Trading bots traders can trust

One platform for forex and crypto

The platform let professional traders work in both forex and crypto markets, so they no longer needed two separate systems. Its users knew trading well: signals, leverage, stop-loss and take-profit were everyday language.

The company's algo team was building trading bots: automated strategies that generate signals and trade on the user's behalf. Bringing those bots to users was a business priority, and it was a new feature for the product. These were experts, not beginners: they would rather see everything than be guided.

A list of requirements, not a defined problem

The brief arrived as three wireframes, each a list of what a page should contain. It said what to show, not why, in what order, or what the trader was trying to decide.

Designing for experts in a field I didn't know

I had never used a trading bot. The users had, and they would notice anything that felt naive. Three sources closed the gap:

1. Competitors. I reviewed existing bot platforms and mapped which features each offered, to separate what traders expect as standard from what would set us apart.
2. The team's traders. Conversations with in-house traders taught me their vocabulary and how they judge a strategy.
3. Desk research. Reading on bot types, signals and performance metrics filled in the rest.

The hardest part wasn't trading. It was the ambiguity. Every answer raised the next question, so I kept a running list of open questions and closed them one by one before drawing screens.

A bot trades real money on the user's behalf

A manual trade is one decision. A bot is a standing decision that keeps acting when the trader isn't watching. Before switching one on, a trader has to be confident it works.

The goal: let traders build confidence in a bot, with evidence they already know how to read, before it touches real money.

Scan many bots, act on one

Traders run several bots at once, across crypto and forex. Each card answers one question: is this bot doing its job? A status label sits next to the name, Running or Paused, so the state reads before any number. Below it, in the order a trader checks:

1. Money. What the bot was given, as an amount and a share of the balance, and its profit or loss so far.
2. Reliability. Signals with a win rate, and how many were won and lost.
3. Behaviour. The split between buy and sell signals, drawn to scale.
4. Now. The last signal: side, entry, leverage, take-profit and stop-loss.

A paused bot keeps its history but fades back, so the running ones stand out. Search by name or pair, a Running / Paused filter and sorting by P&L narrow a long list.

Trade-off:the card carries only the numbers needed to compare bots. Everything else lives on the bot's own page, one click away.

Build, test and report on one page

A bot's life is a loop: check how it is doing, read the result, adjust the strategy, check again. Splitting those steps across the product would break the loop, so each bot has one page with three parts:

1. Overview. What the bot was given, its signals, the buy and sell split, and the last signal in full.
2. Balance. How the bot's balance moved, from the last 24 hours to all time.
3. Strategy. The indicators the bot trades on, each with its parameters, and the timeframe. Indicators are added, edited and removed right here.

The same report serves a test run and live trading, so a trader compares like with like: the numbers that convinced them in a test are the numbers they watch once the bot is live.

That answers when a bot is ready: when its test numbers hold up in the same report the trader will watch once it's live.

Trade-off: a dense page. Accepted for expert users, who would rather see everything than click between screens.

A report that answers a trader's questions in order

The report is ordered by the questions a trader asks about a bot:

1. Is it making money? Balance over time, with the period's high and low marked as lines, and the change for the day.
2. How reliable are its signals? Signals by state, closed, live and pending, with how many were won and lost.
3. How does it win and lose? Total and average profit and loss per trade, and the buy and sell ratio.
4. What is it doing now? The last signal in full: pair, market, type and leverage, side, entry, stop-loss and take-profit. Copy to order turns it into a trade.

The on/off switch sits at the top of the side column, above the market and pair. Turning a bot on or off is never buried.

Colour carries meaning only: green for profit and wins, red for losses, orange for pending. Everything else stays neutral, so a loss is never missed.

Risk limits in plain terms, with no surprises

A trader sets how much a bot may use and when it must stop. Each control says what it does, not just what it is called:

1. Allocation. An amount and its share of the available balance, typed or set with a slider, so the size of the bet is always in view.
2. Stop loss. “Pause the bot when losses reach” a set amount.
3. Take profit. “Pause the bot when profit reaches” a set amount.

Editing a live bot changes only what comes next. The dialog says so: new settings apply to new signals, and open positions keep their stop-loss and take-profit.

Trade-off: these limits act on the bot as a whole. Per-trade stop-loss and take-profit still come from the strategy, so there are two levels to understand.

An assistant that answers in a trader's format

Traders asked the assistant about the market as much as about bots. Its answers are structured instead of a wall of text:

1. Asking. An empty chat opens with example questions, and suggestions appear while typing, so a trader sees what the assistant can do.
2. Facts first. A news question returns a short summary next to the numbered sources it came from, each with its time and a link to read more.
3. Opinion on request. Analysis is a separate step, behind its own button, so the summary stays factual.
4. Feedback. Every answer can be rated, so weak answers are visible to the team.

Trade-off: one more click for analysis. Accepted, because in trading a summary that quietly mixes in opinion can move money.

Shadowing the team's traders

To move fast, we sat beside in-house traders while they worked through the designs, and said nothing. We didn't explain or guide; we watched where they went. If a trader had to search for something, its place was wrong, and we moved it. Watching them work surfaced problems early, while the designs were still cheap to change.

From a feature list to a product flow

The requirements became three connected surfaces, built around one idea: a trader should trust a bot because of evidence, not because of a promise. When I left, the bots were in internal testing with the company's traders. These are the metrics I would use once they reach users:

1. Readiness. Share of bots that are tested before they go live.
2. Assistance. Share of AI answers rated helpful.
3. Retention. How long live bots stay switched on.

What I learnedIn a domain you don't know, the ambiguity is the work. Turn it into a list of open questions and close them one by one, before you draw a single screen.