Messaging

Messaging measures your company's messages in real sales and customer conversations. A leader declares what the team should say, or ArcGlass discovers what the team is actually saying. Every call is then read for two things: whether the message was said (delivery) and what the buyer did right after (response). Each message lands on one chart and gets one action: keep it, lead with it, push it, fix the words, or retire it. Every count opens to the moments behind it, in the buyer's own words.

Key facts

• What it is: message measurement from conversations. For each message: is it being said, to how many accounts and by how many reps, and what do buyers do when they hear it.
• Built from: your meeting and call transcripts, your website, and the email threads where buyers pass your message on to colleagues.
• Who it's for: CEOs, COOs, CMOs and VPs of marketing, product marketers, and the sales leaders who own what the field says.
• Evidence or nothing: every delivery and every reaction keeps the exact words, found in the turn of the conversation they came from. A quote that can't be found is dropped.
• Measured against your own company: delivery against each message's own target, response against how your own buyers usually respond. No industry benchmark.
• No scores on screen: results are counts of real accounts and reps, in words, plus one action.
• Seven declared messages at most, so evidence concentrates instead of spreading thin.
• Any language: matching is by meaning, so a call in Portuguese can match a message written in English.
• Where: in the console's left navigation, under What you hear → Messaging.

The problem it solves

Every company writes messaging. Almost none can tell whether it's being said, and none can tell what buyers did when they heard it. Three things happen constantly, and no other tool names them:

1. Leadership changed the pitch and the field didn't. The deck says one thing in March; reps are still saying the old thing in September.
2. A message works and nobody says it. Buyers engage every time they hear it, and it comes up in one account in ten.
3. A message is said constantly and does nothing. It's on the homepage, every rep opens with it, and no buyer has ever responded to it. This one is the most expensive, because volume looks like success.

Questions it answers

• Which message is landing, and which accounts prove it?
• Which message does everyone say that no buyer responds to?
• Which message works when it's said, but is rarely said, and what words should reps use?
• Which message do buyers dispute?
• Who is still saying the message we retired?
• Do deals that heard a message move on more often than deals that didn't?

The words Messaging uses

How it works

Two ways in, one loop.

1. Declare, or discover. A leader declares up to seven messages. ArcGlass also reads what your reps actually promise on every call, and what your website says, and groups them into discovered messages.
2. Every conversation is read. As each call is processed, ArcGlass finds the turns closest in meaning to each tracked message, and one AI read per call decides what was said and what came next.
3. Delivery means a real promise. A delivery is someone on your side making the message's promise. Narrating a demo, answering a factual question, logistics, or repeating the buyer don't count. Each delivery is marked as declared, part of it, or drifted.
4. Response is what the buyer did next. The reaction comes from the buyer's turns right after the delivery, and is judged on what the buyer did, not on enthusiasm words.
5. Every quote is checked. A delivery or a reaction must quote the transcript word for word, in the turn it cites. A delivery whose quote isn't there is dropped; a reaction whose quote isn't there becomes "no reply".
6. Unreadable isn't negative. When a recording doesn't separate the buyer's turns from yours, the delivery still counts, but no response is read, so silence is never counted against a message.
7. One action each. Every week, each tracked message is placed on the chart and given its action, and the action becomes a task with an owner.

Reactions, best to worst

The first two count as engaged. Each account counts once, with its best reaction: one account saying "that's the report my CFO wants" once and "okay" four times is an account that named what it changes. A talkative account can't outvote a quiet one.

Declaring messages

Declare a message opens a full-screen session: "What are you asking the field to say?"

1. Blind first. You type your top messages before seeing any suggestion.
2. Suggestions, one at a time: claims from your homepage and product pages first, then the strongest discovered messages. Each shows a rep line, a website line and a buyer line, and how many accounts it reached. Accept, edit or skip it.
3. For each message: its audience (personas, stages and segments), a start date, the message it replaces if any, its owners, and a product area.
4. The look-back. Saving reads the message against every conversation you've already recorded, with progress as it goes, so you don't wait a quarter for new calls. You can leave; a notice says when the read is ready.

