Crypto market making models: in-house vs. outsourced

The hard choice is not “do we want control?” Everyone wants control. The real choice is whether your team can run a trading desk without pretending it is just another growth tool. In-house vs outsourced market making sounds like a vendor comparison. It is not. It is a decision about latency, inventory risk, exchange politics, capital discipline, and who gets blamed when the book looks like a desert at 3 a.m. UTC.

What founders think happens, and what the order book actually does

Founders usually walk into crypto market making with a clean mental picture: deploy bots, quote both sides, earn the spread, keep the token looking liquid. Nice diagram. Almost charming.

The order book does not care.

A real market-making setup has to continuously place and cancel bids and asks across venues while watching price drift, inventory imbalance, volatility, exchange downtime, and the behavior of every sniper bot parked around the book. If your token trades on two centralized exchanges and one DEX pool, you already have three different liquidity realities. The CEX books need quoted depth and spread control. The DEX pool uses an Automated Market Maker model, where liquidity sits in a pool and pricing moves by formula. These are not the same game with different buttons.

On a CEX, the visible book is politics and mechanics at once. Exchanges want tradable markets because dead books embarrass them. Many major centralized exchanges expect a designated market maker before they list a project. They may not say it in theatrical language, but the message is usually blunt enough: show us who will maintain liquidity, or enjoy the waiting room.

On a DEX, there is no traditional order book market maker in the same sense. Liquidity provisioning means capital in pools, often exposed to impermanent loss, fee dynamics, and mercenary LP behavior. You can still run strategies around it. You can still hedge. But you are not “making a book” in the classic bid-ask sense. You are feeding an AMM and living with its math.

Liquidity is not a mood. It is capital, software, execution, and discipline showing up every minute the market is open.

That is why the in-house versus outsourced question matters. If you get it wrong, the damage is not limited to a bad vendor invoice. You can end up with blown inventory, ugly spreads, failed exchange obligations, and a chart that screams “do not touch.”

The technical architecture of in-house market making

Running crypto market making in-house sounds attractive because nobody likes handing token inventory and market behavior to an outside counterparty. I get the appeal. I have sat across from market makers whose pitch deck was cleaner than their execution. I have also seen internal teams convince themselves they could build a trading stack because they had one strong backend engineer and a founder who once traded perps.

That is not a market-making desk. That is a liability with a dashboard.

An in-house setup needs more than a bot that places orders. At minimum, it needs:

1. Exchange connectivity that does not fall apart under load.

You need stable API integrations across every venue you trade. Each exchange has its own rate limits, quirks, maintenance windows, partial fills, and delightful little edge cases that only appear when liquidity matters most.

2. A quoting engine that adapts to volatility.

Static spreads are amateur hour. The bot has to widen, tighten, skew, and pull quotes based on market conditions. If volatility spikes and your engine keeps quoting like it is lunchtime in a dead market, you are donating inventory.

3. Inventory management across venues.

If you sell too much token on Exchange A and accumulate too much quote asset on Exchange B, the dashboard may still look “active,” but your risk is crooked. Rebalancing is not a nice-to-have. It is the plumbing.

4. Latency-aware execution.

Serious HFT infrastructure runs on millisecond and sub-millisecond thinking. Not every token needs ultra-low-latency colocation drama, but if your system reacts slowly, predatory flow will find you. It always does.

5. Risk controls that actually stop trading.

Kill switches, exposure limits, maximum inventory thresholds, stale-price protection, and venue-level shutdown logic. The best risk system is the one that ruins a founder’s day before the market ruins the treasury.

6. Monitoring by people who understand books, not just uptime.

“The bot is online” means nothing if it is quoting nonsense. You need humans who can read spread, depth, fill quality, drift, hedge slippage, and venue behavior.

The uncomfortable part: building or licensing high-frequency trading software capable of managing order books across multiple exchanges simultaneously is not a weekend sprint. The tooling, talent, and maintenance burden add up. The hidden costs are where internal market making starts to look less heroic.

A rough internal stack looks like this:

If you already have a trading engineering team, a risk lead, and the stomach to run treasury inventory through live markets, in-house may be rational. If not, you are not choosing control. You are choosing to learn market structure with your token as tuition.

