This article describes a FinancialCrime.org investigation conducted in collaboration with a trading-systems specialist. The brokerage described in this report has been given the pseudonym “Meridian Markets” and the affiliated entity has been called “MMX Liquidity Services” to protect the integrity of ongoing regulatory proceedings and the privacy of individuals involved. Specific penalty and restitution figures have been adjusted. The core facts, methodology, and outcome are accurately reported. A detailed technical submission was filed with the relevant derivatives regulator, which subsequently conducted its own controlled testing and obtained a settlement.
How This Investigation Started
Jonas spent years as a trading-systems tester, mostly reviewing execution quality for small asset managers — the kind of unglamorous, technical work that involves comparing fill prices against independent market data feeds across thousands of trades and looking for patterns that deviate from what a fair execution model should produce. After leaving a bank, he built a paid newsletter for professional traders.
His investigation began with a complaint from a reader.
The reader traded contracts-for-difference (CFDs) and rolling spot FX through Meridian Markets, a fast-growing online brokerage that had blanketed the market with advertising. Meridian’s marketing was everywhere — podcasts, sports sponsorships, airport lounges, social media influencers. Its website promised “institutional-grade execution for ordinary investors,” “no dealing desk intervention,” “best available pricing,” and “execution aligned with the client.” The branding was polished, the platform was slick, and the growth was rapid. Meridian had gone from a niche platform to one of the more visible retail brokerages in its market within three years.
The reader’s complaint was specific. Losing trades seemed to execute instantly. Winning trades seemed to hang.
He was not describing an occasional bad fill. He was describing a pattern: when the market moved against him between the moment he clicked “buy” and the moment the trade was confirmed, the trade went through at the worse price without hesitation. When the market moved in his favour during the same window, the platform would frequently requote — offering a new, less favourable price — or simply delay execution until the favourable movement had dissipated.
Jonas was sceptical. Retail traders frequently blame brokers for bad timing. Confirmation bias is powerful: you remember the trades that went wrong and forget the ones that went right. Execution complaints are the background noise of retail trading.
But this reader was different. His complaint included timestamps accurate to the millisecond, screen recordings of the requote behaviour, and exported order logs from the platform showing submission times, confirmation times, and the price at each stage. He had been collecting evidence for months before reaching out.
Jonas looked at the logs. He saw enough anomalies to ask for my help. We decided to run a controlled test.
The Controlled Experiment
Anecdotal evidence — even well-documented anecdotal evidence from one trader — is not proof of systematic manipulation. The reader’s logs showed a pattern in his trades. We needed to determine whether that pattern was specific to his account, specific to Meridian, or simply a feature of volatile market conditions that would appear at any broker.
To answer that question, we designed a controlled experiment that would isolate Meridian’s execution behaviour from normal market dynamics.
Methodology
Over six weeks, we opened small accounts at Meridian and four competitor platforms — brokers of comparable size, offering similar products (CFDs and rolling spot FX), and targeting similar retail client profiles. All five accounts were funded with identical amounts.
We built a controlled testing environment with three components.
Synchronised clocks. Each trading terminal was time-synchronised to a common NTP (Network Time Protocol) server, ensuring that timestamps across all five platforms were accurate to within a few milliseconds of each other. This was essential for comparing execution speeds: if we submitted an order at the same moment on two platforms, we needed to know that “the same moment” actually meant the same moment.
Independent price feeds. We subscribed to two independent market data feeds — one from a major interdealer broker and one from a data aggregator — that provided real-time prices for the same instruments we were trading. These feeds served as our ground truth: the actual market price at the moment of order submission and at the moment of execution. By comparing the broker’s fill price against the independent price, we could measure slippage — the difference between the price we expected and the price we received — and determine its direction.
Scripted order entry. To eliminate human variability in order timing, we built scripted tools that submitted identical orders across all five platforms simultaneously, triggered by the same market event. The scripts handled the mechanics of clicking, submitting, and logging. We controlled the timing. The instruments were the same, the order sizes were the same, and the entry points were the same across all five platforms.
Trade selection
We did not place trades randomly. We specifically targeted scheduled volatility events — moments when the market was likely to move sharply in a short period, creating the conditions under which execution asymmetry would be most visible. These events included central bank interest rate announcements, consumer price index and inflation data releases, crude oil and natural gas inventory reports, and major equity market opens.
