Activist 13D + Insider Buying: 3 Real Clusters in 17,592 Filings (2018–2026)

A full scan of the 13dwatch database — 17,592 Schedule 13D filings from 2018 through 2026, joined to every Form 4 insider transaction on the same issuer — surfaces 3 events where an activist 13D is followed within 90 days by a non-self-filed open-market insider purchase. The "smart-money cluster" pattern is real. It is also rare enough to be a product.

What was measured

The 13dwatch pipeline holds 17,592 activist filings and 1,079 Form 4 insider transactions as of 2026-05-06, queryable at /api/health. For this study, an admin-gated analytical endpoint joined every filing to the insider-transaction table on issuer_cik and counted, for each filing, the insider rows occurring within ±90 days of the filing date.

A "true cluster" means a Form 4 row with transaction code P (open-market or private discretionary purchase, per SEC Form 4 Table II) where the insider's filer name does not match the activist's filer name. The same-name match is removed because a 5%+ activist is automatically a Section 16(a) reporter and shows up as their own "insider" — that is reporting overlap, not signal. The previous post in this series, Form 4 Insider Buying While Activists File, set up the methodology in detail.

The headline number

Source: /api/admin/insider-overlap-study?window=90 on the 13dwatch worker, snapshot 2026-05-06.

The conditional rate (0.30%) is the meaningful figure for a hedge fund analyst, because it controls for the recall ceiling: insider Form 4 ingest only covers 423 of the 4,264 unique issuers in the activist filings table. Among issuers where the join can match, ~1 in 333 activist filings is followed by a discretionary management open-market buy. Among filings where any insider rows appear in the window, ~1 in 11 is the real pattern. Everything else is reporting overlap, vesting, derivative settlement, or sale.

The 3 candidates

Sources: /api/filing/0000898432-26-000286 (CBK); /api/filing/0000919574-26-000467 and /api/filing/0000919574-26-000468 (SVRN).

The OceanPal pair is one underlying event: two activists filed simultaneously on 2026-01-26 (a wolf-pack 13D — see When Multiple Activists Circle the Same Stock). The Co-CEO bought ~80 days later. So the distinct-cluster count across 8.3 years is 2.

The Robertson × Commercial Bancgroup case is the clean showcase: an outside investor announces an 8.5% stake on Apr 27, the issuer's CFO buys 3,300 shares on the open market 3 days later. Three-day lag is short by 90-day-window standards. The CFO's purchase was at $29.02/share for $95,760 — meaningful for a community bank holding company executive but not a structural-position purchase.

All 3 candidates are on small or micro-cap issuers. There are no large-cap names on the list. The pattern concentrates where management ownership matters most — which is also where outside activist pressure has the most marginal effect on personal incentives.

Where the noise comes from

Of 33 filings showing any insider Form 4 inside the 90-day window, the SEC transaction-code distribution is:

Code P is 5.8% of all transaction-code occurrences inside the matched set. Anything labeled "insider buying alongside an activist" by a less rigorous source is, with high probability, counting one of the other 94%. That is the entire methodological gap post #5 opened. This study closes it on the full database.

The reporting-overlap effect

Of the 33 filings with any insider Form 4 in the window, 7 (21%) had the activist itself as the only insider. A 5%+ activist is automatically a Section 16(a) reporter under 15 U.S.C. § 78p(a), so any time the activist trades the issuer's stock, the row appears in both the activist filings table and the insider transactions table. Counting those as "insider buys alongside the activist" is a pure double-count.

The remaining 26 (79%) did have at least one independent insider in the window. Most of those independent insiders were selling, vesting, or settling derivatives. Three were buying.

Why this matters

The data study answers a specific empirical question: does the cluster pattern that hedge fund commentary often invokes ("when an activist files and management buys, that's the real signal") actually fire? Yes — at a base rate of about 1 event per 5,860 filings, or about 1 per 333 filings if the issuer's insiders are even tracked. When it fires, it is unmistakable. When it doesn't, no amount of hand-waving aggregation can make it appear.

