RESEARCH · VENUE INTELLIGENCE · AUGUST 2026

Mosaic ATS: A Deep Evaluation

A new non-displayed midpoint venue that sorts every order into one of three interaction levels using a proprietary informed-trading score, and requires institutional subscribers to declare their expected participation rate on every single order.

Sapinover Research · August 9, 2026 · Sources: Mosaic ATS, LLC Form ATS-N/UA filed March 26, 2026 (CIK 0002065275, File No. 013-00210) and FINRA Rule 4552 ATS Transparency Data, 12 reporting weeks

01Operator Background

Mosaic ATS, LLC is a Delaware entity, SEC File No. 013-00210, operating under CIK 0002065275. It has exactly one Form ATS-N on EDGAR, an updating amendment filed March 26, 2026. That is unusually thin for a venue of this structural complexity, and it means the entire public record of how Mosaic works sits in a single document.

The venue does not run on its own infrastructure. Three outside relationships carry it:

• Mosaic Platforms supplies shared employees and the ATS trading technology under an expense-sharing agreement.
• Instinet provides clearance and settlement.
• Ocean builds the Constructed NBBO that every execution is priced against.

That third relationship deserves attention. Mosaic does not price off the SIP by default. It prices off a Constructed NBBO that Ocean assembles from fully redundant direct feeds for all exchanges, combined with SIP data, with the venue continuously arbitrating between the two for latency and feed anomalies. If the Constructed NBBO fails, pricing falls back to the SIP alone. For a venue whose entire execution model is midpoint, the quality of that constructed reference price is the product.

02Structural Architecture

Mosaic operates a non-displayed central liquidity book providing anonymous matching in all eligible NMS securities. Orders arrive over FIX as either Firm or Conditional. Every order is non-displayed and pegged to the midpoint of the Constructed NBBO. There is no displayed quote, no auction, and no price other than the midpoint.

Midpoint-pegged orders may carry a limit price or not. Where a limit is supplied, the venue uses the more conservative of the limit and the current midpoint when testing eligibility. Market orders execute at the midpoint. That is the entire pricing surface.

The last two conditions are the interesting ones. Mosaic will not cross unless displayed liquidity exists on both sides of the public market. A midpoint venue is only as honest as the quote it references, and requiring two-sided displayed interest before matching is a direct defence against pricing off a degenerate or one-sided book. It also means Mosaic is structurally incapable of trading in the thinnest moments, which is a deliberate trade of opportunity for integrity.

03MERIT and the Three Levels

Here is what makes Mosaic genuinely different. Every order entering the venue is classified as one of two things. An Investor order is a subscriber broker-dealer acting in an agency or riskless principal capacity for an institutional or retail investor. A Risk Provider order is a subscriber acting for its own account in a principal or market making capacity.

Classification is not inferred from behaviour. Subscribers attest to it in their subscriber agreements and certify it per workflow during onboarding, and a single subscriber can be classified both ways across different workflows. Mosaic then evaluates orders for trading compatibility through a proprietary scoring system it calls MERIT, the Market Execution Rating for Informed Trading.

The book is then physically partitioned into three interaction segments by those classifications:

Subscribers cannot opt in or out of the assigned segments.

That sentence is the whole design. On most venues, counterparty segmentation is a service you configure. On Mosaic it is the market structure itself, and it is not negotiable. The venue also declines to reclassify subscribers unilaterally. Instead it monitors whether activity is consistent with the declared classification, and where activity looks materially inconsistent, it can require the subscriber either to adjust behaviour or to voluntarily re-certify the workflow under a new classification, as a condition of continued access. The operator retains sole discretion to limit or restrict access to protect the integrity of the framework.

This is an enforcement model built on attestation rather than surveillance-driven reclassification. It puts the compliance burden on the subscriber and keeps the venue out of the business of unilaterally relabelling its own clients, which is a meaningfully different posture from the tiering systems run by the large bank pools, where the operator scores you and moves you.

04User Urgency

The second novel element is a required order parameter. For any order classified as Investor, the subscriber must supply a User Urgency value on an order-by-order basis. It maps to the expected participation rate of the parent order the subscriber is working in the broader market, defined as parent order quantity over expected market volume in that symbol across the expected duration of the order.

A companion optional parameter, Contra Category, lets a subscriber restrict interaction on an order-by-order basis with counterparties carrying a relatively higher User Urgency, as verified by MERIT. So the venue asks the buyside to declare how aggressively it is working an order, independently verifies that declaration through its scoring system, and then lets other participants screen on the verified value.

