River
Y CombinatorBacked by Y Combinator
FREE TEMPLATE

MEDDICC Deal Qualification Scorecard

Four sheets that treat the scorecard as an instrument: what each score came from, when it expires, and whether five reps read it alike.

Free download  ·  No account needed

Instrument Read, Q3 2026

68 open deals, 5 raters, 3 calibration deals

Commit book, everything recorded£892,000
Commit book, current sourced evidence only£156,000
Median evidence age in commit73 days, against a 45-day clock
Elements that cannot carry a gate2 of 8, both required at full marks by the old gate

1. Does the instrument read the same thing twice

ElementRound 1 spreadRound 2Verdict
Economic Buyer3 of 31Fixed by one sentence
Champion3 of 31Fixed by one sentence
Paper Process21Tight

2. Does every element separate anything

Identify Pain reads 3 on 61 of 68 open deals. Nobody enters a deal with no pain into a pipeline, so the element was being scored after the selection had already happened. It carried the same weight in the gate as the elements that vary.

3. The three commit deals that fall

AccountValueRecordedCurrent
Pentworth Council£310,0002011
Erskine Water£240,0002214
Cadwell Utilities£186,000219

Pentworth's champion left in July and the cell still reads 3. Illustrative figures for one fictional field service software vendor.

The rubric is a solved problem. Search the phrase and page one hands you the same build eight times: eight elements, a nought to three scale, an evidence field that is mandatory above a 2, a gap, an owner. Then custom CRM fields with the property names spelled out, and a stage gate that blocks commit until every element clears 3. It is a good build. It is also finished, and it stops exactly where the interesting question starts.

That number goes into a forecast, and nobody has established that it reads the same thing twice. Put one anonymised deal and one evidence bundle in front of five reps and Economic Buyer comes back anywhere from 0 to 3. Measuring that has a standard method: different operators, the same objects, and repeatability and reproducibility reported separately. A manager auditing afterwards is not it, because the manager is one more rater with no calibration of their own.

Two more failures hide in the same total. An element reading 3 on nine deals in ten separates nothing, because displaying four levels is not resolving four levels. And a score is a statement about a date: what a problem costs a buyer barely moves in a quarter, and who signs stops being true the morning they resign. So every element gets its own clock, and the pipeline is read twice, once on everything recorded and once on what is still current. Where the scores come from is the discovery call's job.

The commit number, read twice

Qualification Read by Deal with a total on everything recorded and a total on current evidence, the calibration round that fixed two elements in one sentence each, and the distribution that exposes an element separating nothing.

Qualification Read by Deal

Illustrative rows for Norbury Systems, a fictional field service software vendor. Expired and unsourced scores count as zero in the current column, whatever number sits in the cell.

DealAccountValueStageRecordedCurrentFallGate
N-2134Pentworth Council310,000Commit2011-9Fails
N-2102Erskine Water240,000Commit2214-8Fails
N-2041Cadwell Utilities186,000Commit219-12Fails
N-2088Marchwood Group156,000Commit2220-2Passes
N-2150Ashmore Logistics96,000Qualify1312-1Fails
N-2117Balfern Estates52,000Validate1615-1Passes

Two failures that read identically in a total. Cadwell fails on the total and needs four things. Ashmore passes on the total and fails on one required element, because its Economic Buyer is a job title typed into a web form. That deal needs one call.

Roll-up: £736,000 of the £892,000 in commit falls below its own gate. Cadwell's largest single problem is an Economic Buyer score sourced to an April call and a Champion score with no artifact behind it at all.

Rater Calibration

Five raters, three anonymised closed deals, one identical evidence bundle each, scored silently before anybody spoke.

ElementR1R2R3R4R5SpreadVerdict
Economic Buyer033133Unusable
Champion031323Unusable
Decision Process132212Loose
Metrics333231Tight
Identify Pain333330Agreed and uninformative

A spread of 3 on a four-point scale is the full range. Economic Buyer and Champion were also the two elements the old commit gate required at full marks, so the gate was being decided by the two things the team could not measure.

The fix was one sentence each, not training. Confirmed meant we have met them to three raters and they have said they sign it to two. Champion did not distinguish friendly from advocating. Rewritten to name an observable, re-run on the same three deals: both spreads fell to 1.

A spread of 0 is not automatically good. Identify Pain agreed perfectly because everybody always scores it 3, which is a constant rather than a calibrated element. That is the failure that hides best.

Element Discrimination

Distribution of every element across 68 open deals, cross-referenced against the rater spread from the calibration round.

