Buyer's guidePrivate Capital
Choosing portfolio monitoring software: criteria, scoring and a pilot on your own files
Portfolio monitoring software collects financial and operating data from portfolio companies, standardises it and turns it into views for deal teams, operating partners, finance and LPs. Choosing one is less about dashboards than about data: how KPIs are defined, how management accounts are ingested and how restatements are handled. This guide sets out evaluation criteria, a scoring template, a pilot run on your own files and the case for building instead.
On this page
- What monitoring has to answer, and for whom
- Capabilities to evaluate and how to test each one
- Data model questions to put to every vendor
- Testing AI features on your own management accounts
- Buy a platform, build on a warehouse, or extend existing tools
- Which route fits your firm
- Integration and security questions to settle before contract
- A six-to-eight-week pilot on your own files
- Questions and answers
- Sources
What monitoring has to answer, and for whom
Four groups read monitoring output, and they ask different questions. Deal teams want to know whether each company is tracking its investment case. Operating partners want early signs of where to intervene, such as falling conversion or rising working capital. Finance needs reliable inputs for quarterly valuations and LP reports. LPs want consistent company-level information in the fund's reporting.
Before any demo, write down the ten questions monitoring most needs to answer, who asks each one, how often and how precisely. A weekly cash view for a stressed company and a quarterly valuation input are different requirements, and many platforms do one well and the other adequately.
For the wider market, ColdAI's Private Equity Software Report 2026 maps vendors by category, including portfolio monitoring, document automation and data ingestion, valuation, ESG monitoring and LP portals1. This guide does not rank vendors; it gives you criteria and a test method to apply to a shortlist.
Capabilities to evaluate and how to test each one
| Capability | What good looks like | How to test it | Warning sign |
|---|---|---|---|
| Collection portal | Company finance teams upload packs or fill templates, with reminders and sign-off | Have two portfolio company CFOs submit a real pack | Submissions only work by email to a vendor analyst |
| Document ingestion | Management accounts in PDF or spreadsheet are parsed into fields with source links | Run your own packs, including scans and spreadsheets with merged cells | Accuracy is quoted, but only on the vendor's sample files |
| KPI standardisation | Company-specific metrics map to firm-wide definitions, with the mapping visible | Map one company's custom KPIs into your dictionary during the demo | Definitions are fixed by the vendor and cannot be versioned |
| Validation | Rules and anomaly flags catch breaks, outliers and missing periods before publishing | Seed a pack with a known error and see if it is flagged | Data publishes with no review step |
| Valuation support | Valuation inputs, comparables and prior marks sit beside the KPIs that drive them | Rebuild last quarter's valuation workpaper for one company | Valuation is a free-text field unlinked to data |
| ESG data | ESG metrics collected through the same workflow with definitions and evidence | Collect one ESG metric your LPs ask for, with its evidence | ESG sits in a separate module with separate logins |
| Reporting outputs | Exports to your LP reports, board packs and data warehouse without rekeying | Produce one page of your current quarterly report from the platform | Exports are images or locked PDFs |
Capability categories follow those used in ColdAI's Private Equity Software Report 20261. The tests are suggestions for a pilot, not a ranking of any vendor.
Data model questions to put to every vendor
Most failed monitoring projects fail here, months after go-live, when the first add-on acquisition or restatement arrives.
Testing AI features on your own management accounts
Demonstrations of AI extraction, anomaly detection and commentary use clean files, so build your own test set: recent packs from several portfolio companies, mixing clean PDFs, scans and spreadsheets with merged cells. Record the expected values for a fixed list of fields before the vendor sees the files.
Score extraction field by field, and check that every value links to the page or cell it came from; a correct number without a source still has to be checked by hand. For anomaly detection, plant known errors such as a sign flip or a missing month. For commentary, check that every sentence traces to a figure in the platform.
Where models run, whether your data is retained or used for training, and whether AI features can be switched off per company belong in the security gate, not the feature score.