For an emerging hedge fund manager or registered investment adviser running a sub-$500M book, the practical takeaway is that monitoring 13D filings without a code-P filter on the joined Form 4 stream is an exercise in producing false positives. A serviceable workflow is: pull every new 13D, look up insider Form 4 activity on the same issuer in the 90 days before and after, filter to transaction_code = "P", exclude rows where the Form 4 filer matches the 13D filer. What survives is short. What survives also tends to be predictive — though that is a question for the next study with longer post-filing return data.

What 13dwatch does with this

The cluster pattern is exactly what the 13dwatch live feed is built to surface. The other ten thousand activist filings — the ones with no insider buy — are not the product. The product is the 1-in-5,860. As of 2026-05-06, the live count of standing real clusters is small, and the signal is fresh: the Robertson × Commercial Bancgroup setup happened nine days ago. Hedge funds and RIAs interested in real-time alerts on this specific cross-source pattern can request pilot access — pricing is $1,500/month.

Frequently asked questions

How is a "cluster" defined in this study?

A cluster is a Schedule 13D filing followed by at least one Form 4 insider transaction, on the same issuer (issuer_cik match), within 90 days, where the transaction code is P (open-market discretionary purchase) and where the Form 4 reporter's name does not match the 13D filer's name. The same-name filter removes the trivial case of a 5%+ activist filing both a 13D and a Form 4 on themselves.

Why is the rate so low?

Three reasons. First, most activist 13D targets do not also have actively trading insiders on file. Second, of insiders who do trade, most are vesting equity grants, settling derivatives, paying tax-withholding obligations, or selling — not buying on the open market. Third, the 13dwatch insider-transactions table covers a 4.6-year window (2021-10-01 to 2026-05-04) versus the activist filings table's 8.3-year window, which caps recall on the older filings.

Are the candidates investable?

This study is research, not investment advice. The candidate list is the output of a pattern-detection rule, not a recommendation. A real investment thesis on any of the named issuers would require fundamental analysis, valuation, and an understanding of the activist's stated intent (Item 4 of the 13D filing).

How does this relate to the academic literature?

Brav, Jiang, Partnoy and Thomas (2008) document approximately +7% abnormal returns around hedge-fund activist 13D announcements with no reversal in the following year. Cohen, Malloy and Pomorski (2012) document significant abnormal returns from "opportunistic" insider trades. The intersection — both effects firing on the same issuer — is the cluster pattern. The academic work establishes that each leg has signal in isolation. This study quantifies how often they actually occur together: rarely, but unambiguously when they do.

How can I see clusters in real-time?

The public /api/feed endpoint returns the 60 most recent activist filings with the 90-day insider join already computed. Filter results client-side for insider_activity.buys > 0 (this counts code-P rows in the worker's aggregation). Cross-check that the matched insider's filer name differs from the 13D filer's name. The full backfill across the 17,592-filing database is exposed to pilot subscribers via the admin endpoint.

Why are some activist filings labeled "SC" or "SCHEDULE" instead of "SCHEDULE 13D"?

The U.S. Securities and Exchange Commission (SEC) renamed the EDGAR form labels in 2025. Pre-2025 filings carry abbreviated labels (SC, SCHEDULE) and post-2025 filings carry the full SCHEDULE 13D / SCHEDULE 13D/A labels. Both are activist 13D filings and were treated identically in this study.

Methodology and limitations: Source endpoint GET /api/admin/insider-overlap-study?window=90 on the thirteen-d-watch Cloudflare Worker (deployment_id 6cbb3c6ddd8e4bcb8948f144d592ff49, deployed 2026-05-06). Universe: 17,592 rows from activist_filings (Supabase project acouoyxkzrqtcbimveyl), 1,079 rows from insider_transactions. Window: ±90 days around filing_date. Issuer match: issuer_cik after leading-zero normalization. Self-filer exclusion: case-insensitive equality on filer names after stripping non-alphanumeric characters. The 184-issuer joinable subset reflects current Form 4 ingest coverage; the conditional 0.30% rate uses 1,010 filings on those issuers as the denominator. The 0.017% raw rate uses the full 17,592-filing universe. Both are reported because the choice of denominator depends on whether the reader treats absent insider data as "no event" or "missing data." Code "X" and "H" rows in the 33-filing overlap pool are not standard SEC Form 4 Table II codes and have been flagged as likely ingest artifacts. Research, not investment advice. Long Street Consulting LLC.