Framed plainly: Mosaic has taken the thing institutional traders most want to know about a potential counterparty, how much of the day's volume that counterparty still has to execute, and turned it into a disclosed, verified, screenable field. Whether subscribers declare it accurately, and how effectively MERIT catches those who do not, is the question the venue lives or dies on. The filing does not describe MERIT's inputs or methodology, which is unsurprising for proprietary logic but leaves the central mechanism unauditable from the public record.

05FINRA Volume Profile

Mosaic has reported under FINRA Rule 4552 for thirteen weeks, from the week of April 13 through the week of July 13, 2026. Across that span it traded $46.0 million of notional in 6,011 trades across 627 symbols, at an average trade size of roughly $15,500, or about 168 shares.

The shape is unmistakable. April is a pair of single test trades worth tens of dollars. May and June are sporadic, ranging from four trades to a couple of hundred, with isolated bursts on May 25 and June 22 that look like individual large prints rather than sustained flow. Then July changes character entirely. The week of July 6 brought $31.9 million across 2,178 trades in 365 symbols, and the week of July 13 brought $46.8 million across 3,304 trades in 435 symbols. Between them those two weeks are 84% of every dollar the venue has ever crossed.

The second week being larger than the first is the part that matters. A single outsized week reads as a one-off, a block or a test at scale. Two consecutive weeks at that level, with the second up roughly 47% on the first and trade count still climbing, reads as a venue that turned something on and kept it on. A book that had touched 56 distinct symbols across its first eleven weeks touched 435 in its thirteenth. The most reasonable reading is a move from limited pilot to broad subscriber activation, though the filing record does not confirm that and we are inferring it from the tape.

The composition is heavily large-cap. Measured across the eleven weeks for which FINRA has published both tiers, Tier 1 names carry 92.4% of notional and Tier 2 the remaining 7.6%. For a venue whose pitch is protection from informed counterparties, concentrating in the most liquid and most heavily quoted names is coherent: those are the symbols where a Constructed NBBO is most reliable and where a midpoint is least likely to be stale.

EQT is the outlier worth noting: $5.19 million in nine trades, an average of about 11,200 shares per print. Against an overall venue average of 214 shares, that is block activity, and it suggests the Conditional order type is being used for size rather than the book functioning purely as a small-clip midpoint sweeper. HPE shows the opposite profile, $5.35 million spread across 119 trades. Both patterns coexisting in a venue this young is a sign the three-level structure is being exercised in different ways by different subscribers.

06Open Questions

Four things the public record does not settle.

• MERIT is a black box. The scoring system that verifies User Urgency and governs trading compatibility is described by name and purpose but not by input, weighting or methodology. That is normal for proprietary logic, and it also means the venue's central claim cannot be evaluated from outside.
• Classification rests on self-attestation. Mosaic explicitly does not reclassify unilaterally. Its remedy for inconsistent activity is engagement, then a requirement to adjust behaviour or voluntarily re-certify, backed by discretion to restrict access. That is a governance mechanism rather than a technical control, and its effectiveness will not be visible in any public dataset.
• Single filing, thin history. One ATS-N/UA on record means no amendment trail to read. When Mosaic changes something material, that filing will be the first real look at how the venue evolves.
• One week of real volume. Twelve reporting weeks sound substantial until you notice that eleven of them are pilot-scale. Any conclusion about liquidity quality, fill rates or subscriber mix rests on a single week of data.

07Verdict

Mosaic is the most opinionated new book on the U.S. tape. Where most venues compete on latency, fee schedules or order-type surface area, Mosaic competes on a single thesis: that adverse selection is the problem worth solving, and that solving it requires knowing who your counterparty is and how hard they are pushing. Its answer, mandatory classification, a non-optional three-level book, a required urgency declaration and a proprietary verification score, is more structural than anything the incumbent bank pools run, and it is imposed rather than offered.

The trade-off is equally clear. Mosaic will not match without two-sided displayed liquidity, prices exclusively at a midpoint it constructs itself, and confines matching to regular hours. It has deliberately narrowed when and how it will trade in exchange for controlling who trades with whom.

On the tape, the venue is real but early. Ninety-three million dollars over thirteen weeks is a rounding error against a market that clears trillions, and two weeks supply most of it. But those two weeks answered the question the earlier data could not. A single outsized week is a spike. Two in a row, the second larger than the first, is a launch. What remains open is whether a venue can sell enforced counterparty segmentation to a buyside that has historically preferred to configure its own protections, and whether this level of activity holds once the novelty is gone. The next few FINRA prints will answer that faster than any filing will.

##References

Sapinover builds intelligence on U.S. equity market structure and the overnight session. This analysis is for informational purposes only and is not investment advice or a solicitation.