{"id":11556,"date":"2026-10-27T09:20:20","date_gmt":"2026-10-27T00:20:20","guid":{"rendered":"https:\/\/blog.themusio.com\/?p=11556"},"modified":"2026-10-27T09:20:22","modified_gmt":"2026-10-27T00:20:22","slug":"before-you-upload-your-first-real-call-what-a-vp-of-sales-should-know","status":"publish","type":"post","link":"https:\/\/blog.themusio.com\/?p=11556","title":{"rendered":"Before You Upload Your First Real Call: What a VP of Sales Should Know"},"content":{"rendered":"\n

Uploading a recording of a real customer call is the part of EOS<\/strong> a VP of Sales is most likely to care about, and the part with the most things worth knowing first. Our earlier walkthrough covered what happens after a call is uploaded. This guide is about before: what you’re agreeing to when you pick a file, whose name the call lands under, who gets scored, and where the limits are.<\/p>\n\n\n\n

It isn’t an argument for uploading customer calls. It’s a list of what the product does today, so you can decide with clear eyes.<\/p>\n\n\n\n

Who can upload, and whose name it’s filed under<\/strong><\/h2>\n\n\n\n

Anyone signed in on your team can upload a recording, whether they’re an owner, an admin, a trainer, or a trainee. The call is always filed under the person who uploaded it.<\/p>\n\n\n\n

That has a consequence worth stating bluntly: a manager can’t upload a call on a rep’s behalf.<\/strong> There’s no “upload for this rep” control. If a manager uploads a recording of a rep’s call, it shows up under the manager’s name, and the rep doesn’t see it as their own. It also won’t appear in that rep’s rows or in that rep’s needs-attention flags, because attribution follows the uploader, not the voice on the recording.<\/p>\n\n\n\n

If your plan is “managers collect calls and upload them for the team,” know that this isn’t how it works today. In practice, each rep uploads their own.<\/p>\n\n\n\n

What the upload asks for, and what it doesn’t<\/strong><\/h2>\n\n\n\n

The upload is a file picker. You choose an mp3, wav, m4a, or mp4 file of up to 500 MB, one at a time, and then select Upload & score. There’s no bulk upload.<\/p>\n\n\n\n

It doesn’t ask which rep was on the call, which product it was about, or what day it happened. The date on the session is the day it was uploaded, and the length comes from the transcript. Which products get checked is worked out afterward from what the rep actually says, against your approved products only.<\/p>\n\n\n\n

There’s no cap on call length. What limits you is the file size, and how many minutes fit in 500 MB depends on how the file was recorded. Compressed audio like mp3 or m4a fits hours. Uncompressed WAV or video can reach the limit in under an hour. An unsupported type or an oversized file is rejected before upload, with a message saying so.<\/p>\n\n\n\n

Who gets scored: speaker labels<\/strong><\/h2>\n\n\n\n

This is the most important thing in the guide.<\/p>\n\n\n\n

EOS doesn’t ask you to mark which voice is the rep and which is the customer. It works that out automatically from the words, deciding who sounds like the customer (asking, reporting a problem) and who sounds like the rep (helping, explaining). It expects two voices.<\/p>\n\n\n\n

It’s a guess, and it can be wrong. Nobody can correct a label afterward, and the report doesn’t warn you when a label is shaky. Soft skills and claim checks both use whoever was labeled the rep, so a wrong label means the wrong person’s words are scored. A third voice, such as another colleague joining, isn’t a supported case. Their words may be folded into one of the two speakers or attached to the previous speaker’s turn, and that can pull a colleague’s or a customer’s statements into the rep’s score.<\/p>\n\n\n\n

The practical takeaway is modest. Calls with two people are what EOS is built for. We can’t say that separate stereo channels or avoiding speakerphone would improve it, because we haven’t measured either, so we’re not going to advise them.<\/p>\n\n\n\n

\n

See what the verification behind a call report checks.<\/strong> Our 12-page report, Beyond Roleplay: The Rise of the Sales Knowledge Engine<\/em>, covers claim extraction, fact-grounded verification, and the questions worth asking any vendor about how real conversations are evaluated. [Download the whitepaper \u2192]<\/strong><\/a><\/p>\n<\/blockquote>\n\n\n\n