Outsourced professional market makers: what you are really buying

Professional market makers sell three things: infrastructure, inventory discipline, and exchange credibility. The good ones also sell scar tissue. That matters. A firm that has managed books through exchange outages, liquidity shocks, unlock events, and hostile flow knows things your internal team will otherwise learn by bleeding.

The usual commercial model is not mysterious. Professional market makers often charge a monthly retainer, commonly somewhere from $2,000 to $20,000 or more, depending on the venue set, required depth, token profile, and service scope. That fee may be combined with a performance component, or with a loan of tokens so the market maker has inventory to quote and manage.

The token loan is where founders need to wake up. You are creating counterparty risk. You are handing inventory to a trading firm under contract terms that are often covered by NDA and rarely transparent from the outside. The contract may define return obligations, trading limits, reporting, termination rights, and what happens during extreme market conditions. Or it may be a beautifully vague document that leaves you exposed when the relationship gets ugly.

Outsourcing can be clean. It can also be predatory.

There are firms that understand market integrity: keep spreads reasonable, maintain depth, hedge exposure, report honestly, and avoid turning the book into a carnival. There are also outfits that dress up pump-and-dump behavior as “liquidity support.” They will talk about awareness, momentum, market confidence, and other polite words for “we intend to move the chart until retail shows up.”

Do not confuse the two.

A serious outsourced market maker should be able to discuss:

• target spread ranges by venue and market condition;
• visible and effective depth at defined price bands;
• inventory limits and hedge methodology;
• reporting cadence and metrics;
• treatment of token loans;
• exchange communication responsibilities;
• what they will not do, especially around price manipulation.

That last point matters. Market making does not guarantee price appreciation. It should not be sold as chart management. It is liquidity management. If a vendor promises a higher token price, they are either lying, manipulating, or using the kind of strategy that eventually creates worse problems than a wide spread.

A good market maker makes trading less painful. A bad one makes the chart look exciting right before it becomes evidence.

Outsourcing is often the more practical crypto market maker option for early and mid-stage projects because the infrastructure already exists. But “more practical” is not the same as “low risk.” You are swapping operational risk for counterparty risk. That can be a good trade. It is still a trade.

Cost structures: the invoice is only the visible part

Founders like to compare in-house payroll against an outsourced retainer. That is the wrong frame. The retainer is visible. The internal cost is smeared across engineering, treasury, infrastructure, monitoring, and mistakes.

Outsourced professional market makers may quote $2,000 to $20,000+ per month. That range is wide because the work is wide. A single small exchange with modest depth requirements is not the same as maintaining books across several CEXs, hedging externally, and providing detailed reporting for a token with volatile flow.

In-house costs rarely appear as one clean line item. They arrive as:

• developer time spent maintaining exchange integrations instead of building product;
• licensing or building trading software;
• cloud or low-latency infrastructure;
• security reviews for systems that can move treasury funds;
• operational coverage across weekends and market shocks;
• treasury capital sitting as inventory on venues;
• losses from bad quoting, stale prices, or poor hedging;
• compliance and reporting work when exchanges ask questions.

Here is the blunt comparison I use when teams ask whether outsourcing is “expensive.”

The sneaky part is that internal losses often feel like “market conditions” until someone audits the fills. A bot that quotes too tight during volatility can lose inventory. A rebalancing process that depends on slow withdrawals can leave one venue starved. A hedge that executes late can turn a neutral strategy into directional exposure.

Outsourced costs are easier to resent because they arrive as invoices. Internal costs are easier to ignore because they arrive as explanations.

Inventory risk and delta-neutral strategies

Market making is not simply placing equal bids and asks. If only.

The core problem is inventory. A market maker quotes both sides because it wants to earn the bid-ask spread while minimizing directional exposure. In practice, fills are uneven. Buyers lift asks. Sellers hit bids. Suddenly the market maker holds more or less of the token than intended. If the price moves against that inventory, the spread earned may be irrelevant.

This is why market-making algorithms generally lean on delta-neutral strategies. The objective is to reduce directional risk: keep exposure balanced, hedge when needed, and avoid becoming a long-only bag holder with better software.

