknowns
Each week inspect rejected results and a wider independent search. Label errors as wrong scope, classification mismatch, location mismatch, duplicate, expired notice or missing source. Measure relevant reviewed results divided by all reviewed results. Keep the sample and denominator visible.
That ratio measures queue precision. Recall requires a separate benchmark of relevant opportunities, and cannot be inferred from an empty inbox. Missing classifications and values should be review categories, not automatic rejection rules.
Your next action
Write one monitoring specification, replay it against known notices, and send the useful results through the 15-minute notice review. Use the bid/no-bid checklist for pursuits requiring team time. For missed-source questions, use the procurement data source audit.
If evaluating Duke, bring that specification and three known notices to a product discussion. Ask to reproduce the results and inspect the source record before relying on an alert workflow.
Reviewed 13 September 2026. Editorial method: proposed operating practice; portal field behavior is supported by the linked TED documentation.