Incoming call notifications arrive as the phone rings. The caller is matched against the CRM, so the notification names the client and the person taking the call rather than showing a bare number. A rep sees who is calling before picking up; a manager sees that an account worth watching is on the line. Whether caller identity appears in notifications at all is a setting each company controls.

Review assignment makes checking the analysis routine. A configurable share of calls is automatically assigned to a reviewer — normally the rep’s own manager — who is notified and can work through them from a filtered list. Because the sample is random, the corrections that come back show where the analysis holds up and where it drifts, across ordinary calls rather than only the ones that already looked wrong.

Caller identification was rebuilt underneath both. Numbers arriving in different formats now resolve to the same client, which is what keeps client history, notifications and CRM matching pointing at one company rather than three.

One change under the surface: the separate alert for missed buying signals has been retired as its own mechanism. It now ships as a pre-defined notification rule running on the standard alerting system — the same alert, arriving the same way, but adjustable like any other rule and routed to whoever should receive it.

Sales Signals now assigns calls for review on its own. You set the share of calls to be checked, and that fraction is drawn at random and assigned to a reviewer, normally the rep’s own manager. The reviewer is notified, and the calls waiting on them can be filtered out of the call list.

Because the sample is random, the corrections show where the analysis holds up and where it drifts, on ordinary calls rather than on the handful that stood out for some other reason. Everything that follows from individual calls — weekly summaries, client trends, recurring themes — relies on getting that analysis right.

Now, the moment a call comes in, the operator who picks it up gets a push notification with a link straight to that client’s page. One tap, and the context is there: call history, open tasks, relationship summary, focus for the next call, and a sentiment chart showing how the relationship has been trending. Call details remain accessible only to those who already have access.

None of this needs to be prepared in advance. It is built from the calls you’ve already had and arrives within seconds — waiting for the moment it becomes useful, just before you say hello.

Transcript ratings and phrase corrections strengthen the foundation. Speech recognition handles ordinary conversation well, but sales calls are full of model numbers, product names, and team-specific shorthand. Transcripts can now be rated and incorrect terms corrected, with corrections improving transcription of later calls. Quality ratings are also tracked alongside analysis, making it easier to distinguish genuine sales signals from problems with the source material.

More precise theme grouping makes those signals more meaningful. Objections and questions are now grouped more precisely, so distinct concerns are less likely to disappear into one broad theme, while labels better describe what the calls actually have in common.

Objection and question breakdowns make these patterns actionable. On a client page, managers can see which objections and questions came up across the relationship, whether they were resolved, and jump directly to the relevant call. The same breakdowns can be used across the call list — for example, to find every deal where budget came up this month.

Notification rules you define yourself put the right signals in front of the right people. Rules use the same filters as the call list — a specific rep, risk level, or missed signal — and can be assigned to individual recipients and delivered by email or browser push. Churn risks can go to the account owner, coaching flags to the team lead, without turning every alert into everyone’s problem.

August made Sales Signals more reliable, more precise, and more controllable: better source material, sharper analysis, deeper investigation, and notifications that fit the way each team works.

You can now define your own notification rules, built from the filters you use on the call list — a specific rep, a risk level, a missed buying signal, or a combination of them. If you can find the calls, you can build the rule.

Each rule has its own recipient, so churn-risk calls can go to the account owner and coaching flags to the team lead, rather than everything landing in one inbox that everyone eventually mutes. Recipients only ever receive calls they could already open themselves — a rule can’t widen anyone’s access.

Notifications arrive by email, by browser push, or both. Push lands within seconds of a call being analysed, for the alerts worth acting on the same day. Email bodies are editable, with a preview, and every rule can be tested with a single click.

You can now rate the quality of any transcript and submit the terms that came through wrong. Corrections feed back into how later calls are transcribed, so the vocabulary specific to your business gets more reliable the more your team flags it.

The ratings matter beyond the individual call. Transcript quality is tracked alongside every analysis, which means a run of poor recordings on one line or one handset shows up as a pattern rather than as unexplained noise in the results — and it tells you when a weak analysis was a transcription problem rather than a call problem.

Both take a few seconds from the call page, and neither requires anyone to listen to the recording again.

