The coverage audit: checking what your alerts are missing

Every bid team knows its win rate. Almost none know its coverage rate, which is the share of relevant tenders the team actually saw while they were still open. Win rate measures how well you bid. Coverage measures whether you got the chance to.

The awkward thing about a missed tender is that it leaves no trace in your pipeline. Nobody files a report about the contract they never heard of. So coverage degrades quietly. A portal changes its categories, a buyer words something unusually, a country you sell into publishes below the threshold you watch, and the first sign of trouble is a competitor announcing a win in your sector that you have no memory of seeing.

This is the routine for finding those gaps on purpose. Run it once a quarter. It takes about ninety minutes.

What makes the audit possible

Award notices. When a public body finishes a procurement, it publishes what it bought, from whom, and usually for how much. That notice appears whether or not you ever saw the original tender.

Awards are therefore a receipt for the market you claim to cover. If a contract in your sector, in your region, at your size was awarded last quarter and you have no record of seeing it while it was open, that is a miss, and it is countable. Learning to read those notices closely pays off well beyond this audit, which is the subject of what published award notices tell you about your competitors.

Step 1: Define what "should have seen" means

Before pulling any data, write your coverage claim in one sentence. Something like: "We should see every open tender for building services in Denmark, Sweden and Norway, above €50,000."

Be specific about sector, geography and size, and keep it honest. This is the claim you are about to test, not an aspiration. Vague claims produce vague audits, and a claim you have quietly widened over the years is usually the reason the first audit stings.

Step 2: Pull last quarter's awards

Go to the portals that publish awards for your markets and search the previous quarter, filtered to your categories and countries. On TED, search contract award notices by CPV code and date range. On SAM.gov, use the award data. In the UK, awards appear on both Contracts Finder and Find a Tender depending on value. Most national portals have an equivalent, usually labelled "awarded" or "results".

Export what you get rather than browsing it. The point is to work through a list systematically instead of stopping when you recognise something.

Two filters are worth applying immediately. Drop awards below your size threshold, and drop the ones whose scope is genuinely outside what you do. Be strict here in your own disfavour. If it is arguable, keep it in.

Step 3: Check each award against your own records

For every award that survived filtering, ask one question. Did we see this while it was open?

Your evidence is whatever you actually keep: the alert inbox, the bid tracker, the shared folder, someone's memory. Mark each award as one of three outcomes.

• Saw it, bid it. Working as intended.
• Saw it, chose not to bid. Also working as intended. Qualification is a separate problem, covered in the ten minute bid or no bid decision.
• Never saw it. A miss.

Do not soften the third category. "We probably saw it" is a miss. If you cannot produce evidence that you saw it, you did not see it in any way that mattered.

Step 4: Trace every miss to a cause

This is the part that changes what you do next. For each miss, work out why it never reached you. In practice it is nearly always one of five things.

The classification was wrong or unusual. The notice carried a CPV, NAICS or UNSPSC code you do not watch, often a generic services code, sometimes one that is simply mis-assigned. Check the code on the notice against the codes in your alert setup.

The wording did not match your keywords. Buyers describe the problem, not your product category. A software procurement titled "case management modernisation" will not trip a keyword alert for "software development". Read the title and ask whether your saved search could plausibly have matched it.

It was published somewhere you do not watch. Below-threshold contracts do not appear on TED, and regional and municipal portals publish independently. If the award came from a portal that is not on your list, you have a coverage hole rather than a filtering problem, which is a different fix and a bigger one. Deciding which portals your firm actually needs walks through that design exercise.

It was in a language you do not search. Your keywords are in English and the notice was in Finnish. Machine translation on your side does not help if the alert matched against the original text.

Nobody was watching that week. Holiday, illness, or a fortnight when everything went sideways. This one is uncomfortable but common, and it is the only cause with a purely human fix.

Write the cause next to each miss. You are building a distribution, not an anecdote. The five causes are worth understanding in their own right, and the systematic sources of missed work covers each one with its own diagnostic.

Step 5: Count, then fix the largest bar

Two numbers come out of this.

Coverage rate is seen divided by seen plus missed. Anything above 90 percent is respectable. Below 70 percent means your pipeline is substantially a matter of luck.

Misses by cause is the count in each of the five buckets. Put them in order.

Misses by cause, one quarter

Fix the largest bucket, and only the largest. If most misses are classification, widen your codes and re-test next quarter. If most are wording, your keyword strategy is the problem and no amount of extra portals will help, which is what tuning alerts so the team still reads them is for. If most are portals, add the portal. If most are language, add search terms in the local languages of the markets you claim. If most are gaps in attention, the fix is a rota, not a tool.

Fixing one thing at a time is not caution for its own sake. Change three things and next quarter's number cannot tell you which of them worked.

Then write the numbers down somewhere you will find them next quarter. This audit compounds only if you can compare it to the last one.

The record sheet

Keep it simple. One row per award, one tab per quarter.

At the bottom: total assessed, total missed, coverage rate, and the largest cause.

What good looks like after a few rounds

The first audit is usually unpleasant. Teams who have never run one often find coverage well below what they assumed, and the causes cluster in one or two buckets rather than spreading evenly, which is good news, because it means one fix moves the number a lot.

By the third quarterly audit you should expect two things. A coverage rate you can state out loud in a management meeting, and a shortlist of structural gaps you have decided not to close because the work is not worth it. That second list matters as much as the first. Knowing which parts of the market you are deliberately not watching is a strategy. Not knowing is a blind spot with better branding.

Common questions

How often should we run a coverage audit?

Quarterly is the right cadence for most teams. It is frequent enough that a change you make in one quarter can be measured in the next, and rare enough that the ninety minutes is easy to justify. Run it monthly only if you have just made a large change to your monitoring and want to know quickly whether it worked.

What is a good coverage rate?

Above 90 percent is respectable, and below 70 percent means a large share of your pipeline depends on luck. The absolute number matters less than the direction across quarters, because your coverage claim and your market both change over time.

Why use award notices instead of just checking the tender feeds?

Because tender feeds only show what already reached you, so they cannot reveal what did not. Award notices are published independently of whether you saw the original tender, which makes them the only free and systematic record of the market you claim to cover.

How long does a coverage audit take?

About ninety minutes per quarter once you have run it twice. The first one takes longer, usually two to three hours, because you are still deciding what counts as in scope and still learning where each portal hides its award data.

Does this work for below-threshold contracts?

Partly. Many buyers publish awards for below-threshold work voluntarily or under national rules, but coverage of those awards is uneven. Treat the below-threshold portion of your audit as a sample rather than a census, and weight your conclusions accordingly.