The logic was straightforward. During normal market conditions, the price difference between order submission and execution is typically small — a few tenths of a pip — and the direction of the movement is essentially random. Slippage during calm markets is noise. But during volatility events, the price can move materially in the milliseconds between submission and execution. If a broker’s execution engine treats positive slippage (price moves in the client’s favour) and negative slippage (price moves against the client) symmetrically, both should pass through to the client with roughly equal frequency and magnitude. If the engine treats them asymmetrically — capturing positive slippage while passing through negative slippage — the asymmetry will be most visible during volatile conditions, when the magnitude of slippage is largest.
Over six weeks, we placed several hundred trades across the five platforms, concentrated around these scheduled events. Each trade was small — we were not trying to move the market or trigger unusual counterparty behaviour. We were trying to measure how the platform handled the gap between submission and execution.
What we found
The results from four of the five platforms were unremarkable. Slippage was present — it always is during volatile conditions — but it was roughly symmetrical. Positive slippage and negative slippage occurred with comparable frequency and magnitude. The fill quality varied somewhat between platforms, but the variation was within normal range and showed no consistent directional bias.
Meridian was different.
The pattern was subtle but, across hundreds of trades, statistically unambiguous. When market movement between order submission and execution favoured the client — that is, when the market price at the moment of execution was better than the price at the moment of submission — Meridian’s system responded in one of three ways. It requoted, offering a new price that was less favourable than the current market price. It rejected the order outright, returning a “price not available” message. Or it delayed execution — adding 200 to 800 milliseconds of latency beyond the platform’s normal processing time — during which the favourable movement often dissipated.
When market movement favoured Meridian — that is, when the market price at the moment of execution was worse for the client than the price at the moment of submission — the trade executed at the worse price without hesitation. No requote. No rejection. No delay. The negative slippage passed through to the client with remarkable consistency.
We called it the one-way latency valve.
Positive slippage almost never reached the client. Negative slippage always reached the client. The execution engine was not random, not symmetrical, and not aligned with the client. It was a filter — a mechanism that sorted market movements by who they benefited and treated them differently depending on the answer.
Quantifying the asymmetry
Our statistical appendix documented the asymmetry in several ways.
Slippage distribution. Across all Meridian trades during volatility events, negative slippage (client receives a worse price than submitted) occurred in approximately 73% of trades that experienced any slippage. Positive slippage (client receives a better price than submitted) occurred in approximately 8%. The remaining trades executed at the submitted price. At the four competitor platforms, the positive/negative split ranged from 42/58 to 48/52 — close to the symmetrical distribution that a fair execution model would produce.
Requote and rejection rates. During periods when the market moved in the client’s favour between submission and execution, Meridian requoted or rejected approximately 61% of orders. During periods when the market moved against the client, Meridian requoted or rejected approximately 4% of orders. At competitor platforms, requote rates showed no significant correlation with slippage direction.
Latency asymmetry. The average time between order submission and execution at Meridian was 127 milliseconds when the market moved against the client (negative slippage) and 489 milliseconds when the market moved in the client’s favour (positive slippage). The difference — 362 milliseconds — was consistent across instruments and event types. At competitor platforms, execution latency showed no significant correlation with slippage direction.
Aggregate dollar impact. Across our test trades at Meridian, the net slippage — the cumulative difference between the price we should have received (based on independent market data at the moment of submission) and the price we actually received — was negative. The one-way valve extracted value from every volatility event, trade by trade, millisecond by millisecond.
The amounts on individual trades were small — fractions of a pip. On any single trade, the difference was indistinguishable from normal market friction. It was only in aggregate, across hundreds of trades, with the statistical rigour to measure directionality, that the pattern became visible. This is precisely what makes the practice effective and difficult to detect. No individual client will notice a few tenths of a pip on a single trade. A broker processing millions of trades per day accumulates those fractions into a substantial revenue stream.
The Second Finding: The Affiliated Counterparty
While we were analysing execution data, we pursued a parallel line of inquiry: who was on the other side of Meridian’s trades?