For token projects, the inventory problem has a political layer. The market maker may need token inventory to quote asks and quote asset inventory to quote bids. If the project supplies tokens through a loan, the firm can support the book without buying all inventory in the open market. But the project then needs contractual clarity: how much inventory, what usage rights, what reporting, what return terms, what happens if liquidity dries up, and what happens if the token price collapses.

In-house teams face the same risk, only without the vendor wrapper. They must decide how much treasury to place on each exchange, how much quote currency to allocate, and how aggressively to rebalance. Every transfer has friction. Every exchange account is a security surface. Every delay can create slippage.

There are three practical inventory questions I ask before trusting any model:

1. Where does the hedge happen?

If the token is only liquid on one or two venues, hedging may be limited or expensive. A delta-neutral strategy needs instruments or correlated markets that actually trade.

2. What happens when volatility doubles?

A model that works in calm conditions can become a meat grinder when spreads widen and flow turns one-sided.

3. Who has authority to stop?

If the risk engine says pull quotes but the founder wants the book to “look active,” someone has to win. If it is not the risk engine, congratulations, you are doing treasury theater.

Inventory risk is also why “earning the spread” is not a business model by itself for a token project. Professional firms may profit from spread capture, rebates, and sophisticated execution. A project running an internal desk is usually trying to maintain market quality, not build a prop trading business. Confusing those motives leads to strange decisions.

Exchange compliance and liquidity depth requirements

Exchange listing conversations are never as clean as public-facing listing guides suggest. The exchange wants markets that trade. The project wants distribution, legitimacy, and volume. The market maker sits in the middle, turning that mutual dependence into an order book.

For many centralized exchanges, a designated market maker is part of the listing reality because minimum liquidity standards need to be met. That can mean maintaining a reasonable bid-ask spread, providing depth within certain price bands, and avoiding the embarrassing situation where a modest market order moves the token several percent. Exact requirements vary, and specific contractual terms are usually private. But the direction is consistent: no serious venue wants a book that looks abandoned.

Depth matters more than founders think. Spread gets the screenshots because it is simple: tight spread good, wide spread bad. But depth is where execution quality lives. A one-basis-point spread with $300 sitting behind it is cosplay. If a trader buys $25,000 and suffers ugly slippage, the book is not liquid in any meaningful sense.

A healthy market-making program watches both:

• Top-of-book spread: how far apart the best bid and best ask are.
• Depth near mid-price: how much size is available within practical price bands.
• Slippage on realistic order sizes: what a buyer or seller actually experiences.
• Venue fragmentation: whether liquidity is concentrated or scattered uselessly.
• Quote stability: whether orders vanish whenever flow appears.
• Inventory skew: whether the maker is quietly running out of one side.

The exchange cares because bad liquidity damages its own trading experience. The project should care because bad liquidity punishes real users. Nothing kills confidence like a token that can be bought but not sold without falling through the floor.

This is where outsourced firms often have an advantage. Known market makers already have exchange relationships, operational playbooks, and reporting formats that listing teams recognize. An internal team may be technically competent but still lack credibility. In exchange politics, competence without reputation is slower.

That does not mean outsourcing is automatically better. It means the project must price the value of exchange familiarity. Sometimes that value is real. Sometimes it is just a logo slide.

Hidden operational hurdles in proprietary trading infrastructure

The ugliest problems in in-house market making are not the glamorous ones. They are boring, repetitive, and lethal.

Exchange APIs behave differently. Balances do not reconcile cleanly. One venue delays withdrawals. Another changes a symbol format. A websocket drops during volatility. An order cancel request times out, but the order is still live. Your dashboard says exposure is flat, but one connector is stale. Then a large seller arrives.

This is not theory. This is the texture of the job.

A proprietary trading infrastructure needs constant maintenance because crypto venues are uneven. Even large exchanges can have interruptions, rate-limit changes, and strange matching behavior under stress. Smaller venues can be worse. If the token is listed across several exchanges, the internal team has to normalize data, reconcile fills, and keep risk calculations accurate in real time.

Security is another quiet monster. Market-making systems need API keys with trading permissions, and sometimes withdrawal-related operational workflows. A compromised trading key can wreck books even without withdrawals. A bad internal permission structure can let one rushed engineer or operator create financial damage. When the trading stack touches treasury assets, normal startup chaos becomes expensive.