Element0123BlankModalSpreadVerdict
Identify Pain01661090%0No resolution
Economic Buyer392432047%3Spread without agreement
Champion2112629043%3Spread without agreement
Paper Process038134465% blank2Not scored
Metrics4122923043%1Discriminating
Decision Process9222710040%2Discriminating

Three failures, three different owners. No resolution is a rubric problem. Spread without agreement is a rubric problem the distribution hides, because it looks healthy from a distance and the calibration round is the only thing that catches it. Not scored is neither: the element works and nobody asks, so it goes to the call agenda.

Competition sits at 3 on 36 of 68 deals in a market with two live competitors. The distribution passes and the number is still worth a look, because a 3 awarded for assuming the incumbent is an assumption, which is a 0.

What's in the pack

01

Element Definitions

Eight elements where every level names an observable that either appears in the record or does not, each carrying its own expiry clock and a worked example.

02

Scorecard

One row per element per deal, with the artifact the score came from, the person whose words are in it, the artifact's date, its age in days and its currency state.

03

Qualification Read by Deal

Two totals per deal, one on everything recorded and one on current sourced evidence, with the fall between them and a gate verdict written as a sentence.

04

Rater Calibration

The same anonymised deals scored by everybody at once, with the spread per element and one line saying which word in the rubric is carrying two meanings.

05

Element Discrimination

The distribution of every element across the open pipeline, separating an element that cannot discriminate from an element nobody fills in.

06

Calibration Standard

How to run the ninety-minute round: which deals, which bundle, silent scoring first, and what to do with an element that returns the full range.

07

Coaching Guidance

What a low score means element by element, and why expired, unsourced and genuinely low are three different conversations with three different fixes.

08

Instrument Read

The quarterly page for whoever defends the forecast: whether the scorecard reads the same thing twice, what it separates, and how old the evidence under it is.

How to use it

  1. 1

    Open in River, or take it blank

    Open the pack in River and hand it your pipeline, or download the Word documents and CSV sheets and fill them in yourself.

  2. 2

    Send the record, not the CRM

    Transcripts, notes, emails and whatever the buyer sent, plus three or four closed deals. Scores get rebuilt from artifacts rather than from the existing cells.

  3. 3

    Age it, then read it twice

    Every score gets a source, a date and a clock. The commit book comes back as two numbers, and the gap between them is the forecast conversation.

  4. 4

    Run one calibration round

    Ninety minutes, five people, three anonymised deals. The elements that return the full range come out of the gate until the rubric names an observable.

Frequently asked questions

Is this template free?

Yes. Four Word documents and four CSV sheets, downloaded as a zip, no signup. River is the optional half: it rebuilds the scorecard from your record, ages it, and runs the discrimination analysis. Neighbouring packs sit in the template library, and one-shot jobs in the tool index.

We use MEDDIC, or BANT, or SPICED. Does this assume MEDDICC?

It ships with eight MEDDICC elements and nothing depends on there being eight. Calibration, discrimination and evidence ageing are properties of any scored instrument, so drop an element or add one and the sheets work unchanged. What the pack will not hand you is somebody else's point thresholds.

How is this different from the discovery call pack?

That one owns the call: which questions to ask, how to score what you heard, and which questions your won deals had in common. This one owns the standing score on every open deal and whether it can be trusted. Run the discovery pack first if your problem is the conversation.

Why not just have managers audit the scores?

Because a manager is one more rater with no calibration of their own, and when the two of you disagree the senior number wins. That teaches the team to predict the manager. Reproducibility is measured across different operators scoring the same objects, which means a round, not an audit.

How many deals do we need before the discrimination read means anything?

Around forty open deals is where a distribution starts having a shape you can read. Below that the sheet still fills, and it reports the count beside every percentage so nobody quotes a modal share computed on eleven deals in a pipeline review and never retracts it.

What happens to deals that fail the gate on current evidence?

They go back a stage, which is unpopular, so the read names the specific missing evidence per deal rather than telling anybody to re-qualify. A deal with one named missing conversation returns in a week. A deal told to re-qualify returns with the same scores and a fresh date. Once it does return, the mutual action plan is where the remaining dates live.

Can we score reps on their average?

The guidance says not to, and the reason is in the data. A rep working public sector deals with published timetables outscores a rep working inbound web forms, and neither fact is about the rep. Whatever gets targeted gets entered, and the entries are the only thing here with any value.

Read your commit number twice

Take the Word documents and CSV sheets blank, or open this exact pack in River and let it age your pipeline against the record first.

Edit with AI