Meridian’s marketing described its execution model as connecting clients to a “professional liquidity network” — language that implied access to a broad pool of external market makers, banks, and institutional liquidity providers. The impression created was that Meridian was an agency broker, passing client orders through to the market and earning only the disclosed spread or commission.
We checked Meridian’s regulatory filings and corporate registry records. A related entity — which I will call MMX Liquidity Services — was incorporated as a separate company but shared overlapping shareholders and directors with Meridian Markets. MMX was registered as a liquidity provider and held the appropriate regulatory authorisations to act as a counterparty to CFD and FX transactions.
Meridian’s Form disclosure (the regulatory equivalent of the ADV in the advisory context) mentioned that the firm “may use affiliated liquidity providers in certain circumstances.” The disclosure did not identify the affiliate by name, did not state that the affiliate was the dominant counterparty for certain products, and did not explain the economic implications of trading against an affiliate controlled by the same shareholders.
When we mapped the execution data against publicly available information about Meridian’s order flow, the picture became clearer. For certain high-volume products — particularly major currency pairs and popular index CFDs — a large share of client orders were not being routed to external liquidity providers at all. They were being internalised by MMX. Meridian’s clients believed they were trading against the market. In practice, for many of their trades, they were trading against Meridian’s own affiliate.
This is not illegal per se. Many retail brokers internalise client flow — it is a common business model in CFD and FX markets, and regulators permit it subject to disclosure and best-execution obligations. The issue was the gap between Meridian’s marketing (“no dealing desk intervention,” “best available pricing,” “execution aligned with the client”) and the reality (affiliated counterparty on the other side, execution engine configured to favour the house).
The one-way latency valve now had a beneficiary. When the execution engine captured positive slippage — when the market moved in the client’s favour but the client did not receive the benefit — the economic value did not evaporate. It accrued to the counterparty on the other side of the trade. And the counterparty, for a substantial portion of these trades, was MMX — Meridian’s own affiliate.
The combination was potent: an execution engine engineered to capture favourable price movements, feeding internalised order flow to an affiliated counterparty that profited from the capture. The one-way valve was not just a latency game. It was a revenue extraction mechanism with the house on the other side.
Client segmentation
Then came the detail that made the case personal.
Several Meridian clients contacted Jonas after his initial newsletter piece (described below). One of them was a former contractor who had worked on Meridian’s platform infrastructure. The contractor sent us a screenshot of an internal dashboard that showed client accounts organised into profitability tiers. The tiers were labeled “toxic,” “neutral,” and “retention-sensitive.”
“Toxic” clients were profitable traders — clients whose trading generated losses for the counterparty (MMX). “Retention-sensitive” clients were high-volume but unprofitable traders — clients who generated significant revenue through spreads and were at risk of switching to a competitor. “Neutral” clients fell in between.
The contractor described — but we could not independently verify — that execution tolerances could be adjusted by tier. “Toxic” clients experienced tighter execution windows (more requotes, more rejections, longer delays). “Retention-sensitive” clients received better execution (faster fills, lower requote rates) to keep them trading. The execution engine was not just asymmetric between positive and negative slippage. It was asymmetric between clients, calibrated to the profitability each client represented to the house.
We did not publish the screenshot. We could not authenticate it — a screenshot of a dashboard, provided by a former contractor, is evidence of what the contractor saw, not necessarily evidence of what the system actually did. Screenshots can be fabricated. Memories can be imprecise. We treated it as intelligence, not evidence.
But we used it to sharpen the questions in our regulatory submission. Rather than simply alleging asymmetric execution, we could now ask the regulator to look for specific things: client-segmentation tables, execution-tolerance configurations by client group, and change logs showing when and why execution parameters were modified.
The Newsletter Piece
Before filing with the regulator, Jonas published a cautious newsletter piece. He did not name Meridian. He described the methodology — controlled testing across multiple platforms, synchronised clocks, independent price feeds, scripted order entry — and the general finding: that at least one retail platform appeared to be operating asymmetric execution filters that systematically captured positive slippage while passing negative slippage to clients.
He warned readers that the practice might be more widespread than the industry acknowledged, and suggested specific steps traders could take to test their own execution quality: compare fill prices against an independent data feed, track slippage direction over a large sample of trades, and pay attention to requote rates during volatile versus calm conditions.
