Applying the lens in politics
Germany does not lack digital strategies. It lacks a state that can learn faster than it legislates.
Where the state loses to its own rulebook
In 2017 Germany passed the Online Access Act (Onlinezugangsgesetz), which ordered federal, state, and local administrations to bring 575 public services online by the end of 2022. When the deadline passed, the federal audit office counted 33 of the 575 available nationwide — and in 2023 Bitkom president Ralf Wintergerst described the digitalization of German administration as, "to put it pointedly, a failed state". Missing strategy is a weak explanation — strategy papers existed in abundance. Something about the way the change was attempted keeps producing this result, across ministries and across decades. Three pieces of the mechanism recur wherever a state tries to steer a complex society by specification.
The law names the outcome, the deadline, and the procedure — and none of it executes. Writing "575 services online by 2022" into a statute creates an obligation, but an obligation is not a mechanism: nothing in it tells a Kommune with three IT staff what to build first, and nothing carries what one administration learns into what the next one attempts. The specification gets mistaken for the work it describes.
SOURCES · Bundesrechnungshof 2023 · netzpolitik.org 2023Every new idea must clear legal review, works-council review, privacy review — across 16 federal states and multiple regulators — before anyone learns whether it works. Each review is individually reasonable; their sum is a system in which trying something costs more up front than doing nothing, so innovators wait in limbo for rules to be clarified. A process built to reduce risk ends up freezing the status quo — which is also a risk, just one that appears on nobody's checklist.
SOURCES · Bitkom 2023Rules are born constantly and almost never die. The fax machine survived in German administration into 2024 not because anyone defended it, but because the written-form requirements around it had no owner whose job was to retire them. Each rule was reasonable when written; the rulebook as a whole is something nobody chose, and shrinking it takes machinery no one has built.
SOURCES · heise 2024 · Bayern StMD 2023What the lens says
The three traps share one assumption: that somewhere near the top, enough knowledge can be gathered to design the right answer in advance. Each service to be digitized touches different local software, different staffing, different case law in each of the 16 Länder. The knowledge needed to put one Bürgeramt's processes online is scattered across its clerks, its IT contractor, and the citizens who will abandon a confusing form — and most of it only becomes visible once there is a version to react to. No ministry can collect it up front, however carefully it consults.
What a state can do — better than almost any other actor — is run comparisons. Sixteen Länder building sixteen variants of a service is usually filed under fragmentation. Read through the lens, it is a search strategy: judging finished attempts against a common yardstick takes far less knowledge than predicting which design will win, and the judging can be done from the center even where the designing cannot. The yardstick can be plain — for a residence-registration service, say, the share of applications completed fully online and the days from submission to decision: two numbers every Land can report and every citizen can check. Attempt, comparison, adoption of the winner — that loop is a feedback loop, the mechanism by which a system learns from its own results, and it is precisely the piece the Online Access Act never built. The statute specified the destination and omitted any way for anyone to learn the route.
In one sentence: a state steers a complex society less by designing outcomes than by setting conditions and selecting among the variants that grow — closer to gardening than to engineering. Why top-down design fails in complex systems generally is the theory panels' subject; this page stays at working depth. So what does selection look like as actual policy machinery?
2016 Top-down vs. bottom-up, feedback loops → 2026 See emergence in a running model →
What to do instead
Pilot the policy in one Land, one city, one agency — with a real budget, a written success criterion, and a kill date. A regulatory sandbox is a probe, not a loophole.
M3 Bottom-up before top-down.Treat federalism as the laboratory it already is: let sixteen variants run, measure them against the same yardstick, then standardize the winner nationally — instead of negotiating one grand design for years.
M6 Prune the rules.Give every new rule a sunset date and every ministry a pruning quota. A rule that must argue for its own renewal is a rule someone has read.
Retire these
Writing deadlines into law and calling it delivery— a statute is not a feedback loop.Waiting for the perfect nationwide solution— while Estonia keeps shipping imperfect ones and improving them.Regulating a technology before anyone has run it here— you cannot select among variants that were never allowed to exist.