That grouping has now been rebuilt. Themes are separated more precisely, which means fewer distinct concerns collapsed into one broad heading and fewer variations of the same concern split across several. Labels are generated from the group as a whole, so what you see on a client or rep page describes what those calls actually have in common.

The rebuild also handles calls that move between languages mid-conversation — common on export and cross-border accounts, where the same objection can appear in two languages and previously counted as two separate themes.

Client pages now break out the objections raised and the questions asked across every call with that client, each marked as resolved or unresolved, answered or left hanging. Open any one of them and you go straight to the call it came from, at the point it was raised.

The same breakdown works across your whole call list. Filter by objection type to see every deal where budget came up this month, or by question type to find the technical questions your reps are fielding but not closing out.

Together they answer a question a call-by-call view can’t: not what came up, but what keeps coming up and never gets dealt with.

Client pages bring all conversations with the same client into one view. Instead of looking at calls in isolation, managers can now see how sentiment changes over time, how frequently the client is contacted, and how the relationship is developing. Signals that are easy to miss in individual calls become visible across the entire relationship.

Client insight notes turn that history into a short, actionable briefing: where the relationship stands and what to focus on next time you speak. Useful when you have twenty seconds before the phone rings — or hundreds of clients to keep track of.

Theme grouping makes patterns across conversations easier to understand, grouping objections and questions by meaning rather than identical wording. The objection your team hears most often no longer hides as twenty separate one-offs.

HubSpot support provides the CRM context needed to make client-level analysis work. Calls now resolve to real CRM records rather than phone numbers, so a company calling from three different lines is still one client.

Together, these updates add the client relationship to the picture Sales Signals already provides — from how the team performs, to what is happening across conversations, to what is happening with each individual client. That makes it possible to spot relationships that are cooling down and act before they turn into lost business.

Every client now carries a short written note: where the relationship stands, what’s unresolved, and what to focus on next time you speak to them. It’s generated from the full call history with that client, not from the last conversation alone.

It’s most useful in the moments where nobody has time to read back through months of calls — a rep taking over an account, a manager stepping into a difficult client before a meeting, or anyone returning to a client they haven’t spoken to since spring.

The note sits at the top of the client page, above the trends and breakdowns. Read it in twenty seconds, or keep going for the detail behind it.

Client pages change the unit of analysis. Every call with the same client is now aggregated into one view: how sentiment has moved over time, how often you’re in contact and whether that’s rising or falling, which stage the relationship sits at, and what’s still open.

Clients are matched through your CRM rather than by phone number, so the history holds together when someone calls from a different line or a switchboard.

Today’s calls are included as they’re analysed, not held back until an overnight job runs — so the view you open in the afternoon reflects the morning.

This is the foundation for early churn warning: the signals that a relationship is cooling appear weeks before the orders stop, but only if you’re looking at the relationship rather than the call.

Sales Signals now groups objections, client questions and coaching points by what they actually mean. “Too expensive”, “that’s above our budget” and “we’d need to find the money” land in one theme, counted once, with a label generated from the group describing what those calls have in common.

The effect is that the frequency numbers start being worth reading. A theme that shows up in a quarter of your calls is visible as a theme, rather than distributed across twenty near-identical entries that each look insignificant on their own.

Connecting takes an access token and nothing else. The connection is verified during setup, so you find out immediately whether it works rather than discovering days later that tasks were never arriving.

The more consequential part is contact matching. Calls are resolved to real CRM records rather than to phone numbers, which is what makes client-level history trustworthy — the same company reaching you from a switchboard, a mobile and a direct line is one client, not three. Contact names are pulled through in full, so a client page shows who you actually spoke to.

HubSpot joins Bitrix24 as a supported CRM, and both use the same integration pattern underneath, so further platforms are additive rather than bespoke.

Alerts for missed buying signals fire on the individual call rather than waiting for a monthly review. When a client shows clear buying intent and nobody acts on it, the manager hears about it while the conversation is still recent enough to pick up — a quote can still go out that afternoon, and the rep still remembers the call well enough for coaching to land.

Follow-up tasks in your CRM handle the commitments that get made and then forgotten. When a call ends with an agreement — send the quote, call back Thursday, check stock — the task is created in the CRM, assigned to the person who owns it, with a link back to the call. Where no clear commitment was made, no task appears, so nothing invented ends up in anyone’s pipeline.