The piece was deliberately understated. Jonas did not allege fraud. He did not name the broker. He described a methodology and a finding, and let readers draw their own conclusions.
The response was significant. Several Meridian clients — who recognised their broker from the behavioural description — contacted us with their own order logs. Their data was consistent with our findings: the same asymmetry, the same requote patterns, the same latency difference between favourable and unfavourable price movements. The additional data strengthened our statistical sample and confirmed that the pattern was not limited to our test accounts.
The former contractor contacted us after the newsletter piece, providing the dashboard screenshot and the description of client segmentation described above.
Building the Regulatory Submission
Our submission to the derivatives regulator was the most technically detailed package I have been involved in preparing. The audience — a regulator’s market-conduct unit — was sophisticated enough to evaluate raw data, statistical analysis, and execution-engine mechanics. We gave them everything.
What the package contained
Test-account credentials and order logs. We provided full access to our test accounts at Meridian, including every order submitted, every execution received, and every requote or rejection recorded. The regulator could log in and verify the data independently.
Independent price-feed comparisons. For every trade in our sample, we provided the independent market price at the moment of order submission and at the moment of execution, sourced from two commercial data feeds. This allowed the regulator to independently calculate slippage for each trade and verify our directional analysis.
Timestamp normalisation code. Because comparing timestamps across platforms requires accounting for clock drift, network latency, and platform-specific time recording conventions, we provided the code we used to normalise timestamps across all five platforms and the independent data feeds. The regulator could run the code, inspect it, and verify that our timestamp comparisons were accurate.
Screen recordings. Video recordings of representative trades showing the requote behaviour in real time — the order submitted, the market moving favourably, the requote appearing, and the re-submitted order filling at a less favourable price.
Statistical appendix. A formal statistical analysis of slippage distribution, requote rates, rejection rates, and execution latency, with tests of significance demonstrating that the asymmetry at Meridian was not attributable to random variation at any reasonable confidence level.
Corporate structure analysis. Registry documents showing the ownership and directorship overlap between Meridian Markets and MMX Liquidity Services, and the disclosure analysis showing the gap between Meridian’s marketing claims and its regulatory filings regarding affiliated liquidity.
Proposed examination request. Rather than simply describing what we had found, we proposed specific documents and data the regulator should request from Meridian: the execution engine’s configuration files and parameters, client-segmentation tables (if they existed), affiliate-liquidity contracts between Meridian and MMX, revenue-sharing or transfer-pricing arrangements between the two entities, change logs showing modifications to execution parameters (particularly around volatility events or quarterly performance periods), and internal communications related to slippage handling, requote policy, and client categorisation.
This last element — the proposed examination request — served the same function as the “Likely Defences” section in the Asterion investigation. It gave the regulator a roadmap for what to look for and where to look. Regulators are not short on authority. They are often short on specificity — knowing which questions to ask and which documents to request. We provided both.
What the Regulator Found
The regulator did something we had hoped for but could not have demanded: rather than simply requesting an explanation from Meridian, it ran its own controlled trades through supervised accounts and compared execution against independent market data. This independent replication of our methodology was crucial, because it meant the regulator’s findings were based on its own evidence, not ours.
The regulator’s controlled testing confirmed the asymmetry. Positive slippage was systematically captured; negative slippage was systematically passed through. The requote rates correlated with slippage direction. The latency asymmetry was real and consistent.
The regulator also subpoenaed internal communications — Slack messages, email, and configuration records — that provided the intent evidence our external analysis could not reach.
The internal documents showed that Meridian’s execution settings had been modified after a bad quarter — a period in which the firm’s counterparty book (via MMX) had experienced losses from client trading activity. The modifications tightened execution tolerances for trades that would have resulted in positive slippage to clients — effectively widening the one-way valve.
Two internal messages, quoted in the regulator’s findings, did particular damage.
One product manager wrote that “giving back positive slippage is a charity model.” The message was sent in the context of a discussion about whether to allow favourable price improvements to pass through to clients, as some competitors did. The product manager’s position — which prevailed — was that retaining positive slippage was a legitimate business decision, not a client harm.
