Home / DPA Insights / Portfolio Intelligence
The Quarterly Pack Arrives. The Decision Already Happened Somewhere Else.
Why periodic monitoring is a structural disadvantage - and what the alternative actually looks like.
The problem is not that the data is wrong. The problem is that by the time it arrives, the moment for the decision has already passed.
How to read this page
01 / The lag
Why does portfolio data reach the investment committee too late to act on?
Multi-manager private markets portfolios have a monitoring problem that is structural, not operational. GP reports arrive on different schedules, in different formats, with varying levels of look-through detail. By the time data reaches the investment committee, it is 45 to 75 days stale. The quarterly pack is assembled manually. Already old by the time the IC sits down to discuss it.
In that environment, portfolio monitoring is a periodic event, not a continuous capability. The questions that matter most - aggregate sector exposure across managers, unexpected geographic concentration, early signs of credit stress in a portfolio company - only surface if someone thinks to ask them during a scheduled review cycle. The architecture does not make those questions easy to ask between cycles.
02 / Intake
How do you standardise GP reports that arrive in different formats?
We ran two experiments. The first addressed intake. GP reports arrive in heterogeneous formats - inconsistent line items, non-standard taxonomies, varying detail. We built an AI-augmented ETL pipeline combining OCR extraction, ML-based data labelling, and automated outlier detection. Every report standardised at intake against a controlled taxonomy; anomalies flagged before data enters the monitoring environment. The first version broke on edge cases and required a redesign that separated the rules layer from the AI standardisation layer. The resulting architecture reduced manual QC intervention by 68% while maintaining 99.5% accuracy across 14,000+ companies and 38,000+ records.
03 / Decision environment
What does continuous portfolio monitoring actually look like?
The second experiment addressed what you do with clean data once you have it. We built a live decision-intelligence environment consolidating 10 portfolios and 200+ companies into a single real-time view - multi-factor performance attribution, sector and geography exposure drill-down, anomaly alerting. Design intent: the IC arrives with data already assembled, validated, and visualised. Analysis starts at the meeting, not before it.
The two experiments connect directly. The first tells you what the data needs to look like. The second tells you what to do with it once it does.
04 / LP expectations
What do the 2025 ILPA reporting changes mean for oversight teams?
ILPA’s January 2025 GP-LP Reporting Framework update reflects where LP expectations are heading: more granularity, faster timeliness, less tolerance for data that cannot be interrogated on demand. Oversight teams that rely on periodic packs will find it increasingly difficult to answer questions their LPs are already asking. The investment is not in new monitoring software. It is in the data normalisation and intelligence layer that makes any monitoring environment trustworthy at the moment a decision needs to be made - not three reporting cycles after it.
Frequently asked questions
Why is a quarterly reporting pack too late to act on?
In multi-manager private markets portfolios, GP reports arrive on different schedules, in different formats, with varying levels of look-through detail. By the time the data reaches the investment committee it is 45 to 75 days stale, and the pack itself is assembled manually. The data is not wrong. It simply arrives after the moment the decision needed to be made.
Why is aggregate exposure across managers so hard to see?
Because monitoring is structured as a periodic event rather than a continuous capability. Aggregate sector exposure across managers, unexpected geographic concentration and early signs of credit stress in a portfolio company only surface if someone thinks to ask during a scheduled review cycle. The architecture does not make those questions easy to ask between cycles.
How do you standardise GP reports that arrive in different formats?
Standardisation has to happen at intake, not at reporting. We built an AI-augmented ETL pipeline combining OCR extraction, machine learning based data labelling and automated outlier detection, with every report standardised against a controlled taxonomy and anomalies flagged before the data enters the monitoring environment. The design lesson was that the rules layer has to be separated from the AI standardisation layer, otherwise the pipeline breaks on edge cases.
Is the answer to buy new portfolio monitoring software?
No. The investment is not in new monitoring software. It is in the data normalisation and intelligence layer that makes any monitoring environment trustworthy at the moment a decision needs to be made, rather than three reporting cycles after it. Clean, standardised intake determines what the environment built on top of it can be trusted to answer.
What do the 2025 ILPA reporting changes mean for oversight teams?
They reflect where LP expectations are heading: more granularity, faster timeliness, and less tolerance for data that cannot be interrogated on demand. Oversight teams that rely on periodic packs will find it increasingly difficult to answer questions their LPs are already asking.
