Pipeline insights
Pipeline insights is the workspace’s own answer to the question every sales leader eventually asks about an agentic setup: it produced all this, so what happened next. It counts two things side by side - what your digital colleagues produced in a period, and how far that work actually travelled before a human picked it up - and then names the specific place the chain is leaking.
What it measures
Section titled “What it measures”Supply is what the agents delivered: accounts qualified as a fit, contacts qualified as being in the buying group, buying signals found, and the research hours that work replaced, converted at an agreed minutes-per-contact rate that is stated on the figure itself.
Delivery follows the same cohort forward, stage by stage, to the point a person touched it. Accounts run from qualified, to carrying a positive buying signal, to having a contact who can actually be reached, to being worked by a rep. Contacts run from qualified, to reachable, to worked by a rep. Every later stage sits inside the one before it, so each step is a conversion rate rather than a separate count.
The words carry precise meanings, and they are worth knowing before you read a number:
- Qualified means a qualification agent returned a positive verdict during the period. If an account is researched again later, that counts again in the later period.
- Carrying a buying signal means research answered at least one of your buying-signal criteria positively.
- Reachable means the contact can be contacted today: a primary email verified as safe to send, or a phone number on file. Addresses that came back risky or catch-all do not count, because sending to them is what damages a domain.
- Worked by a rep means a human logged an email, a call, or a meeting in the CRM on or after the latest qualification. Outreach that never reaches the CRM, LinkedIn messages being the common case, is invisible here.
Alongside the chain it splits your reachable contacts by channel - emailed only, called only, both, or never contacted - and tracks new opportunities created in the period, how many of them sit on accounts your agents sourced, and the pipeline value attached to them.
Why it exists
Section titled “Why it exists”Agentic prospecting fails quietly in a way that traditional prospecting does not. When a team builds its own list, the person who built it also works it, so supply and effort move together. When agents build the list, they can keep delivering perfectly good accounts into a workspace where nobody opens them, and every dashboard still looks busy. The gap does not announce itself: credits get spent, records get created, research gets written, and the meeting count stays flat.
That is the gap this measures. A qualified account that no rep ever touched is worse than a missing account, because you paid for it and it created the impression of coverage. Reading supply and delivery on one page turns an argument about whether the agents are working into a specific, checkable claim: they delivered this many reachable contacts, the team touched this many, and here is the difference.
It also settles which lever to pull. Weak supply is a targeting problem, and the fix is in your criteria and your ecosystems. Strong supply with weak delivery is a capacity or habit problem, and the fix is how work is split across the team. Those two failures feel identical from the outside and need opposite responses.
The three checks
Section titled “The three checks”Three plain rules run over the period’s numbers, each landing on a green, amber, or red state, or standing down entirely when there is not enough data to judge:
- List quality - the share of the accounts you qualified that actually scored as a fit. A low share means the agents are spending their runs on accounts that were never going to qualify, which is a targeting conversation, not an effort one.
- Activity volume - the share of your reachable contacts that nobody ever touched. This is the leak that matters most, because those contacts were fully paid for and fully ready.
- Channel mix - whether outreach is concentrated in a single channel, and how many contacts with verified phone numbers were never called. Single-channel outreach is the most common reason a well-built list produces silence.
These are arithmetic against fixed thresholds. No model is asked to guess a number or issue a verdict anywhere on the page: a check reports a finding, and Eva turns that finding into an offer of help you can accept or ignore.
Each check is also a door, not just a verdict. A finding names a specific set of records - the accounts that did not score as a fit, the ready contacts nobody ever touched, the phone numbers nobody ever called - and every one of those sets can be opened as an actual list, or handed to Eva to work on, without anyone reconstructing the definition by hand. That is deliberate: a diagnosis nobody can act on the same afternoon becomes a number people learn to scroll past. The two touch-based checks are counted the way you would select the records yourself, so the figure and the list behind it are the same set rather than two nearly-equal ones - which is what makes it safe to assign the list straight out of the report.
What it can and cannot see
Section titled “What it can and cannot see”Every figure is counted from work that already happened - agent run records, and the activities and opportunities that sync in from your CRM. Nothing is estimated or modelled.
Two consequences follow. First, the numbers are computed once a day rather than live, so the report speaks about complete days, and the most recent days can still move as late CRM activity arrives; a fortnight-old figure is settled, yesterday’s is provisional. Second, delivery is only as visible as your CRM: the worked stages read CRM-recorded emails, calls, and meetings, so a team that works accounts outside the CRM will look less active than it is. That is a reason to fix the logging, not to discount the report, because the same blind spot is hiding the work from your forecast too.
Pipeline value is reported in the currency each opportunity was created in, and amounts in different currencies are never added together or converted.
Eva-only: the delivery chain and the account and contact lists share one definition of every stage, so a stage number and the equivalent list filter agree by construction. “Reachable” is a property of having a safe primary email or any phone, not of having been enriched; “worked” is anchored to the latest qualification date, so a touch logged before a record was requalified does not count it as worked. Stages are strict subsets, but on a freshly recomputed day the trailing window can briefly break that invariant, which is why untouched counts are clamped at zero rather than shown as negative. The activity-volume and channel-mix checks are anchored differently from the delivery chain on purpose: they count contacts never touched at all, and phones never called at all, with no date bound, whereas the chain and the coverage bar are anchored to the latest qualification. That is why the two figures can differ on the same screen, and it is what makes each check’s number equal to the list its click-through opens. When a customer says the report undercounts their activity, the first question is whether their reps log in the CRM at all - LinkedIn-only outreach and calls made outside the CRM are the usual explanation, ahead of any bug.