What should a sales call scorecard include?

The guides we read for this piece answer the question with a template: a handful of sections, a weight on each, sometimes a pass mark in the eighties. The templates are fine. What they skip is that a real scorecard does several jobs at once. Part of it coaches, part of it protects the company, part of it just labels what happened so somebody can count it later. Put all of that on one scale and the score ends up averaging things that were never meant to be averaged.

So this piece does the work in the other order. First, what kinds of line belong on a scorecard and what each kind is allowed to do to the number. Then size, weights and auto-fails. At the end there is a compact discovery-call scorecard you can copy and cut down.

A scorecard holds five kinds of line, and only two of them should move the score

A sales call scorecard mixes scored behaviors, auto-fail checks, outcome lines, labels and lines the rep does not control. Only scored behaviors and auto-fails should change the number. The other three are recorded and counted, but if they feed the score it starts measuring the buyer, the call type or the result, and stops measuring the rep.

The outcome row is the one teams argue about. "Next step agreed" belongs on the form, because without it you cannot see where deals stall. It does not belong in the ranking of what to coach, because it is the thing the other lines are supposed to explain. Why that ranking goes circular is covered in the piece on post-demo silence. Kixie's scorecard guide from July 2026 reaches the same place from the other side: score controllable behaviors and review pipeline and conversion metrics separately.

The last row is the easiest to leave out. Some things happen on a call because of who the buyer is, and a form that scores them penalizes the rep who drew the harder buyer. After a few weeks of that, the team reads the scorecard as luck. Run the form you use today against a week of recordings and it sorts itself into those five kinds.

Five to eight lines per call type is less than teams actually check

Claap's call scoring guide, published October 14, 2026, puts the right size of a scorecard at 5 to 8 lines per call type: a short list of behaviors a manager can coach, on one type of call. Other vendor guides go up to about fifteen. The forms we see in production are much longer than either, and the length is not automatically a mistake.

Over September 2026, across eight sales teams whose forms stayed unchanged for the month, the yes-or-no part of the scorecard ran from 24 lines to 128. Seven of the eight sat between 24 and 51. Labels and free-text notes come on top of those counts.

Part of the gap closes once you sort the lines into the five kinds. A short guide describes the coachable core. A production form carries that core plus whatever else the company needs checked on every call: compliance steps, outcome lines, behaviors split finer than a coach would bother with. We have not measured how that split falls across the eight teams, so take it as a way to read your own form, not as a finding about theirs. A form with 40 lines can still have a coachable core of eight, and that core is the part a weekly coaching conversation should use.

The length that does cause damage is length that does not fit the call. A line written for an outbound discovery call and pointed at an inbound support queue comes back blank on most cards, and a blank line skews the percentage in ways a report never shows. A long form that fits the call does no harm; the trouble comes from lines aimed at somebody else's conversations.

A default worth starting from, and it is a recommendation, not a measurement: four to seven sections in the order the call runs, three to eight scored lines in each, and a mandatory set small enough that a manager can say it from memory. If nobody can list the mandatory lines without opening the form, there are too many of them.

Weights belong on sections, and a weight you cannot explain is a guess

Put weights on sections, not on individual lines, and give each section a weight you can justify in one sentence about what it costs the business when that section goes wrong. Line-level weights on a 40-line form make each line worth a couple of points, and nobody reading the result can tell which line moved it.

Kixie's sales template is a reasonable shape to start from: opening 10%, qualification and discovery 25%, value presentation 20%, objection handling 15%, closing and next steps 20%, documentation 10%. What carries over is less the numbers than two habits. Discovery gets the heaviest weight, because a call that never establishes the problem leaves the later sections nothing to work with. Documentation stays light, because it says little about the conversation itself.

Then write the one-sentence reason next to each weight. "Discovery is 25% because a call without a stated problem cannot be scoped" is a weight. "Discovery is 25% because the template said so" is a guess, and the first time a rep disputes a score the guess will not survive the conversation.

Our own arrangement sidesteps part of the argument. In Spellit Actions the call score is the plain share of mandatory lines passed, with no weights inside it. Weighting lives in a separate composite that starts at 100 and subtracts for a few named things, such as a missed next step or a critical failure. Keeping the two apart means a rep can read the first number line by line, and a manager can see the second one fall for a single reason. The subtractive score itself is described in the piece on reviewing every call.

Auto-fail lines stop meaning anything once most calls trip one

An auto-fail line overrides the score: one occurrence fails the call regardless of what else went well. That makes it the strongest line on the form and the easiest to overuse. Keep the list to a handful of things with legal or trust consequences, each checkable against the transcript and each within the rep's control.

The published guidance disagrees on what an auto-fail should do. Kaizo's guide to auto-fail rules, published July 27, 2026, says an auto-fail sets the whole evaluation to zero, sits in its own section at the top of the scorecard rather than as a weighted line, and that most teams should be able to justify five or fewer. Kixie takes the softer line: a failed critical item may trigger an automatic review even when the total score is high. Both are defensible. Zeroing suits a regulated script, where one breach is the whole story. Routing to review suits sales, where the same words can be a breach in one deal and a fair answer in another.

What neither guide mentions is what happens to the reporting once the list grows. We report critical failures as an average number per call, not as a share of calls with at least one, and the reason is in the data: on a mature project, the share of calls with at least one critical failure climbs so close to all of them that it stops separating one rep from another, while the average per call still moves. When a form has fewer than two critical lines, we drop the per-line view and fall back to broad categories, because an average over one line is just that line's pass rate under another name.

Voxjar's call scoring page offers a usable alarm: if a third of calls trip an auto-fail, the list is too broad, or one of its lines describes something the rep does not control. Counting how often each of your auto-fail lines trips on real calls takes a week of recordings.

One more detail decides whether auto-fails get counted correctly at all. Some lines are written so that "yes" is the bad outcome: "a technical problem occurred", "a critical error was made". Forms we receive mix both directions, and a report that averages yes answers adds the bad events to the good ones. Either rewrite every line so that yes means the rep did the right thing, or store the direction next to each line and check that every report reads it.

A copyable discovery-call scorecard fits on one screen

The example below is a discovery-call scorecard with five sections, twelve scored lines, two auto-fails and three lines recorded outside the score. The last column is the part to keep even if you change everything else: what a passing quote would sound like. A line without one is not ready to score.

Three things about it. Every scored line describes an event a second listener could find in the recording, which is the property that keeps two scorers from reading the same line two ways. The mandatory set is five lines, short enough to remember. And the rep's proposal of a date and the buyer's agreement to it are two separate lines on purpose: the first is a behavior, the second is an outcome, and only the first goes into the score.

What to do next

Print your current scorecard and write one of five words next to every line: behavior, auto-fail, outcome, label, outside. Count each group. If the outcome, label and outside lines are feeding the score, move them out of it first. Do this before touching any weights, because weights applied to the wrong kinds of line only move the noise around.

Then take the mandatory lines and, for each one, write the sentence a rep would have to say to pass it. Any line where you cannot write that sentence gets rewritten or demoted before the next scoring run. Finally, count your auto-fails. If there are more than five, decide which of them would really fail a call on their own, and turn the rest into scored lines.

Score disputes usually start on the lines nobody can quote. Plenty of forms carry a few lines that have never been passed or failed with a sentence from the recording. Bring the form you use today, and we will find the quote behind each of its lines in your recordings.