Client impact · Market-rate PE
Exit models stop pulling their own numbers
One post-investment team covered RMB and USD funds, and the same portfolio company was recomputed three times a year — once for valuation, once for KPIs, once for an exit case. We changed three things: a standard collection pack, valuation inputs held as versioned records, and one project spine running from deal pool to portfolio to exit.
The institution
Two currencies, dozens of deals, one post-investment team
The money is market-raised RMB and USD fund capital. Three lines run through the manager: investment, post-investment and IR, and finance. The structure is manager — fund — deal, with funds in the tens and live portfolio companies in the dozens.
The front end was ahead of most peers: origination, diligence files and investment committee records were already managed. The chain breaks after closing. Financial statements and operating metrics are chased company by company, formats differ, and revenue recognition or a restructuring of the holding chain silently destroys comparability.
The binding constraints come from outside: LPs expect periodic reporting, valuations have to survive audit and the delegated investment committee, and every input in an exit case must trace back to a collection event or a parameter version.
Before the change: one company, three calculations
The table describes how effort was spent and carries no percentages. Effort saved in post-investment and valuation can only be estimated once we have looked at your actual collection and recompute log from the last quarter.
| Step | Who carried it | How the effort was spent | Why it never closed |
|---|---|---|---|
| Post-investment collection | Post-investment staff, company by company | Non-standard financial statements and KPIs returned repeatedly; history re-keyed after recognition or restructuring changes | No agreed collection pack, so every reporting season re-aligned definitions from scratch |
| Valuation recompute | Head of post-investment and finance, in separate models | Comparables, discount rates and liquidity discounts held in personal models; recomputed at quarter end and hard to reproduce | Assumptions outside the ledger mean nobody second can rerun the same method |
| KPI alerting | Risk, reading reports after the fact | Thresholds with no baseline and no owner; warnings closed only by a manual quarter-end sweep | Alerts lived outside the record, so neither false positives nor misses could be counted |
| Exit modelling and disclosure | IR and finance | Exit cases re-pulled data from scratch; DPI and IRR disagreed with fund accounting; LP pack and internal monthly pack had two sources | The origination pool and the portfolio book never shared one project identity |
The judgement
Post-investment bottlenecks on evidence, not on analysis
The team is short of defensible inputs, not short of analytical methods. So the delivery weight sits on collection evidence and parameter versions: every number answers "which collection, which parameter version, reviewed by whom".
Valuation judgement stays with the investment committee and the post-investment lead. The system makes that judgement reproducible; it does not make it.
- Every figure carries its source and evidence
- Valuation assumptions versioned, never personal
- An alert closes only with an owner and a conclusion
- Exit modelling reuses the portfolio ledger
Three steps we took
Standardise the pack, version the inputs, connect the spine
The order matters: step one does not touch the investment committee, step two does not change valuation policy, step three only removes duplicated data pulls.
01 · Collect
A standard post-investment pack
Each company gets one defined metric set and submission template. Statements go through OCR with the original file retained, and machine extraction runs alongside manual review for one cycle. Data vintage and source are stored as fields.
- Deliverable: metric set and collection template
- Deliverable: OCR extraction with review log
- Reviewed by: post-investment lead and finance
02 · Value
Assumptions become versioned records
Method, comparables, discount rate and haircut parameters enter the ledger with an effective version, so a recompute can be reproduced. Risk alerting runs against the 18-indicator definition set, with hits routed into a disposition queue.
- Deliverable: valuation parameter ledger
- Deliverable: alert queue with dispositions
- Reviewed by: risk and the delegated IC authority
03 · One source
A project spine that reaches exit
Deal pool, document repository, portfolio book and exit model share one project identifier. Exit cases read from the portfolio ledger, and the LP pack plus the internal monthly pack generate from the same underlying record.
- Deliverable: exit-model to portfolio mapping
- Deliverable: LP report template map
- Reviewed by: IR and the finance lead
Modules and security base in this scenario
Module names follow PEPM’s published naming. Scope follows the assessment findings; nothing here is a standard-function guarantee.
Acceptance measures
All four measures, wording untouched
This engagement uses all four measures published on the client impact index. In the consistency formula the tiers are read here as diligence basis — portfolio ledger — exit model. For the filing cycle, "template issued" is the arrival of the LP data request or report template and "consolidated submission" is delivery of the finalised pack.
How we recomputed them here
| Measure | Calculation (as published) | Recomputation in this engagement | Sample and timing |
|---|---|---|---|
| Filing cycle | Calendar days from template issued to consolidated group submission | Timestamps from request arrival to final submission, never a mock report; two consecutive cycles in the same format compared | Both the RMB and USD report sets, including the post-audit annual version |
| Manual fallback hours | Overtime hours during filing season ÷ people involved, from system-captured effort | Counted only for chasing data, re-returns and quarter-end manual recomputes, taken from logged effort rather than recollection | Reported per reporting season, kept separate from implementation hours |
| Definition consistency | Share of identical metrics where the three tiers agree with zero variance | A fixed metric set — revenue, net profit, ownership, cost, latest valuation — compared across the three stages per company; a variance closes only when a collection record or a parameter version explains it | Sample must include thinly reported private companies, not only the well-documented ones |
| Alert precision on review | Share of triggered warnings confirmed as real risk by the business | Each alert stores a disposition owner and a conclusion; confirmed and unconfirmed shares are computed at quarter end, and misses are listed separately so "no false positives" is never read as "no misses" | Reported per alert type across two full quarters |
A valuation is defensible when a second person can reach it again using the same method.Catalpa delivery methodology
500+
institutions served on the platform (company disclosure, not this manager)
14,000+
funds covered (platform book basis, not one client)
5.6RMB tn+
assets covered (RMB, platform book basis)
60%+
share of engineering staff (company basis)
Where this approach does not apply
An individual angel manager with single-digit live deals and no recurring external disclosure is better served by shared documents plus a valuation assumption sheet; modelling the collection process costs more than it returns. It is also the wrong tool for teams that only raise or broker and hold no lasting valuation duty — every benefit here rests on the same company being computed more than once a year.
Questions specific to post-investment delivery
Mechanism answers. They are not valuation, audit or compliance advice.
Past quarters were valued on a different basis. Do we restate them?
Portfolio companies will not report, or only publish an annual account. What now?
How tight do permissions on valuation data need to be?
How are USD funds and multi-currency handling dealt with?
How does this connect to audit and LP enquiries?
Next step
Bring last quarter’s valuation workpaper
A 45-minute assessment: we check which inputs have evidence behind them, which assumptions exist only in one personal model, and what the shortest route from standard collection to a shared exit source looks like.