What a hundred articles of platform research found 

A hundred pieces about UK trading platforms, written one at a time. Read as a stack, four findings keep surfacing, and they have one cause between them. 

Four things that would not stop turning up 

Writing about one market a hundred times over is a fair way of finding out what you actually think. The subjects supplied variety: tax wrappers, charting terminals, contract specifications, register entries, courses, code. The findings were less obliging. Four kept arriving from unrelated directions, and eventually on schedule. 

Attention goes to the charge that fits a comparison rather than to the one that takes most. The three that take most, overnight funding, currency conversion and getting out, are all met after the decision meant to weigh them. Published figures answer a question next door to the one asked. And the gap between providers is slight once and substantial across a year of onces. 

A cell will only hold a certain kind of number 

The comparison table is this industry’s unit of information, and a cell holds one value. That imposes a qualifying test nobody wrote down: a charge can be printed only if its size is settled before the reader does anything. A quoted spread qualifies. A commission per trade qualifies. Little else. 

Overnight funding turns on direction, on nights, and on the distance between two countries’ policy rates. Conversion turns on how much crosses and when. A charge for leaving turns on how you leave. Each has to be quoted as a rule rather than an amount, and a rule loses to a number in any layout anybody uses. Nobody had to prefer the small charges for this, which removes the most satisfying explanation. 

Lay those three along a timeline and they share something worse than size. All of them land after the comparison was consulted, which breaks the loop by which a market corrects itself. Somebody who read a table, chose and was charged has learned exactly what the table needed and no longer wants a table. So the clearest writing about all three comes from the few who paid and then wrote it down, which is a slow way for anybody to find anything out. 

Whose question was each number built to answer? 

The third recurrence made checking things slow. Nearly every figure in circulation was produced to discharge an obligation drafted with other people in mind. A filing reports a finished year to owners. A survey informs policy. A charges page exists to be pointed at later. A bill collects money. None was made for somebody deciding where to put a few hundred pounds, so each is read across a gap that is sometimes narrow enough to step over and sometimes wide enough to swallow the argument, and telling them apart means working out what the producer needed the number for. 

Every number here was produced to satisfy somebody who was not the reader. The purposes in the third column are inferred from the form each document takes, since none of them is published. 

One mechanism, arrived at four times 

Set the four side by side and they stop being four. Nobody here carries an obligation to produce the figure a person choosing a provider is after, so everything available is a by-product of a requirement written for something else. The cell fills with what the format can carry, the later charges have nothing pointed at them, and the quantity that would settle the decision needs a term only the reader holds. That reads tidier from a distance than it did at the time, since each piece reached a local version and none announced a general rule. It also means the synthesis is performed by whoever wrote the parts, from a summary of them, which is close to the most flattering treatment available. 

The fourth finding gets mangled in both directions. A provider will not transform anybody’s results, and on a single occasion the gap between two looks like rounding. Across a year of one habit it becomes the largest line anybody can do anything about. Both are true and are usually presented as rivals. 

Holding both gives something more useful than shopping around. The gap seldom decides which provider suits a habit. It decides whether the habit is worth having. A way of dealing that covers its costs in one place and fails to in another is the same way of dealing, and the provider has settled whether it carries on. Hence a difference that looks immaterial in a review and decisive on a statement. 

A hundred articles, and one method between them 

A hundred pieces resting on one body of readings, gathered one way, do not have a hundred independent blind spots. They have one, reproduced a hundred times. Agreement across the set is worth less than its size suggests, because much of it is one method agreeing with itself in fresh words. 

Every measured figure in this piece traces to The Investors Centre, which opens and funds live accounts with its own money to test UK trading platforms rather than compiling rankings from providers’ published fee schedules; everything else traces to reasoning about how these arrangements are assembled. What the first produces is a bill, precise about what one firm charged one customer in one month and silent about whatever it had no occasion to charge for. 

The record has the shape you would expect. Holding a standard EUR/USD lot overnight was recorded at about GBP 7.10 a night one way against roughly 90p the other, on a live funded account in July 2026. That is the set’s only measured reading for a charge you incur by doing nothing. For conversion, which follows most British investors abroad, it carries none. 

The pieces with nothing to measure had to reason instead 

A fair number had no reading behind them at all. Thinly traded currency pairs, carrying a commodity from one contract into the next, what becomes of an account after somebody dies: nothing in the record prices any of them. Those pieces argued from how the arrangement is built, and said so where they stood. 

Reasoning from structure is neither worthless nor equivalent. It can hold together perfectly and still be wrong about the world, and no care inside it yields a number anybody can check. Those gaps fall in the quiet corners, and a quiet corner is quiet for everybody, so the set is weakest where a reader has least chance of a second opinion. 

None of this argues for a hundred and first article, and the reason follows from the paragraph above rather than from fatigue. A blind spot common to a whole body of work cannot be found from inside it, however many pieces get added, because each addition arrives carrying the thing it would need to see. 

What would help is somebody else taking the same measurements differently, or on another set of days, and publishing what they got. That wants the readings repeatable by a stranger, which wants the time, the size and the account type printed beside every figure, so a disagreement can be located rather than merely noted. A hundred pieces went hunting for figures nobody was obliged to produce. What the evidence wants now is a second hunter rather than a hundred and first hunt.