Another message described the affiliated liquidity provider — MMX — as “the house edge without the casino language.” The author was describing, with uncomfortable candour, the economic function that the affiliate served: it provided Meridian with the economics of being a counterparty to client trades, without the regulatory and marketing complications of describing itself as a dealing-desk broker.
The client-segmentation system was also confirmed. The regulator found internal dashboards that categorised clients by profitability and documentation showing that execution parameters could be, and were, adjusted by segment. “Toxic” clients — those whose trading generated consistent losses for the counterparty — received tighter execution windows, more frequent requotes, and longer processing times. The system was designed to maximise revenue extraction from each client category.
The Outcome
Meridian settled without admitting intentional fraud, but the order was severe and operationally comprehensive.
The firm paid $11.2 million in restitution to affected clients, calculated based on the aggregate value of asymmetric slippage captured over the examined period. It paid a $9.5 million regulatory penalty. The total financial consequence — $20.7 million — was substantial for a firm of Meridian’s size, though likely a fraction of the revenue the one-way valve had generated over the years it operated.
The operational requirements were more consequential than the financial penalties.
Execution symmetry. Meridian was required to eliminate asymmetric execution filters. Positive and negative slippage were to be treated identically: if negative slippage passed through to clients, positive slippage must also pass through. The execution engine’s handling of price improvements and price deterioration must be provably symmetrical, subject to ongoing monitoring.
Affiliate disclosure. Meridian was required to disclose its relationship with MMX Liquidity Services in plain language — not buried in regulatory filings, but in the platform interface, at the point of trade execution. Clients must be told, clearly, when their counterparty is an affiliate of the broker, and what that means for the economic alignment of the trade.
Independent execution-quality monitor. An independent monitor was appointed for two years to review Meridian’s execution quality on an ongoing basis, with access to execution logs, configuration files, and internal communications.
Quarterly execution-quality statistics. Meridian was required to publish quarterly reports including: positive and negative slippage distribution across all client trades, requote and rejection rates by market condition, average execution latency, and the percentage of trades internalised by affiliated entities. These publications would allow clients — and competitors — to evaluate Meridian’s execution quality against objective benchmarks.
Separation of segmentation from execution. Meridian was required to separate its client-segmentation analytics from its execution-engine settings. Client profitability data could still be used for marketing, retention, and risk management purposes. But it could not be used as an input to execution parameters. How a trade is filled must not depend on how profitable the client is.
Personnel and governance changes. The head of trading resigned. The board created a conduct-risk committee with authority over product design, marketing claims, and execution controls. The compliance function was given veto power over any changes to order-handling logic — a structural change that ensured the execution engine could not be modified without compliance review and approval.
We testified privately before the regulator’s market-conduct unit, providing technical context for the execution data and answering questions about our methodology. Our testimony was not public. The settlement terms were published in the regulator’s enforcement report with the findings described in general terms.
What This Investigation Teaches
Asymmetric execution is not a bug — it is a business model
The one-way latency valve was not a software error or a configuration mistake. It was a deliberate design choice that treated the gap between order submission and execution as a revenue opportunity. The execution engine was programmed to ask, for every trade, a simple question: does the price movement between submission and execution benefit the client or the house? If the house, execute. If the client, delay, requote, or reject.
This is not unique to Meridian. Asymmetric execution practices — sometimes called “last look” in institutional FX markets — are a known feature of certain market structures. What made Meridian’s practice problematic was not the existence of execution discretion (which its terms permitted) but the combination of discretion with marketing that promised the opposite. “No dealing desk intervention” and “execution aligned with the client” are specific claims. An execution engine that systematically favours the house contradicts both.
For retail traders, the practical implication is clear: do not trust marketing claims about execution quality. Test them. Track your slippage over a large sample of trades. Compare fill prices against an independent data feed. Pay attention to requote rates, particularly during volatile conditions. If you find a pattern — and you need hundreds of trades to distinguish pattern from noise — you have evidence, not just a complaint.
The affiliated counterparty is the structural issue
The one-way valve was a mechanism. MMX — the affiliated counterparty — was the economic engine. Without internalisation, the value captured by the execution filter would have accrued to external liquidity providers, not to Meridian. It was the combination of asymmetric execution and affiliated internalisation that created the revenue extraction mechanism.
