Analytics is the add-on with which you assemble reports from the product’s own data yourself, without waiting for a developer: sales, service, marketing — whatever cut you need today. A report is described as data (what to count, in which cuts, for which period), saved once, and then it lives on its own: it can be pinned to a dashboard, hung as a panel on a customer card, or delivered by email on a schedule.
Why your own reports
The built-in reports answer the typical questions of each contour — the sales funnel, deal velocity, ticket deadlines, receivables. But life is wider: “how many service contracts of our partners expire this quarter”, “which managers lose deals after the second touch most often”, “how many requests in each department have been unanswered for over a day”. Calling a developer for every such question is slow and expensive; here the report is assembled from your own wording — or, more precisely, you ask the agent to assemble it in words.
How a report is built
A report is described as data, without code. You name five things: which records to count (contacts, companies, deals, invoices, tickets and others), what to show in the columns, which totals to sum up, which cuts to group by, and for which period. Plus the usual filter — for example, active customers only, or one branch only.
Totals follow the meaning of a field: a count works on anything, sums and averages only on numbers and money, and adding up dates or names makes no sense, so the product will not allow it. The filter inside a report is the same one the product’s lists use: if you set a condition in a report that matches a condition in a list, the numbers match record for record. That matters for trusting the figures: a report and a list must never “differ slightly”.
You can group by ordinary fields and by periods: day, week, month, quarter, year. There are one or two cuts, and that is not an arbitrary restriction: two cuts (say, “month” and “source”) give a readable table, while five turn it into a puzzle in which you yourself easily make a mistake. Conditions on totals (“show only rows where the sum is above a million”) are allowed when there is just one cut.
How to read the numbers
Every number has an origin, and the product names it. In the answer you can see: which period was counted and by which date (creation, closing or the event date), as of which moment it was counted, in which currency and with which rights. The line “with your rights” matters more than it seems: a report never shows more than a person is allowed to see in the lists. That is why the same report gives a manager and a team lead different numbers — and each of them sees that they are looking at their own rights, not someone else’s.
There is also protection against false precision. A cell assembled from too few records warns about it honestly instead of posing as an established fact: a win rate over eleven deals proves nothing. And if the matching rows turned out to be more than fit into the report, the answer says outright “first N of M” — instead of a silent cut that would make the figure look complete.
Dashboards
A dashboard is a screen where several reports stand side by side: revenue, the funnel, service workload, repeat sales. Every tile of a dashboard points to a ready report and takes its numbers from it, computing nothing itself — so the figure on the dashboard and the figure in the report itself are always the same. Filters set on a dashboard are passed into the reports, not applied to an already finished table: the totals are recalculated honestly.
A dashboard has a viewing mode: “with my rights” (each person sees their own picture) or “as a particular owner” (everyone sees that person’s picture). The mode is stated right on the screen, so there is no confusion about whose numbers these are.
A report can also be attached to a customer or deal card — then it is shown right on the card. Such a panel must always name its binding: the panel answers about this particular record, not “about whatever happened to fall into the report”. If a report did not manage to compute within its allotted time, the panel shows the last computed version and states the moment it was made at.
History, exports and delivery
The picture changes over time, so reports can be taken like snapshots: once a day the product remembers the report’s numbers for that day. Snapshots show the dynamics — “how it grew month after month”, not only “how it is now”. Every answer names its basis outright: “live” or “snapshot as of such a date”. You cannot read “how it was in March” from live data — live data is always about today; that is exactly what snapshots exist for. They are kept for a limited time, and the term is a visible number.
A report is exported to a file for a spreadsheet — CSV and XLSX, which Excel opens — and also to a printable version for sending on or taking into a meeting. Exports have limits: how many exports a day and how many rows in each; these are visible numbers you can learn in advance rather than at the moment of refusal. The file lives for 24 hours and is then deleted — data does not multiply across disks uncontrollably. The file’s header names the data moment and the rights it was made under.
On top of that, a report can be delivered by email on a schedule. Every recipient gets their own letter; the rule can decide not to send an empty report, or not to send it again if nothing has changed since last time. Every delivery lands in the journal, and a failure shows up as its own notification — there are no silent failures.
Honest limits
- Reports are cut by rights. Nobody sees more through a report than they see in the lists, and every answer says so itself. That is a rule, not a shortcoming.
- Small numbers warn about themselves instead of pretending to be precise.
- Live data is always about now. The question “how was it in March” is answered only by snapshots, and the answer says honestly that the series starts later if there are no snapshots for that period.
- A removed field does not pass unnoticed. If a field a report was built on has been removed, the product will not silently rebuild the report: it marks the problem and refuses, naming the field. An export behaves the same way — a column will not come out blank without an explanation.
- The contour reports have not moved anywhere. The deal funnel, sales velocity, ticket deadlines, receivables and marketing attribution remain their own sections with their own figures. The shared engine answers everything else, and where a ready report already exists it honestly points at it instead of repeating it.
- Data is handed to an external reporting system as versions, not as a live connection. The product publishes a dataset stating how many rows it holds and as of which moment; when the set of columns changes, the external system receives a divergence warning instead of quietly wrong numbers.