Sales manual

The sales manual is TopClosers’ source of truth. Everything that gets scored, compared and recommended comes from it: if your manual says budget must be confirmed before price is presented, that is what the AI looks for in every call. Without a manual there is no analysis.

The Sales manual screen belongs to the agency owner. Closers do not edit it — they see the result in the analysis of their own calls.

What it is

Two different things, and it pays not to mix them up:

The manual

The document that describes how your team sells. You upload it as a PDF, plain text or Markdown.

The rubric

What the AI extracts from the manual: the scorable sections of your sale, each with a weight and success criteria.

The score

Every call is scored section by section against that rubric, from 0 to 10.

The manual supplies the context and the vocabulary; the rubric supplies the rules used to grade. You can edit the rubric as much as you want without touching the document again.

How to set it up

1. Open Sales manual from the side menu. If there is none yet, you get the welcome screen with the three steps of the process.

2. Upload your manual. Drag the file onto the upload area or click to browse. PDF, .txt and .md are supported, up to 20 MB. You can give it a Title so your team recognizes it; leave it empty and the file name is used.

3. Click “Upload and build my rubric” and keep the page open. Reading the document and drafting the rubric takes a few seconds.

4. Review the proposed rubric. It arrives split into two blocks: Phases of the sale and General criteria. None of it is final — it is a draft for you to correct.

5. Adjust names, criteria and weights. The editor footer always shows the Total weight, which must reach 100%.

6. Click “Save rubric”. From then on, every new analysis uses this version.

Scanned PDFs

If the PDF is a scanned image there is no text to read and the upload fails. Export a version with real text (for example, printing to PDF from your word processor) and try again.

Phases and general criteria

The rubric takes two kinds of section. Both are scored the same way, 0 to 10, and both weigh on the final score.

• Phases of the sale. The stages of your process, in order: opening, discovery, presentation, close… Each phase is judged over its own segment of the call.
• General criteria. Standards that do not belong to a single moment: tone, methodology, who does the talking, closing and follow-up discipline. They are judged over the whole call, from the first word to the last, and a slip at the end counts as much as one at the start.

The practical test: if you can point to the minute where it should happen, it is a phase; if it must hold across the whole conversation, it is a general criterion.

They are labelled in the analysis

In a call’s detail view, general criteria are marked Whole call, so it is clear they are not judged over a single segment.

Weights

A section’s Weight is its share of the call’s overall score. The weights of every section — phases and general criteria together — must add up to 100%. Until they do, the rubric will not save.

Distribute weight by what actually decides your sales, not by what takes the most time. A three-minute close can matter more than a twenty-minute discovery.

How to write good success criteria

Criteria are what the AI literally looks for in the transcript. The more observable they are, the more consistent the score.

The recommended format — and the one the AI proposes by default — is an additive point checklist:

• One rule per line, as +N pt: observable behavior.
• Between 5 and 8 rules per section, so the score has fine grain.
• At most +3 pt per rule: the lower the ceiling, the more distinct behaviors a section has to enumerate.
• All the rules in a section must add up to 10 points.

Each rule is evaluated as met, partial or not met, and the section’s score comes out of that tally. That is why a vague rule (“does good discovery”) produces erratic scores, and an observable one (“asks about available budget before quoting a price”) produces stable ones.

A real example

A Discovery section weighted at 25%:

+3 pt: asks at least two open questions about the current situation before pitching anything.
+2 pt: quantifies the problem in money, time or volume with a concrete figure given by the lead.
+2 pt: explicitly asks about available budget or investment range.
+2 pt: identifies who else is involved in the decision.
+1 pt: summarizes what was heard and asks for confirmation before moving to the proposal.

And a general criterion called Control of the conversation, weighted at 10%:

+3 pt: the prospect talks more than the closer during the call.
+3 pt: no monologue by the closer runs longer than two minutes.
+2 pt: every objection gets a question before it gets an argument.
+2 pt: the call ends with a concrete next step, with a date.

Write the way you speak

Use your manual’s vocabulary. If your methodology calls “diagnosis” what others call “discovery”, write it that way: the coaching your closers receive will use your words.

Editing, reordering and regenerating

• Add a phase or Add a general criterion creates a new section at the end of its block. A rubric takes up to 16 sections and needs at least one phase.
• The arrows on each card move a section up or down. The order of the phases should match the real order of your sale.
• The trash icon deletes the section.
• Regenerate with AI re-reads your manual and proposes a fresh set of sections. It asks for confirmation because it overwrites what is on screen, including your edits.
• Replace manual uploads a new document and creates a new version of the rubric.

Regenerating is not undo

If you have invested time fine-tuning criteria, stay away from “Regenerate with AI”: there is no way to recover the previous phases. For small changes, edit by hand.

Versions

The manual header shows the active Version and the upload date. Every time you replace the manual a new version is created, and calls from then on are scored against it.

What matters is what happens to what was already analyzed:

• Calls that were already analyzed are never rescored. They keep the score they got with the rubric they were evaluated against.
• Editing or deleting a section does not erase the past. Every section score stores the name, kind and weight it had at the time of the analysis, so an old call’s report still reads correctly even if that phase no longer exists today.
• Changes only affect what comes next. If you want an old call scored against the current rubric, re-analyze it from its detail view (see Call analysis).

This is what lets you compare a closer’s progress without a rubric change distorting the history — but it also means that, right after a big change, new and old scores are not strictly comparable.

Things to know

• The screen is owner-only.
• No active manual means no analysis. Upload it before connecting the recorders, or your first calls will sit waiting.
• Weights must total 100%, every section needs a name and criteria, and the rubric needs at least one phase of the sale.
• Phase names and all coaching are written in the agency language, not in the document’s language. You can change it in agency settings.
• The rubric belongs to the whole agency: there are no separate rubrics per closer or per product.

Best practices

• Upload the real manual, not a summary. The more detail the document carries (scripts, common objections, things never to say), the better the AI interprets your criteria.
• Start with few sections. Six or eight well-written ones beat sixteen vague ones.
• Review the rubric after the first ten calls. If a section always scores 10 or always scores 3, the problem is usually the criterion, not the team.
• Change the rubric in batches, not weekly. Every change cuts the historical series; group your adjustments and apply them at once.
• Share the criteria with the team. A closer who knows exactly what is being scored improves faster than one who only sees the number. See Coaching for the follow-up.