Review and correct call analysis gives managers the other half of that: the ability to disagree. Any part of a call’s analysis can be approved or corrected in place, and the correction becomes the version everyone sees. Trends, comparisons and summaries are built on individual calls, so being able to fix one in a few seconds keeps everything above it honest.

Weekly summary improvements extend the same idea to the reports themselves. Summaries are now broken into sections that can be rated or flagged when a conclusion is wrong, and they track the calls behind them — if those calls change, the summary is marked out of date and can be regenerated.

Taken together, June made Sales Signals answerable. The system surfaces what matters while it can still be acted on, turns agreements into tasks without anyone writing them down, and lets the people who know the client best correct it when it gets something wrong.

They’re now broken into sections, and each section can be rated or flagged when the analysis gets something wrong — so disagreeing with one conclusion doesn’t mean discounting the whole summary.

Summaries also track the calls they were built from. If those calls change, the summary is marked out of date and can be regenerated.

Open a call and you can work through its analysis in place: approve the parts that are right, and change the parts that aren’t. Correct a score, a risk level, a call type, a missed signal. Your correction replaces what the system produced and becomes the version everyone sees from then on — it isn’t logged as a complaint and left sitting next to the original.

That matters most for the numbers built on top of individual calls. Weekly summaries, rep comparisons and client trends are only as good as the analysis feeding them, and a manager who can fix a wrong call type in five seconds keeps those aggregates honest without having to distrust the whole system.

It also makes the daily review a genuine workflow rather than a chore. Ten minutes a morning going through yesterday’s calls leaves the record accurate, the reps’ feedback fair, and any pattern of the system getting something consistently wrong visible rather than tolerated.

Sales Signals now creates the task for you. When a call ends with a commitment — send the quote, call back Thursday, check stock and confirm — that becomes a task in your CRM, assigned to the person who owns it, with a due date taken from what was actually agreed on the call. The task carries a short summary of what was discussed and a link back to the call, so whoever picks it up doesn’t need to ask.

Nothing is invented. If the call ended without a clear commitment, no task is created — a vague “we’ll be in touch” doesn’t become a fake deadline in someone’s pipeline. That absence is worth knowing about too, and it’s why the same call shows up as a next step that was never fixed.

Tasks that fail to reach the CRM are visible and can be sent again, so a connection problem doesn’t quietly cost you a week of follow-ups.

Sales Signals now raises an alert when that happens. If a call contains a clear buying signal that went unhandled, you’re told about it, with the client, the rep and the call attached.

The timing is the point. This isn’t a number that drifts on a dashboard until someone reviews the month — it fires on the individual call, while the conversation is still recent enough to pick up. A quote can still go out that afternoon. A rep can still be told the same day, when they remember the call well enough for the coaching to land.

One call ending without a next step is normal. Next steps being fixed on half as many calls as last month is a different thing entirely — and it’s the kind of change nobody spots in time, because no single day looks wrong.

You can now put a target on any metric Sales Signals tracks, say which direction is bad, and be notified when performance crosses it. Alerts follow the trend rather than individual calls, so a slow week doesn’t trigger anything and a genuine slide doesn’t wait for someone to open a dashboard and notice.

Once an alert has fired, it stays quiet until something actually changes. Repeated warnings about a problem you already know about are how people learn to ignore alerts, and a metric you’re already dealing with shouldn’t keep interrupting you.

Metrics without a target are marked as such, so it’s clear what’s being watched and what’s only being displayed.

Scored as a sales call, that’s a failure. No close attempt, no deal progression, no budget discussed. Scored as what it was, it’s close to perfect.

Sales Signals now evaluates each rep against their own role. Every person is assigned one — salesperson, presale engineer, support engineer or account manager — and their calls are judged on what that role is there to do. Closing discipline for salespeople. Qualification and clean handoffs for presale. Resolution, empathy and escalation for support. Retention and expansion for account managers.

Metrics that don’t apply to a role aren’t reported as zeros or weak scores. They’re left out, because their absence is correct rather than a gap.

This matters most in the businesses where support and presale handle more calls than sales does. Judging that volume by a sales rubric doesn’t just produce unfair scores for the people on those calls — it makes the team-wide numbers meaningless, and the coaching built on them worse than useless.