The counter stays in view, "3 of 7". Past seven, evidence stops concentrating and every message gets a thin read.

The Messaging page

"What we say, whether buyers hear it, and what to change."

• Filters: the time window (30 days, 90 days by default, 12 months, all time, or since the message's start date), segment (all accounts, prospects or customers), team, and origin (declared, discovered or both).
• Three numbers: Tracked messages ("5 declared · 3 discovered"), Reach (accounts that heard any tracked message, "of N with a meeting"), and Strong reactions (accounts that named what it changes, "of N that heard one").
• The brief: three sentences built from the counts: the strongest message, the top message to push or fix with its reason, and the top message buyers raised themselves or disputed. Copy it into a staff meeting or board update. A sentence with nothing to say is left out.
• The chart and the scoreboard, side by side (below).
• Callouts: "Off-script and working" (discovered messages that work), "Buyers raise it, nobody says it" (buyers ask for something you can already say), and "Buyers dispute it" (with both quotes).

Until there's enough to read, at least 30 conversations across at least 10 accounts, the page says reads appear as calls are processed. You can still declare messages, so intent is in place before the data arrives.

The chart: delivery against response

"Delivery and response together say who has to act." Each dot is a message, labeled with its name.

• Up (Delivery, rare → wide): how widely it's said, against the message's delivery target (40% of its audience by default).
• Right (Buyer response, weak → strong): how often buyers engage, against your company baseline: how your buyers usually respond across all your messages.

The lines sit at the delivery target and the company baseline; hover them to see which is which. They carry no numbers, because a share on an invented scale can't be argued with, but "9 of 14 accounts engaged" can be opened and checked. A message exactly on a line counts as up or right.

A message heard in fewer than three accounts has no place yet. It sits in a band under the chart, Not enough data yet, because no data and a weak response are different facts.

One action per message

ArcGlass goes down this list and gives each tracked message the first action that fits:

1. Retire: the message was retired or replaced, and someone still said it in the last 30 days.
2. No action: it was heard in fewer than three accounts.
3. Fix: at least two accounts, and at least 30% of those that heard it, pushed back. Pushback overrides the corner: a promise buyers argue with isn't working, however widely it's said.
4. Lead with: it's in the Keep corner but not on your homepage or product hero.
5. Keep, Push or Fix, matching its corner. The Not enough evidence corner gets no action.

Every action carries one plain reason built from real counts, for example "Works when said; said in 5 of 31 accounts in its audience."

Owners default to whoever declared the message.

The scoreboard

One bordered row per tracked message, with the messages that need someone to act first:

Discovered lists the messages your team says that nobody declared, ranked by how many accounts heard them, with the long tail under "Emerging". Each shows its accounts, reps, whether it's on your site, its response, and whether it's new this week. Track starts measuring one; you can also rename, merge, split or dismiss them.

One message, up close

Click any message to open its panel:

• What it is: the statement, whether it was declared or discovered, its action and owner, its audience, start date, and the message it replaces.
• Four numbers: heard (accounts and reps), engaged, pushed back, and retold.
• How buyers reacted, in words: "4 named what it changes for them, 5 gave a reason, 2 asked how, 3 acknowledged, 2 pushed back".
• The trend since its start date, week by week, marked where it was declared, where your website changed, and where the team was enabled on it.
• The accounts that heard it, each labeled as a customer, in evaluation or in discovery.
• Every delivery, newest first: the company, date and rep; the rep's words; how closely they matched the message; and the buyer's words with their reaction, or "no reply". Play the recording from that moment. If ArcGlass got one wrong, mark it Not this message or Wrong reaction; a correction is never overwritten by a later re-read.
• Buyer language: up to five phrases buyers use for the same idea.
• Best deliveries: the three with the strongest reactions, each with Copy as talk track.
• On the site: the pages it appears on, how prominently, and since when.

On each account page, a Messaging tab shows which tracked messages that account heard, how it reacted to each, and when: "Heard 4 tracked messages across 6 meetings; engaged with 2."

