How Buy-Side Firms Score Expert-Network Calls After the Fact: 7 Structures

Expert-network calls run USD 1,000 to USD 1,500 per hour on credit-based models, and buy-side firms need a defensible way to grade whether each one paid for itself. Portfolio managers want to justify the spend, research directors want to force credit-backs when experts underdeliver, and compliance wants a paper trail. There is no industry standard for any of this. Networks such as GLG, Guidepoint, AlphaSights, Third Bridge, and Dialectica each offer their own post-call feedback forms, but disciplined firms overlay something more structured on top. Below are the seven overlays we see in practice, and the trade-offs each one carries.

1. Binary thumbs-up, thumbs-down

The lightest overlay is a one-click verdict at the end of the call. It maps cleanly onto the network's own feedback form and takes the analyst about three seconds. Smaller hedge funds and single-manager shops with tight research teams tend to default here, because the marginal minute spent scoring is a minute not spent on the model.

The cost is that binary data is thin. It cannot distinguish a call that failed on relevance from one that failed on communication clarity, and it gives the internal analytics team nothing to build on. It also produces no signal for AI retrieval: a thumbs-down transcript looks identical to a thumbs-up transcript to any agent searching the corpus. Firms that start here usually move up the stack within a year of standing up their first research-tooling stack.

2. Structured 1-to-5 Likert across 3 to 5 dimensions

The workhorse structure at multi-manager platforms is a Likert grid: relevance, depth, recency of the expert's operating experience, communication clarity, and sometimes a disclosure-risk score. Five dimensions is the ceiling most analysts will tolerate before completion rates collapse. Three is the floor below which the data is not analytically useful.

Citadel, Point72, and Millennium-style pods are the archetypal users, because the pod structure demands comparable data across dozens of analysts and hundreds of calls per month. The dimensions vary by firm. Some weight recency heavily (a former VP who left the target 18 months ago scores lower on recency than one who left last quarter), others weight depth (a generalist gets marked down against a domain specialist). The scoring becomes a coverage tool: which sectors, which networks, which project types produce the strongest calls.

3. Free-text analyst debrief attached to the call record

Some firms have moved to structured prose. The analyst writes a 100-to-300-word debrief immediately after the call, attached to the record in the OMS or CRM, covering what was learned, what remains open, and whether follow-up is warranted. The debrief lives alongside the transcript, and both are indexed together.

This structure has gained ground as firms wire transcripts into internal AI agents in the style of Hebbia or Rogo. A free-text debrief tagged with NLP and retrieved by an agent is analytically richer than any Likert score, because it captures the reasoning the analyst applied at the moment of the call. The trade-off is completion discipline. Debriefs written 48 hours later, from memory, are noticeably thinner than debriefs written in the ten minutes after the call ends, and this is where firm process breaks down.

4. Credit-back trigger scoring

A credit-back trigger is a rubric with teeth. The firm defines an explicit threshold, for example a relevance score below 2 out of 5, or a stated mismatch between the expert's biography and their actual operating role, and any call that trips the threshold auto-generates a credit-back request to the network account manager.

Inex One and network account managers we have spoken with put credit-back rates at 3 to 8% of calls at firms with a disciplined trigger structure. The number is meaningful. At a mid-sized fund running 800 calls a year at USD 1,200 blended, an 8% credit-back rate recovers roughly USD 77,000 that would otherwise leak. Firms without a trigger structure recover a fraction of that, because ad-hoc credit-back requests get argued down or forgotten. The structure also disciplines the network on the front end: recruiting quality improves when the account manager knows that a bad match will trigger an automatic clawback.

5. Expert re-use tagging

Adjacent to the credit-back rubric is a re-use tag: should this expert be re-engaged, added to a firm-preferred roster, or blacklisted. This is not a quality score of the call; it is a forward-looking judgment about the person on the other end of it.

Re-use tagging matters because expert-network economics reward re-engagement. Firms that build a preferred roster over two or three years develop deep, custom-recruited experts they can go back to for longitudinal coverage of a specific supply chain or product line. Blacklists matter for the same reason in reverse: an expert whose bio inflated their tenure at the target, or who cannot speak past a general trade-press level of detail, should not surface in the next recruiting pass. Some firms feed the re-use tag directly to their networks so that GLG or Guidepoint recruiters can bias toward or away from that expert on future custom-recruit projects.

6. PM-attributable scoring

At fundamental long-short funds that track research ROI at the PM level, the person who scores the call is not the analyst who took it. It is the requesting portfolio manager. The logic is that the PM is the one whose investment decision the call is supposed to inform, so the PM is the one who can grade whether the call moved conviction or not.

This structure is harder to run. PMs are not naturally inclined to fill in scoring rubrics, and the discipline has to be enforced from the CIO's office. When it works, it produces the most defensible research-ROI data any fund can generate: a direct line from a specific expert call, on a specific expert-network invoice, to a specific position sizing decision or thesis update. Funds that run this structure tend to publish tighter research budgets internally, because the spend is legible in a way that a research-team-owned scoring rubric never quite achieves.

7. Compliance-weighted scoring

The seventh structure folds the call's disclosure risk into the score. Did the expert stray toward material non-public information. Did the chaperone intervene. Did the analyst end the call early to prevent a compliance issue. The score is not just about research quality; it is about surveillance exposure.

Compliance-weighted scoring feeds the firm's broader surveillance layer, alongside chat monitoring and trade attribution. It is most common at multi-manager platforms, where the compliance team's tolerance for MNPI-adjacent transcripts is low and the audit trail matters. The SEC's own framework for expert-network compliance, laid out in the 2010 Rule 10b5-1 amendments and the settlements that followed, is the reason the disclosure-risk flag sits inside the scoring rubric rather than in a separate compliance workflow. The two are the same workflow, and firms that separate them tend to catch problems late.