Scoring eighteen predictions from 2018

Start with the number itself, because it is the most revealing thing on the page. Eighteen.

Nobody has eighteen convictions about a single year. Eighteen is what a list looks like when the author is hedging: name enough things and the arithmetic guarantees a few hits, and nobody goes back to count the misses. The genre survives on that asymmetry.

The arithmetic is not subtle. Give each item an even chance and a list of eighteen produces around nine winners by luck alone, which is enough material for next year’s piece about how the technology was called correctly. Shorten the list to three and the author has to be right, which is why nobody shortens the list. The length of a prediction list is inversely related to the confidence behind it, and it is the first thing to look at before reading a word of the content.

So rather than working through eighteen entries and awarding eighteen grades, this piece marks the property that separated the predictions that landed from the ones that did not. It turns out to be a single property, it is not technological, and it was knowable in 2018.

The forecasts that arrived on schedule were the ones with a statutory deadline behind them. Everything else drifted.

The one that landed, and can be counted

Open banking went live in the United Kingdom at the start of 2018, not because a market demanded it but because a competition remedy required the largest banks to expose account data and payment initiation through common interfaces. There was a date, and there were named organisations obliged to meet it.

Eight years later it is one of the few 2018 predictions with a measurement series attached. Open Banking Limited’s review of 2025, published on 29 January 2026, reports 16.5 million user connections by December 2025, up from 12.1 million a year earlier, 351 million payments across the year, a 57% increase, and 24 billion successful API calls, up 27%. Weighted availability stayed above 99.50% every month and average response times improved to 324 milliseconds, the fastest recorded. Swept variable recurring payments nearly doubled.

Now the detail that makes this the strongest entry on the list. The same document explains that user connections are counted per bank brand and are not deduplicated across brands, because Open Banking Limited cannot identify individuals using services through more than one brand.

That is a publisher volunteering the weakness in its own headline number. It is what a real measurement regime looks like, and almost nothing else forecast for 2018 has one. The technology prediction was correct, and it was correct because it was really a prediction about a regulator’s timetable.

The measurement exists for the same reason the technology does. A remedy imposed on named firms requires somebody to check whether the firms complied, which requires availability figures, response times and volumes, published on a schedule. A voluntary market standard would have produced adoption claims from vendors and nothing that could be audited. Britain ended up with an unusually good dataset about a payments innovation as a by-product of enforcement, which is not how anyone would design it and is nevertheless how it happened.

The one where the deadline itself moved, in public

Digitising tax was on every 2018 list in some form, usually dressed up as the disappearance of the tax return. It is arriving, slowly, and its progress is the clearest illustration of why a statutory deadline is worth more than an industry consensus even when it slips.

HMRC’s current eligibility guidance, updated on 26 March 2026, sets out the phasing. Qualifying income above £50,000 for the 2024 to 2025 tax year brings a taxpayer into Making Tax Digital for Income Tax from 6 April 2026. Above £30,000 for 2025 to 2026, from 6 April 2027. Above £20,000 for 2026 to 2027, from 6 April 2028.

Three thresholds, three dates, descending over three years. This is a programme that has been reshaped repeatedly since it was first proposed, and every reshaping happened in a published document that anyone can read against the previous one.

Compare that with a technology forecast, which is republished each January with the date quietly advanced and no acknowledgement that a previous version existed. A slipping statutory deadline is a better planning input than a stable industry consensus, because you can see it slip.

The one that was never a technology prediction

Blockchain dominated the 2018 lists, and in hindsight most of those entries were not describing a technology at all. They were describing a price.

The clearest evidence for that reading is what the regulator eventually had to do. On 6 October 2020 the Financial Conduct Authority announced a ban on the sale, marketing and distribution to retail consumers of derivatives and exchange traded notes referencing certain unregulated cryptoassets, effective 6 January 2021. Its stated reasons were that the underlying assets had no reliable basis for valuation, that market abuse and financial crime were present in the secondary market, that price movements were extremely volatile, and that retail consumers did not adequately understand what they were buying. It estimated retail consumers would save around £53 million a year.

Then, on 1 August 2025, the same regulator announced it would lift the ban on retail access to crypto exchange traded notes, with effect from 8 October 2025, on the basis that the market had evolved and the products had become more mainstream and better understood. The prohibition on retail access to cryptoasset derivatives remains.

Seven years. A ban, and then a partial reopening, both grounded in judgments about consumer comprehension rather than about whether distributed ledgers work. The 2018 forecasters were not wrong about the engineering. They were making a market call and filing it under technology, and the only dates the sector ever produced were the ones a regulator imposed on it.

The twelve or so that cannot be marked at all

Artificial intelligence in everything. Smart cities. Additive manufacturing at scale. Commercial drones. Augmented reality in the workplace. Voice as the primary interface. Autonomous vehicles. Chatbots replacing the contact centre.

Each of these appeared on 2018 lists, and none of them can be given an honest grade, for the same reason: they were never stated in a form that could fail. There was no threshold, no date, and no named organisation obliged to do anything. A prediction that cannot fail cannot succeed either, and marking it would be a performance rather than an assessment.

Some of these subjects do now have a public record, because a statute or a regulator eventually attached one, and those are the ones worth revisiting. That exercise is done in our marking of the 2019 forecasts, where the entries are individually gradable precisely because acts of Parliament, filings and regulator decisions have accumulated behind them.

The rule this leaves you with

Before acting on any forecast about the next eighteen months, ask two questions of it.

What deadline does this depend on, and who is legally obliged to meet it? If the answer to the second is nobody, the forecast is describing an appetite rather than a schedule, and appetites are not procurement inputs.

And where would I look next year to check? If there is no published series, no regulator’s return and no filing that would show the claim to be wrong, the claim is not a forecast in any useful sense. It is a mood.

Applied to 2018, that pair of questions would have separated the hits from the misses in an afternoon, without any knowledge of technology whatsoever. That is an uncomfortable finding for a technology publication, and it is the reason this desk publishes as much procurement and regulation as it does. The rest of that work sits under emerging technology.