The talk-track loop

The point of Messaging is to change what reps say next, not to report on what they said last month.

1. A message lands on Push: buyers engage when they hear it, and reps rarely say it.
2. Its words owner sets a talk track, usually one of the best deliveries, edited if needed.
3. Before the next meeting, the rep's brief shows it under Worth saying when the account is in the message's audience and hasn't heard it yet: the statement, the talk track, and one line of evidence, such as "9 of 14 accounts engaged when they heard it". At most two per brief.
4. Say this instead. When a rep said a retired message in the last 30 days, their own next brief for that audience shows the new talk track in its place. It goes to that rep only, never a list of reps.
5. The loop closes itself. Delivery is measured again every week. When a message moves from Push to Keep, its Push task closes with a note, such as "Now said in 19 of 31 accounts."
6. Said more, landing less. If a message is said more after its talk track goes out but buyers engage less, its words owner is told: the wording is now the problem, not the reach.

Tasks

Each action becomes a task in ArcGlass Tasks, owned as above: "Push: …", "Fix the words: …", "Lead with on the homepage: …" or "Retire: …", followed by the statement. The task carries the reason and the two strongest deliveries with their buyer lines. There's one open task per message and action: a new action closes the old task with a note, and Keep closes any open task.

Deal movement

When a CRM is connected, Messaging compares deals that heard a message with comparable deals that didn't, at the same stage and in the same segment, and counts how many moved on at least one stage within 30 days: "Moved on: 12 of 20 deals that heard it, 6 of 21 that didn't." It's shown as two counts side by side, never as a lift, a rate or a cause, and only when each side has at least five deals.

Where Messaging shows up elsewhere

• Briefs: "Worth saying" and "Say this instead", before the meetings where they matter.
• Engage: each deal shows the messages its account heard and how it reacted.
• Product Signals: a product area shows "What we say about this", the tracked messages in that area with their response. Demand the company can't yet answer stays in Product Signals as an unmet need instead of becoming a message.
• Landscape: once a competitor comes up in at least three accounts that heard a message, the message's response is split by whether that competitor was in the conversation.
• Pitch: Build a pitch from this opens the pitch studio with your tracked messages, strongest first.

How it stays current

• Every call, as it's processed: read for deliveries, reactions and raises, and the affected messages are updated.
• Every email thread, as it's analyzed: read for retells.
• Every week: discovery finds new messages your team is saying, your website is read again for changes, and a snapshot is saved for the trend.
• Every night: every window, the company baseline, the chart, the actions and the tasks are recomputed.
• When you declare a message: the look-back reads it against every conversation you've recorded.

Who can do what

• Anyone who can read conversations can open Messaging and copy talk tracks. It's a workspace-wide view: its counts and quotes cover every conversation that's been read.
• Members, Managers and Organization Owners can declare, edit, track, retire, merge, split and dismiss messages, correct deliveries and reactions, set talk tracks, and create tasks.
• Managers and Organization Owners can also change delivery targets.

When something isn't connected

Most companies don't have every source connected, and Messaging is built for that. Anything that isn't connected is left out rather than shown as a zero:

• No CRM: no Deals column and no deal movement.
• No email: no Retold.
• No website read: no site chips, and Lead with gives way to Keep.

Frequently asked questions

Glossary

• Message: one promise your company makes, in a sentence or two.
• Declared message: a message a leader stated, with an audience and a start date.
• Discovered message: a message your team says that nobody declared.
• Delivery: one time a rep, an email or your website put the message in front of a buyer.
• Reaction: what the buyer did right after a delivery.
• Engaged: named what it changes, or gave a reason it matters.
• Raise: the buyer brought the message up first.
• Retell: the buyer passed the message on to a colleague in email.
• Delivery target: how widely a message should be said, 40% of its audience by default.
• Company baseline: how often your buyers usually engage with your messages.
• Talk track: the best way to say a message, for reps to use.

Related

• Product Signals: what customers ask for and what lands, from the same calls.
• Landscape: the companies your customers mention.
• Conversation types: each call read against its playbook.
• Tasks: where Messaging's actions go.