The portfolio that reports itself.

It's the second week of the quarter, and somewhere in your firm there is a file called Fund I KPI Consolidation v7.xlsx.

Thirty portfolio companies. Thirty templates that started identical and drifted. Three call the same number "Revenue", "Net Revenue", and "Sales". Eleven haven't replied to the request email. Two replied with a PDF.

Someone on your ops team turns that into one board-ready view. Every quarter. By hand.

The uncomfortable truth: most funds don't have a portfolio monitoring problem. They have a collection problem — and dashboards built on chased, unreconciled data just make the chase prettier.

Here is how CapHive ends the chase.

Collection is a workflow, not a mailbox

The quarter starts with one click: notify every portfolio company that hasn't reported. Each company gets a secure upload link — valid 15 days, expiry date printed in the email, no portal login, no password reset ticket. Each company can name designated KPI contacts, so the reminder reaches the person who actually owns the numbers.

From that moment, every company has a status badge: Link Sent → Draft Saved → Submitted → Reviewed. Your ops team stops asking "who's missing?" — the request tracker answers at a glance.

Companies save drafts as they work; nobody loses progress to a browser tab. When they hit Submit, the fund team is emailed instantly and the report locks. And the reminder engine has one rule your inbox never learned: it stops. Submitted companies are skipped in every future reminder cycle — automatically, every time.

Even the workbook itself is checked at the door: if the uploaded file's date headers are missing or malformed, the Submit button stays disabled and the company sees exactly what to fix — before bad structure ever reaches your team.

Nothing touches a dashboard without approval

Here is the design decision that matters most: collection and consumption never share a pipe.

Every uploaded number lands in a staged state — visible to your team, highlighted for review, but invisible to every dashboard, performance table, and growth calculation. An authorized approver reviews the staged values (changes are flagged; identical re-uploads are skipped entirely) and promotes them in one action.

Only on approval does the platform recompute what depends on the data: quarter, YTD, and annual totals roll up from monthly actuals — or from quarterly actuals when that's all a company reports, never both, so a year is never double-counted. Growth percentages recalculate. An audit note records who approved what, and when.

One company on one screen

Open any portfolio company and its dashboard is assembled from widgets you control: the KPI grid (choose categories, periods, and Actual/Budget/Projected tags), the cap table with ownership charts, document embeds, data cards.

The grid doubles as a charting surface. Tick the KPIs you care about, mix chart types — Revenue as bars, margin as a line — and the chart follows every grid filter live. Save the layout ("Revenue vs Margin, 12M") and reopen it next quarter in one click, or load several saved views side by side to compare eras of the same company.

Because charts use the standard investor-facing KPI names from your mapping layer, the legend says "Revenue" even when the company's workbook said "Net Sales". Consistency is enforced by the system, not by the analyst's memory.

The whole portfolio on one screen

The same widget system builds the aggregated portfolio dashboard: a multi-company investment table grouped however you slice it, portfolio IRR, fair-market-value distribution, sector and company breakdowns.

Two capabilities make this scale operationally:

Named dashboards. Every dashboard type supports multiple saved configurations — an operational Default, a condensed Board View, a Budget View — switchable from a dropdown, bookmarkable by URL, with per-user defaults so the managing partner lands on the board view while ops lands on the working view. One data foundation, many audiences, zero rebuilt decks.

Agent Charts. Describe a chart once, in plain English — "line chart of Revenue and EBITDA, last 12 months" — and the AI generates it for every portfolio company that matches a tag. Thirty companies, one definition, thirty consistent charts, each embedded on its company's dashboard and regenerable the moment new KPIs are approved. Chart standardization stops being a manual discipline.

⚙️ Under the hood

The reason this holds up at scale is a set of deliberately boring guarantees:

• A KPI mapping layer translates each company's raw metric names to your standard names — matching is case- and space-insensitive, so "gross profit" and "Gross Profit" are the same metric, and every dashboard legend speaks one language.
• A date parser that meets companies where they are: monthly, quarterly, half-year, annual, YTD, and day formats — Jan 2024, Q1 FY24, H1 2023, YTD Nov 23 — 20+ recognized formats across 6 period families, detected anywhere in the first 10 rows of the sheet.
• Idempotent re-uploads: if a company resubmits the same workbook, unchanged values are skipped — no duplicate approvals, no phantom "changes" to review.
• Rollups that refuse to double-count: annual totals compute from monthly actuals, or from quarterly actuals when monthly data doesn't exist — never both for the same year.
• Scoped access: upload links grant access to exactly one company's report for one period. Fund-side, every dashboard and agent operates inside the viewer's tenant and role permissions — enforced by the application layer, not by AI goodwill.

Each of these is invisible on a good day. Together they are why the numbers on the board slide survive an audit question.

The agents that keep it honest

Dashboards show you what the data says. Four support agents make sure the data deserves to be believed — each answering one question:

1. Did it arrive? The Portfolio Company Agent — fully rule-based, no AI involved — sweeps the portfolio: KPIs submitted for the period? Required documents attached? Valuation recorded? Missing items are flagged as blocking, and reminder emails go out only on the days you configure (say, day 7, 14, and 21), never more than once a day.
2. Is it healthy? The Data Quality Agent runs nightly integrity checks — stale valuations past your threshold, stale FX rates, commitment overflows — and emails a health score with a fix-it list.
3. What does it say? The Portfolio Scoring Agent reads each company's approved KPIs, documents, and valuations against a sector-specific rubric you define, and produces a structured 0–100 score with written justification — decision support, reviewable and editable, never auto-actioned.
4. What is management not telling you? The Investor Presentation Analyst reads board decks across periods, compares them, surfaces quietly dropped risks and missed commitments, and hands you 8–12 grounded pre-board questions plus a management accountability score.

Every run is logged with full execution history. Deterministic agents need no AI key at all; AI agents are pinned to the model and provider you choose, per entity.

📊 The impact

	The email-and-spreadsheet quarter	The monitored pipeline
Who's missing?	Reply-all archaeology	Status badge per company
Reminders	Manually, apologetically	Automatic, and they stop on submit
"Revenue" vs "Net Sales"	Analyst fixes it in the workbook	Mapping layer fixes it once
Board pack charts	Rebuilt by hand each quarter	Saved layouts + tag-based generation
Data trust	"It should be right"	Staged → approved, with an audit note
Between quarters	Silence	Nightly health score + agent sweeps

What used to be a quarterly chase across 30 companies — with a consolidation workbook only one person understood — becomes a pipeline: statuses instead of follow-ups, approvals instead of hope, dashboards that update the moment data is promoted.

What to take from this

A genuine question for portfolio ops teams: what share of your quarter is collecting the numbers versus actually reading them?

Fund Ops, Unstacked #3 — one CapHive feature, properly explained.