Digitising, digitalising, and the difference that decides a budget

In 2016 HMRC faced a question about order. It could require business taxpayers to keep digital records first. It could replace the legacy systems that administer those taxes first. Or it could do both at once.

The National Audit Office recorded the reasoning in its June 2023 report on the programme. HMRC considered whether to sequence the two elements, and concluded that sequencing them would not meet the published timetable, so it proposed to do both at the same time. The auditors’ verdict on the consequence sits a few lines later: HMRC did not fully assess the scale of work required at the outset of the programme, or the additional complexity of introducing digital record keeping at the same time as replacing its legacy systems.

That is what the digitising and digitalising distinction is for. Not as a taxonomy. As a warning that a plan is describing two different kinds of work and paying for one of them.

Requiring a taxpayer to keep records in software changes the medium of an existing process and leaves the process itself alone. Replacing the systems that administer a tax changes what the organisation does, in what order, with which people. Different risk, different skills, different failure modes, one programme board, one timetable, one line in the Budget.

The bill for treating them as one

HMRC’s original 2016 estimate for introducing Making Tax Digital across three business taxes was £226 million. Its 2023 estimate of the cost of introducing it for VAT and Self Assessment business taxpayers with incomes over £30,000 was £1.3 billion, described by the NAO as a 400% increase in real terms since 2016, with £642 million already spent by March 2023. The timetable for Self Assessment had been pushed back four times since 2015 and was at least eight years behind the original plan.

Now separate the two halves and the picture becomes legible in a way the headline figure hides.

The digitising half largely worked. Digital record keeping for VAT went in from April 2019 for larger traders, accounting for almost all of the VAT collected, broadly in the original timescale. Smaller traders followed in 2022, three years later than planned, a delay the NAO attributes to HMRC’s concerns about making changes during the EU exit process and the pandemic and about burdens on businesses during that period. Research reported by HMRC in March 2022 suggested a high likelihood that the change was generating additional VAT revenue, estimated at £185 million to £195 million in 2019-20, with the caveat that some of the increase may have been due to other factors.

The digitalising half is where the money and the delay went. Moving 3.2 million VAT records off the legacy system took 15 months longer than planned, because HMRC did not anticipate the extent of the data issues involved. Introducing the VAT changes cost around £295 million, or £322 million at 2022-23 prices, by March 2023, which is roughly £70 million more than HMRC had originally expected the entire programme to cost across all three taxes. Delays in moving those VAT records in 2022-23 then reduced HMRC’s capacity to build the Self Assessment system, which is the mechanism by which a data migration became a policy delay.

And the benefit slipped with it. In 2016 HMRC expected an annual return of £600 million in additional tax revenue from VAT and Self Assessment by 2020-21. It now expects to reach that level in 2027-28.

There is a further detail that inverts the usual assumption about which half is the risky one. The digitising half is the half whose scope had to keep shrinking. The original plan required digital record keeping for Self Assessment, VAT and Corporation Tax by 2020, with the Self Assessment changes due to complete by 2018. In July 2017, on HMRC’s advice, the government pushed Self Assessment back to at least April 2020, citing concerns that businesses and tax agents were not ready. In December 2022 it moved again, so that instead of requiring every Self Assessment business taxpayer with income over £10,000 to keep digital records from April 2024, only those above £50,000 would from April 2026 and those above £30,000 from April 2027. Requirements for those under £30,000 were still under review, and no date had been set for general partnerships or for Corporation Tax at all.

Each of those retreats was sensible in isolation. Together they describe a change that turned out to be about people’s habits, software readiness and agents’ capacity rather than about record formats, which is the thing digitising is always assumed not to be about.

Seven years of a benefit stream deferred, on a business case approved on the strength of it. Not because anybody was incompetent, and not because either half was a bad idea. Because the timetable that bound them together was fixed before anybody had assessed the scale of either.

The cost that leaves your balance sheet and lands on somebody else’s

There is a second difference between the two words, and it produces most of the opposition a programme meets.