Then there is staffing. Crypto trades continuously. A book can degrade while the team sleeps. Outsourced firms usually have coverage because that is their business. An in-house team has to build rotations, escalation rules, and incident response. If the answer is “our engineer gets Telegram alerts,” you do not have operations. You have a sleep deprivation experiment.

The other hidden hurdle is governance. Who decides quoting parameters? Who approves inventory transfers? Who can override the bot? Who talks to exchanges when depth slips below expectations? Who explains to the board why the treasury is holding more token and less stablecoin after a week of one-sided selling?

These questions sound administrative until the market moves. Then they become the only questions.

When in-house makes sense

I am not anti-in-house. I am anti-delusion.

In-house market making can make sense when the project already has serious trading DNA. That means the team understands order books, has engineering capacity, can build or license robust software, and treats risk as a first-class function rather than a slide in the finance deck.

It can also make sense for projects with unusual requirements: specific tokenomics, complex cross-venue strategies, sensitive treasury constraints, or a long-term plan to internalize liquidity operations as a strategic capability. In those cases, outsourcing may feel too blunt. A vendor can run the book, but it may not understand the project’s full capital structure or long-term market design.

Still, the bar is high. If you go in-house, I would want to see:

• a tested quoting and risk engine before listing day;
• venue-specific connector testing under stressed conditions;
• clear inventory limits by exchange;
• independent monitoring of spread, depth, and slippage;
• written authority for kill switches and emergency parameter changes;
• treasury controls around API keys and transfers;
• someone accountable for execution quality, not just bot uptime.

The phrase “we can build it” is not enough. Everyone can build the first version. The second version, after real fills, broken APIs, angry exchanges, and a one-sided market, is where the truth appears.

When outsourcing is the cleaner trade

Outsourcing is usually cleaner when speed, exchange requirements, and operational maturity matter more than absolute control. If a project is preparing for CEX listings and does not already have a trading desk, hiring a professional market maker is often the rational move.

The project gets existing infrastructure, experienced operators, and a counterparty that exchanges may already know. The book can go live faster. Reporting can be standardized. The internal team can focus on product, distribution, token utility, and the other problems it is actually built to solve.

But outsourcing only works if the contract and incentives are tight. I want to know exactly what the market maker is being paid to optimize. Tight spread? Minimum depth? Reduced slippage? Venue coverage? Reporting? If the mandate is vague, the result will be vague. If compensation rewards the wrong behavior, the book will show it.

Projects should also resist the temptation to select vendors based on swagger. Market making attracts a specific kind of confident talker. Some of them are excellent. Some are just fluent in opacity. Ask for concrete reporting examples. Ask how they manage inventory risk. Ask what happens in stressed markets. Ask what they refuse to do. The refusal list tells you more than the promise list.

The best outsourced relationship is boring in the right way: defined scope, visible metrics, disciplined inventory, no miracle language, no secret plan to “support price.” The book improves. Traders experience less slippage. Exchanges stop complaining. Nobody needs to pretend liquidity is community engagement.

The real decision: capability or dependency

The in-house vs outsourced market making debate is not about ideology. It is about where you want the risk to live.

In-house gives more direct control and less external counterparty exposure, but only if you have the machinery and people to deserve that control. Without them, it becomes a fragile internal prop desk funded by treasury and optimism.

Outsourcing gives speed, expertise, and exchange credibility, but introduces vendor dependency, token-loan risk, and incentive problems. With a good firm and a hard contract, that trade can be sensible. With a bad firm, you may not know the damage until the book has already taught everyone.

For most token teams, the honest sequence is simple: outsource first, learn what good liquidity metrics look like, build internal monitoring from day one, and only consider bringing execution in-house when the organization can challenge the vendor intelligently. Not emotionally. Not because the retainer feels annoying. Intelligently.

Crypto market making is not brand work. It is not a growth hack. It is market structure under pressure. Either you run it like a trading operation, or you pay someone competent to do it while you watch them closely. The middle path — half-built bots, vague vendor promises, and founder pressure to “make the chart healthier” — is where tokens go to get quietly punished.