Hong Kong SFC Colleague Network - Professional Relationship Explorer

In finance, a job, a hire, or a warm reply usually comes through someone you've worked with. Search anyone SFC-licensed to map their direct colleagues: everyone who held a licence at one of their firms at the same time, and how those colleagues connect, so you can find an introduction, see who to reach out to, and follow how careers move.

Hong Kong SFC Firm Talent Flow - Hiring and Attrition Explorer

Firms hire from, and lose people to, a fairly predictable set of rivals, and those talent corridors are a map of opportunity. Search a firm to see where its SFC-licensed staff came from and where they went: the rivals it hires from, and the ones it loses people to.

Hong Kong SFC Talent-flow Network

Search across the register to see how licensed people move between firms. Node size ranks firms by PageRank or transfer volume, colour compares business categories or detected communities, and every edge is a measured transfer corridor.

Search for an individual above to explore their professional network

Search for a firm above to see who it hires from and loses people to

Loading the published talent-flow network...

Talent centrality ranking

Quality-weighted PageRank and raw transfer-volume ranks are shown together for the selected time window. The table lists every mapped (backbone) firm active in the window; # is the rank within this list, while the PageRank and volume ranks count every SFC firm with at least one transfer, so those can skip numbers.

#FirmPageRank rankVolume rankPageRankPeople
Loading ranking...

How this network is built

Who appears — the searched person's direct SFC colleagues: individuals who held an SFC licence at one or more of the same firms during overlapping periods. Lines between colleagues mean they also once shared a firm (even one the searched person was never at).

How colleagues are ranked — each one gets a tie‑strength score, summed across every firm they share with the searched person. For each shared firm:

  • Days of overlapping tenure at that firm — longer shared time means a stronger tie.
  • Firm size penalty — sharing a boutique means more than sharing a megacorp, so the contribution is divided by log(firm_size + 2).
  • Licence activity match — two people who both do e.g. Dealing and Advising at the same firm are more likely to have actually worked together. We add the Jaccard similarity of their activity sets.

Formula: score = Σfirm overlap_days × (0.5 + activity_jaccard) ÷ log(firm_size + 2)

Only the top 200 colleagues by score are drawn so the graph stays legible. Hover any node to see its score and rank; click to re-center the network on that person. “Active only” restricts to currently-licensed relationships.

Looking at firms instead of people? See the firm talent flow →

How these flows are computed

Source: the SFC public register, licence records only.

  • Stints. A person’s licence records at a firm are merged into continuous stints; gaps of 90 days or less are bridged, so renewals and role switches are not counted as moves.
  • Moves. A move is counted when a stint at one firm ends and the person’s earliest stint at a different firm begins within 365 days. Each departure produces at most one move, to the nearest destination. A licence already held at the destination before the departure is never counted as a move. Moves are dated by the start at the destination.
  • Periods. 3y/5y/10y means 1,095/1,825/3,650 days before today; the exact cut-off date is shown on each button.
  • Before April 2003. The register discloses nothing before 1 April 2003. Records with no start date are treated as starting on that date, so all-time figures are a floor, not a count, and some corridors show a spike at April 2003.
  • Recent months. A departure in the last 365 days cannot be classified yet: the person may still move within the window. Such departures appear as “too recent to classify”, and figures for the most recent 12 months are provisional.
  • Affiliated entities. Corridors marked ‡ link firms whose registered names share leading words (for example two Morgan Stanley entities). These often reflect internal transfers or group licence re-papering rather than open-market hires. The flag is a name heuristic: it can miss renamed or differently-branded affiliates, and the register publishes no group structure. A few firms appear twice under near-identical names; transfers between duplicate records are flagged the same way.
  • What this register cannot see. Bank staff are registered with the HKMA, not the SFC: of 186 registered institutions in this data, almost none have individual records here. Moves to or from banks, overseas roles, and unlicensed positions are invisible. “Left the register” therefore does not mean the person left finance.
  • New to the register counts people whose first SFC licence record ever is at this firm. They are not necessarily new to finance.
  • Returns and repeats. A return to the same firm within 365 days is continuity, not a move. A person who leaves a firm several times is counted once per departure.
  • Coverage. Up to 12 firms per direction and period are drawn; the remainder is aggregated as “Other firms” with its firm count shown.

Looking at people instead of firms? See the colleague network →

How the talent-flow network is built

Source and edition. This is the 2026-H2 edition, computed from SFC public-register licence records. The edition freezes the method; daily data refreshes do not silently change the formula.

  • Moves. Licence records at one firm merge into a stint when gaps are 90 days or less. A transfer is the earliest different-firm stint starting 0–365 days after a stint ends, at most one destination per departure. Transfers are dated by the destination start.
  • Affiliated registrations. Firms connected by at least two people concurrently licensed for at least 30 continuous days are grouped first. Components of ten firms or fewer merge automatically; larger components require an explicit reviewed whitelist. Transfers inside a merged group become self-loops and are removed. The ‡ mark identifies a merged node.
  • Edges and windows. Edge weight is the number of transfers, never overlap days. All history, ten-year and five-year metrics ship together; changing the control does not fetch or rearrange the network.
  • PageRank. Weighted directed PageRank uses alpha = 0.85 on every corridor with at least one transfer. Forward PageRank is the talent-magnet measure; reversing every edge gives the neutral talent-output measure. Node size can instead show raw people joining or leaving.
  • Drawing threshold. Metrics use the full graph, but a firm pair is drawn only when at least one direction has five or more all-time transfers. Many active firms are isolated at this threshold; search still finds them and reports the absence honestly.
  • Ranking table coverage. Ranks are computed for every firm with at least one transfer — 4,061 firms all-time — while the table lists the 822 firms above the drawing threshold. A firm missing from the table usually still has movers, just never five or more with any single counterpart; its complete hires and departures are on its firm talent-flow page. Firms with no recorded transfers at all — small or short-lived firms, HKMA-registered banks whose staff are not SFC-licensed, and firms whose leavers exited the register — are not ranked.
  • Corridor thinning. To keep the picture readable, the map draws each firm's three strongest qualifying corridors per time window, plus the minimal links that keep the displayed network in one piece. Every qualifying corridor still counts in the metrics and stays visible in the ranking table, tooltips, the focused view and firm flow pages.
  • Left censoring. The register discloses no earlier date for records marked “since at least 1-Apr-2003”. All-history counts are therefore floors, not complete career histories.
  • Right censoring. A departure within the last 365 days may still acquire a destination inside the transfer window. Recent outflow conclusions remain provisional.
  • Coverage limit. Staff registered with the HKMA rather than the SFC, overseas roles and unlicensed positions are invisible. A move out of this network is not evidence that someone left finance.
  • Communities. Fixed-seed Louvain runs on the family-collapsed, all-time backbone. Its agreement with accepted AI business categories is NMI pending; category colour remains the default because communities also capture ownership and market structure.
  • Stability. The edition bootstrap retained 90% of the top ten with rank correlation 0.85. The ranking prints PageRank beside raw-volume rank so disagreements stay visible.