Digitising a process often does not remove work. It relocates it, to the end of the process occupied by whoever has least power to object. Making Tax Digital is again the documented case, and the documentation is unusually blunt. HMRC’s May 2022 business case sought further investment. The NAO found that the cost-benefit analysis in it excluded £1.5 billion of upfront transitional costs falling on VAT and Self Assessment business taxpayers with incomes over £10,000, although the detail was in an annex. A separate provisional estimate put the net cost over the first five years to Self Assessment business taxpayers with incomes between £10,000 and £30,000 at £1.2 billion, around £460 each on average based on 2021 costs.

Every commercial organisation has a version of this. A supplier portal that replaces an inbox. A self-service form that replaces a phone call. A customer app that replaces a service desk. In each case somebody is now doing data entry who was not doing it before, and whether anyone counts that depends entirely on whether they sit inside the accounting boundary.

A benefit case built by counting only the costs that fall inside your own organisation is not dishonest. It is incomplete in a direction that reliably flatters the proposal, and the people at the other end of the process work out what has happened to them within about a fortnight of go-live.

The one place the distinction has legal force

Outside programme management there is a single context in which these two ideas are separated by something other than opinion, and it does not sort them the way the consultancy ladder does.

In April 2021 the IFRS Interpretations Committee published an agenda decision on configuration and customisation costs in cloud computing arrangements. Its finding, in the fact pattern examined, was that a buyer configuring software owned by its supplier will frequently have no intangible asset to recognise at all, for the simple reason that it does not control the thing it has been improving. Those costs go through as an expense. The exception is narrow and precise: where the arrangement produces additional code from which the customer has the power to obtain the future economic benefits and to restrict others’ access to them, the customer assesses whether that code is identifiable and meets the recognition criteria.

Notice what the dividing line is. Not ambition. Not whether the change is transformational. Control. Two programmes of identical scale, both changing how an organisation works, land on different sides of that line depending on who owns the software when it is finished. The accounting consequences of that for enterprise buying are worked through in the buying process nobody puts on the website.

For present purposes the point is narrower and more useful. If your programme is described internally as transformational and every pound of it is configuration of a product somebody else controls, the accounts will disagree with the vocabulary, and the accounts will win in the year the profit target is set.

Where the distinction is marketing

Now the part the definitional pieces leave out.

No standard, statute or professional body defines digitisation and digitalisation as a pair. There is no British Standard for it, no accounting definition, no statutory instrument. Which is why every framework using the words defines them slightly differently, and why arguments over whether a given initiative is one or the other never resolve and never need to.

The three-rung ladder that usually carries them, from digitisation up through digitalisation to digital transformation, is a sales structure before it is an analytical one. Its function is to establish that whatever the client has bought so far was only the first rung. It is effective, it is not dishonest, and it predicts nothing about whether a programme will work.

In ordinary British usage the verb digitise has covered both meanings for decades, and nobody outside the industry hears a difference. A board paper that spends a page establishing terminology has spent a page.

So the honest position is that the pair earns its keep in three decisions and is decoration everywhere else. It decides sequencing, because doing both at once has an evidenced cost. It decides who bears the transition, because moving a process onto a screen frequently moves work onto somebody outside your organisation. And it decides the accounting treatment, because control of the software determines whether spend is an asset or a charge.

If the word being argued over does not change one of those three, the argument is not about anything.

Three questions that replace the framework

After this lands, does anybody do different work, in a different order, with different information in front of them? If not, you have bought a new interface for an unchanged process. That is a legitimate purchase and it will not deliver the returns attached to changing the process.

Who controls the software the money is being spent on? Your auditor will ask, and the answer determines a line in the accounts that most programme boards never see until it has moved.

And whose costs are moving? Every process has two ends, and the saving at your end is frequently data entry at the other.

The single most expensive pattern in British transformation is a business case whose benefits require somebody’s working week to change, attached to a delivery plan in which nobody’s working week appears. Making Tax Digital is not that pattern. It is the rarer and more instructive one, where a genuine understanding of both halves existed and a published date decided that they would be attempted together anyway. What that entanglement costs once a rollout begins is traced in pilots clear, rollouts do not. Wider programme coverage sits at digital transformation.