Case study3Commas Fintech · Trading automationSenior Product Designer2024 · 12 months

Complex bots, made launchable

3Commas' flagship product is trading automation — DCA, GRID and Signal bots with dozens of parameters each. Powerful, and for most users impenetrable. I led the redesign of the whole bot line across 50+ screens: fewer steps to launch, and a product that finally explains its own value.

Product

3Commas — crypto trading automation platform. Bots are the company's flagship

My role

Senior Product Designer. Owned the entire bot line: DCA, GRID, Signal and more

Scope

50+ screens: creation flows, strategy management, backtesting, automation workflows

Users

B2C — hundreds of thousands of crypto traders. Testing, feedback, analysis with real users

9→6

steps to configure a bot — and activation up 18%

Product analytics · post-launch

+22%

backtesting usage after the advanced backtesting release

Product analytics · post-launch

+14%

active automated strategies; feature engagement up 18%

Product analytics · post-launch

The problem

A DCA bot is genuinely powerful: entry conditions built on indicators, safety orders, multiple take-profit targets, trailing, stop-loss. But that power came as a wall — nine configuration steps and dozens of parameters before anything ran. Most users never got to see what the bot could do, because they never got one launched.

The deeper issue wasn't the number of fields. It was that the product never explained why. Why these conditions, what a safety order buys you, whether your setup was any good. The bots were the company's flagship — and their value was invisible until after you'd already fought through the setup.

Setup as a gauntlet

Nine steps, dozens of parameters, and financial risk at the end of it. Every extra decision before launch cost activations.

Value hidden behind setup

Nothing showed users what a good configuration looked like or what their bot would have done. The product's worth was a promise, not a demonstration.

Expert vocabulary everywhere

ADX, Ultimate Oscillator, trailing deviation, safety order volume scale — the interface spoke fluent trader, and most users didn't.

One product, every screen size

Traders live on desktop and phone at once. The whole line had to work identically from a 2560px monitor to a 360px screen.

The calls I made, and what they cost

Two decisions shaped the redesign — one about complexity, one about value.

Decision 01

Cut the parameters — or restructure how they're revealed?

The obvious way to shorten a nine-step setup is to delete settings. But this is a trading tool: the parameters are the product for advanced users. Removing them would trade one audience for another.

✓ What I chose

Keep the power, restructure the path

I rebuilt the flow into a four-part structure — General, Entry order, Safety orders, Exit orders — with sensible defaults and advanced settings folded behind progressive disclosure. Nine steps became six. A beginner can launch on defaults; an expert can still tune every parameter, in a structure that finally has a logic to it.

✕ What I rejected

Strip it down to a "simple mode"

Faster to ship and great for a demo — but it forks the product into two experiences, doubles maintenance, and tells your power users the tool is no longer for them.

What it cost

Progressive disclosure is more design work than deletion: every parameter needed a decision about when it appears, what its default is, and what happens when it's ignored. Months of detail work instead of a week of cutting.

How I checked

Competitor analysis and user testing on the flows, then activation tracked in product analytics after release — up 18%.

Decision 02

Explain value with marketing — or build it into the product?

The hardest problem wasn't usability. It was that users didn't see why bots were worth the effort. The company's instinct was more features; my argument was that new features would stay as invisible as the existing ones until the product learned to demonstrate its own worth.

✓ What I chose

Make the product show its value in context

I built value into the flow itself: Trading opportunity and Market data insight blocks that surface real signals ("Buy signals, last 30 days") right where you configure; advanced backtesting so you can see what your setup would have done before risking money; inline video tutorials and tooltips at the exact field where confusion happens, not in a help center.

✕ What I rejected

Onboarding tours and promo banners

Tours get skipped and banners get blind spots. Value shown at the moment of decision beats value claimed at the moment of signup.

What it cost

Real market data in the setup flow is an engineering cost, and every insight block is another surface to keep truthful. We spent roughly four months on the DCA flow alone getting this right.

How I checked

Backtesting usage +22%, feature engagement +18%, active automated strategies +14% after release — tracked in product analytics.

The work

The creation flow

The four-part structure — General, Entry order, Safety orders, Exit orders — with deal-start conditions on indicators (ADX, RSI, Ultimate Oscillator, TradingView signals), multiple take-profit targets, trailing and stop-loss. Defaults carry beginners; disclosure carries experts.

Every state, not just the happy path

Bot status states (starting up, success, error), empty pairs, no results, mismatched quote currencies — the whole line was designed through its edge cases, because a trading tool that only handles the happy path is a support queue waiting to happen.

True responsive, 2560 to 360

The full flow designed and specced at 1920–2560, 1440, 1024, 768 and 360, plus the mobile app — with screener, modal and dropdown behaviors resolved per breakpoint. Traders switch devices constantly; the product doesn't degrade when they do.

A component system underneath

The whole line runs on a component library I built alongside the redesign: group headers and containers, pairs select, condition components, entry/safety/exit settings blocks, target groups, advanced settings — with a tokenized color system behind them. Fifty-plus screens stay consistent because they're assembled, not redrawn.

Why this counts

Flagship-scale consistency isn't a style guide — it's components. The same system thinking I later applied at Cycode started paying off here: one source for conditions, orders and targets across every bot type.

Results

All tracked in product analytics after release.

Bot activation

After the 9→6 step restructure

Base

Before

+18%

After

Product analytics

Backtesting usage

After advanced backtesting shipped

Base

Before

+22%

After

Product analytics

+18%

feature engagement across the bot line

+14%

active automated strategies

50+

screens shipped across all bot types

About these numbers

These figures come from the product analytics we tracked after each release. I've since left 3Commas and no longer have access to the dashboards, so I can't re-pull the exact panels — I'm reporting them as measured at the time. If precision on methodology matters in a conversation, I'll say exactly that rather than improvise.

Reflection

Four months on one flow was the right call — barely

The DCA creation flow took about four months of design work. No dramatic failure — just a product so dense that every parameter interacted with three others, and every simplification had to be tested against expert workflows. I'd budget for that density up front next time instead of discovering it sprint by sprint.

Value display is a product feature, not a marketing task

The biggest lesson: when users don't see the value of a powerful product, the fix belongs inside the flow — signals, backtests, context at the point of decision. Handing the problem to onboarding tours would have been easier and worse.

Where AI would fit today

This project predated AI in my workflow. Today I'd use it for the long tail: drafting the dozens of edge-case states, first-pass microcopy for expert vocabulary, and clustering user feedback. The judgment calls — what to default, what to disclose, what to cut — would still be the job.