Buy a platform, build on a warehouse, or extend existing tools
| Factor | Buy a monitoring platform | Build on a warehouse and BI layer | Extend administrator or CRM tools |
|---|---|---|---|
| Best fit | Many companies with varied KPIs and an operating team needing portfolio views | Firms with data engineers wanting monitoring joined to deal and market data | Small portfolios and lean finance teams |
| Time to first useful view | Weeks to months, mostly KPI mapping and onboarding | Months, because ingestion and validation must be built | Short, but limited to what the tools capture |
| KPI flexibility | Within the vendor's data model | Whatever you design and maintain | Usually low |
| Ongoing ownership | Vendor maintains the product; you own definitions and onboarding | You own everything, including on-call support | Shared with the administrator or CRM vendor |
| Main risk | Lock-in if history cannot be exported | Key-person dependency on the engineers who built it | Outgrowing it without noticing |
Hybrids are common: some firms buy ingestion and collection but keep the standardised data in their own warehouse.
Which route fits your firm
- If
The portfolio is small, the finance team is lean and nobody can maintain a data pipeline.
ThenExtend your administrator or CRM tools and invest first in a KPI dictionary and standard submission templates.
Agreed definitions carry over to any platform you buy later.
- If
Operating partners need comparable KPIs across many companies with different systems.
ThenBuy a platform, and budget as much effort for KPI mapping and company onboarding as for the licence.
Onboarding effort, not software, decides whether the data is complete.
- If
You already run a data warehouse with engineers and want monitoring joined to deal and market data.
ThenBuild on the warehouse, buying document ingestion as a component if extraction is the hard part.
A second store of the same company data creates reconciliation work forever.
- If
LP reporting standards are driving the project.
ThenWeight outputs that feed your LP portal and the ILPA templates most heavily.
Fee and expense data comes from fund accounting, so monitoring must hand over cleanly to it.
Integration and security questions to settle before contract
A six-to-eight-week pilot on your own files
Choose companies and questions
Pick a handful of portfolio companies that represent your hardest formats, and the ten questions from your requirements.
Agree a KPI dictionary subset
Define the metrics the pilot will cover, with units and owners, so every vendor is tested against the same definitions.
Load history and a live cycle
Load recent history, then run one real reporting cycle with portfolio company finance teams submitting through the platform.
Score against the template
Rate each criterion high, medium or low using evidence from the pilot, with weights agreed before demos began. Treat security and data-use terms as pass or fail.
Decide and plan rollout
Choose, negotiate export and exit terms, and plan onboarding in waves, starting with companies whose finance teams are ready.
Questions and answers
How long does onboarding portfolio companies take?
It depends more on the companies than on the software. Agreeing KPI mappings with each finance team, loading history and getting a first clean submission take the most time, and messy charts of accounts take longest. Onboard in waves, starting with willing finance teams, and treat the first full reporting cycle as part of onboarding.
Who should own KPI definitions?
The fund, not the vendor or the portfolio companies. A named person in fund finance or portfolio operations should own the dictionary and approve changes, while each company's CFO owns the accuracy of its submissions. Keep definitions versioned, so that a change to how a metric is calculated does not quietly break comparisons across periods or companies.
Can portfolio monitoring software feed the ILPA templates?
Partly. Portfolio-level data such as cost, valuations and proceeds can feed the portfolio sections of the ILPA Performance Template, and company KPIs often appear in LP letters. Fee, expense and capital account figures come from fund accounting and the administrator. Plan the handover between the two systems explicitly, so both sides agree on identifiers and valuation dates.
Should we wait for vendors' AI features to mature?
No. AI features change quickly, while the data model and portfolio company onboarding are what you live with for years. Choose on the data model, integrations and export rights, then test AI features on your own packs to see what they do today. A platform with weaker AI but a sound data model is the safer choice than the reverse.
Do portfolio companies need to change their own systems?
Usually not. Most platforms accept uploaded management packs or templates, and some connect to common accounting systems if a company agrees. What companies need is a consistent reporting routine and a person who submits and answers queries. Asking them to change accounting systems for the fund's convenience is rarely necessary.