What gets scored<\/strong><\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

Uploaded calls are scored on the same criteria sets your organization uses for practice. There’s no separate rubric for real calls, and items written with roleplay in mind are still scored when their set applies. Your organization’s sales context setting (walk-in, inbound, or scheduled) shapes the practice customer only. It isn’t used when a real call is scored, and customer demeanor doesn’t apply to a real call at all.<\/p>\n\n\n\n

Claims are checked against your approved products only. If a rep talks about a product that’s still pending, or that isn’t in your catalog, the claim tends to come back Unsupported, the same as in practice, unless an official outside source supports or contradicts it. If an uploaded call produces a pile of Unsupported results on one product, check whether it’s still waiting for approval.<\/p>\n\n\n\n

Language and audio quality<\/strong><\/h2>\n\n\n\n

The language is detected from the audio. It isn’t restricted to the eleven training-content languages, and it isn’t taken from your organization’s training language. A call that mixes languages gets one detection for the whole file, with no special handling and no tested rule of thumb for how well it works. The written product-knowledge summary still comes out in your organization’s training language, even when the call was in another one.<\/p>\n\n\n\n

Transcription is best-effort. EOS doesn’t check phone-line quality, background noise, or accents, laughter and other non-speech sounds are dropped from the text, and nobody can edit the transcript afterward.<\/p>\n\n\n\n

One limit applies to Japanese and Korean calls specifically, and it’s narrower than it might sound. Soft-skill scoring breaks the rep’s words into sentences using English-style punctuation (a period, exclamation mark, or question mark followed by a space). It doesn’t treat a Japanese full stop as a sentence end, so a Japanese or Korean line can be merged or sliced in the wrong place, and a criterion can land on the wrong slice. Claim checking is separate. It reads each turn whole and isn’t split that way.<\/p>\n\n\n\n

When an upload goes quiet<\/strong><\/h2>\n\n\n\n
\"\"<\/figure>\n\n\n\n

Two kinds of upload don’t produce a normal report.<\/p>\n\n\n\n

A very short call<\/strong> still appears, with its transcript and no score, and nobody can score it by hand.<\/p>\n\n\n\n

A failed transcription<\/strong> doesn’t appear at all. If the file can’t be read, or the transcript comes back empty, no session is created and nothing says it failed. You see only the first message that scoring will run in the background, and then the call never shows up. There’s no failed state to open, and nobody is alerted inside the product. A missing call is something you have to notice.<\/p>\n\n\n\n

Who can see an uploaded call<\/strong><\/h2>\n\n\n\n

The person it’s filed under can see it and its report, including the transcript and an audio player when the original was stored. Other reps can’t see it. Owners, admins, and trainers can open every uploaded call in the organization. In the sessions list, a Real Call badge appears where a practice persona’s name would be. The organization dashboard itself is limited to owners and admins.<\/p>\n\n\n\n

How uploads show up in your numbers<\/strong><\/h2>\n\n\n\n

Scored uploads aren’t reported separately. A real call that receives a soft-skill score counts toward the same pass rate, team average, and needs-attention flags as practice sessions, and it counts as activity, which can clear an inactivity flag. A too-short call without a score doesn’t move pass rate or team average, though it still counts as activity.<\/p>\n\n\n\n

You can’t filter the sessions list to uploaded calls only, and there’s no view that places one rep’s practice and real-call results side by side. That’s worth knowing if you read pass rate as a measure of readiness for real conversations: it mixes both.<\/p>\n\n\n\n

Before you upload customer calls<\/strong><\/h2>\n\n\n\n

Questions about recording consent, personal data in transcripts, deletion, and retention aren’t settled by this guide. They’re decisions for your own legal and privacy teams, and a conversation to have with us before you start, not after. Please don’t treat anything here as an answer to them.<\/p>\n\n\n\n

A short checklist<\/strong><\/h2>\n\n\n\n