This structure — a retail broker that internalises flow through an affiliate and profits from the other side of client trades — is legal and common in CFD and FX markets. But it creates an inherent conflict of interest that most retail clients do not understand. When your broker is also your counterparty, every dollar you make is a dollar they lose, and vice versa. The broker’s economic interest is directly opposed to yours.
Best-execution obligations are supposed to manage this conflict. The broker is required to obtain the best available price for the client, regardless of its own position. But best-execution enforcement is difficult when the broker controls the execution engine, the liquidity pool, and the pricing — and when the client has no independent mechanism to verify that the fill they received was actually the best available.
This is an area where the weaponised compliance dynamic I have written about elsewhere is particularly visible. The rules require best execution. The broker’s systems generate data that purports to demonstrate best execution. The client and the regulator must take the broker’s data on trust — unless someone conducts the kind of independent, controlled testing that Jonas and I performed. Few people do.
Client segmentation crosses a line
The tiered execution system — where “toxic” (profitable) clients receive worse execution than “retention-sensitive” (unprofitable) clients — is the finding that most disturbed me, and I suspect will most disturb regulators in other jurisdictions as they examine similar practices.
The economic logic is straightforward. A profitable client is profitable because their trades consistently move in the right direction — which means the counterparty (the affiliate) consistently loses money. From the broker’s perspective, these clients are a cost centre. Degrading their execution quality — more requotes, longer delays, tighter tolerances — reduces the counterparty’s losses.
An unprofitable client is the opposite: a revenue source. Their trades consistently move in the wrong direction, generating counterparty profits. These clients should be retained and satisfied, because they represent reliable income. Giving them better execution keeps them trading.
The result is a system that punishes competence and rewards incompetence. The better you are at trading, the worse your execution becomes. The worse you are, the better the platform treats you. This is the exact inverse of what “execution aligned with the client” means — and it is the logical endpoint of a business model that profits when its clients lose.
How to detect this yourself
For traders who suspect their broker may be running asymmetric execution, here is the approach we used, simplified for individual application.
Track every trade you make for at least three months, recording: the price at the moment you clicked submit, the price at the moment of execution, the time between submission and execution, and whether the trade was requoted, rejected, or filled. Use an independent price feed — even a free one — to record the market price at the time of each order.
Calculate slippage for each trade: execution price minus submission price, adjusted for direction (positive if the price improved in your favour, negative if it worsened). Plot the distribution. If positive and negative slippage are roughly symmetrical — similar frequency, similar magnitude — your broker’s execution is fair. If negative slippage is significantly more frequent or larger than positive slippage, the distribution is asymmetric and warrants further investigation.
Compare requote rates during volatile and calm conditions, and — critically — compare them by slippage direction. If requotes concentrate specifically during conditions where the market is moving in your favour, the pattern is not random.
If you find a statistically significant asymmetry and want to pursue it, document everything and get in touch. These cases are difficult but not impossible to bring to regulators — if the evidence is properly prepared.
The broader pattern
The Meridian investigation, like the Asterion investigation, illustrates a recurring theme in the work I do through this site: the most harmful financial misconduct is often not dramatic fraud but the systematic, technically sophisticated extraction of value from clients through mechanisms that are individually invisible and collectively significant.
A few tenths of a pip per trade. A few basis points on share class selection. A few percentage points on a cash sweep rate. Each one is too small for any individual client to notice. In aggregate, across thousands of clients and millions of trades, they are the difference between a business that serves its clients and a business that serves itself while marketing the opposite.
The common thread is the gap between what the firm says — “fiduciary-first,” “no dealing desk,” “execution aligned with the client” — and what the firm does. Closing that gap requires evidence, expertise, and the willingness to do the forensic work. It also requires regulators who are prepared to act when the evidence is presented — and who have the technical capability to verify it independently.
In Meridian’s case, the regulator’s decision to run its own controlled trades — rather than relying solely on our data or Meridian’s self-reported execution statistics — was the decisive step. It produced independent evidence that could not be challenged as biased, self-interested, or methodologically flawed. It is a model that other regulators should consider for execution-quality examinations.
If you have information about execution practices at retail brokerages — asymmetric slippage, client segmentation, affiliated internalisation, or other practices that contradict best-execution obligations — I would like to hear from you. Reach out at [email protected].