The OECD’s January report on European retail hands small shopkeepers one modernisation agenda with two halves. Local Retail, Global Trends treats digitalisation and environmental compliance as a single project, the twin transition, funded from a single balance sheet, a squeeze we have described before. The sharper problem is that the halves move in opposite directions: the digital one deflates at a speed set by competition among model vendors, the green one inflates on a timetable set by statute. A retailer holding one budget and two invoices buys whichever gets cheaper while it waits.

The digital half has the steepest downward price curve in business software. Epoch AI measured the price of a fixed level of machine capability rather than of a named model, across six benchmarks over three years, and found it falling between 9 and 900 times a year depending on which capability is fixed. For graduate-level science questions at GPT-4’s original standard, the rate was 40 times a year. Epoch notes the fastest drops are the most recent and may not persist. Even at a fraction of that rate, a shop that postpones an AI purchase by two quarters buys a better tool for less money.

Retail has now been measured on the return side, and the answer flatters the cheapest rung of the ladder. The OECD’s own analysis of retail firms across five EU countries finds that those with a website show 20 to 40% higher labour productivity than those without, controlling for age, size and region. It calls that a correlation rather than a causal claim. Even read at its most cautious, it hands the shopkeeper a number the green half of the agenda cannot match.

France runs the EU’s longest-standing textile fee, and it points the other way. Clothing, linen and footwear have carried extended producer responsibility there since 2007, and Refashion’s 2026 scale charges 10.98 cents on every pair of men’s denim trousers sold. Per garment that is trivial; per season it is a levy on volume rather than profit, payable whether the stock moves or not. The scheme adds a flat €30 administrative charge that the five-person shop and the five-hundred-store chain pay identically. Fixed costs shaped like that are how a compliance regime sorts firms by size without mentioning size.

The direction of that fee is now written into French law. Law No. 2026-602, promulgated on 8 July 2026, attaches a per-item penalty to the same eco-contribution from 1 September 2026, opening at between €0.25 and €12 an item, per public reporting on the decree. The schedule matters more than the opening level. That ceiling is legislated to climb toward €20 by 2030, which makes the cost of the green half knowable years ahead and knowably larger. The penalty is aimed at ultra-fast-fashion volume and the independent boutique will never trigger it; what it establishes is the instrument, since the lever it pulls is the contribution every marketer already pays.

Most other member states join that curve on a fixed date; the Netherlands and Latvia already run textile EPR schemes of their own. The revised Waste Framework Directive entered into force in October 2025 and gives governments 30 months to stand up textile EPR schemes under which, in the Commission’s words, producers “pay a fee for each product they place on the market”. That fee sits at zero across most of the EU, the Dutch and Latvian exceptions aside. A cost moving from zero to something is the steepest curve in this argument, and it lands in 2028 across the single market. UKFT notes that microenterprises get an extra twelve months, conceding the case before it starts.

Waiting is rewarded on one curve and fined on the other.

The instrument that measures the twin transition can see only one of its halves. Digital adoption is countable at firm level: the website exists or it does not. Environmental progress in retail resists that treatment, and the OECD concedes as much in its own summary, noting that “persistent data and measurement gaps continue to limit the ability to track SME-specific progress”. A dashboard fed by those two sources shows a precise rising line beside a vague one. If the vague line is also the expensive one, the composite reads as balanced progress while only the cheap half is being bought.

The survey evidence already shows what a low threshold buys. The Commission’s June 2024 Flash Eurobarometer on SMEs and green markets found that 93% of EU small and medium firms had taken at least one resource-efficiency measure, which reads as near-universal engagement. The two commonest measures are saving energy and minimising waste, each reported by 66% of firms, and both are housekeeping rather than capital expenditure. An indicator satisfied by switching off lights will keep climbing through a decade in which nobody retrofits anything.

The strongest objection is that these two curves converge rather than diverge. Most of the green burden on a small retailer is information work, and information work is precisely what is deflating. Cheap models should therefore drag the compliance bill down alongside the tooling bill, leaving the twin transition a single purchase after all. That holds only if the binding cost of environmental compliance is measurement rather than transfer, and it is not. A model can draft the declaration and estimate the fibre split; the certification layer absorbing that work is already growing. What it cannot do is pay a per-item levy, re-cut a garment or replace the refrigeration.

The fix is a timing instrument rather than another digital voucher, and member states still set their own EPR schedules. What nobody has proposed is measuring the transition by expenditure rather than by adoption, which would make deferral visible in the year it happens instead of in 2028. Until someone does, the twin transition will post its strongest numbers during the years its expensive half is being postponed, and every one of those numbers will be true. That is the choice facing whoever writes the indicators: count what small retailers have started, or count what they have paid.