Indonesia's government scheduled the national launch of Perlinsos Digital, its digital social protection programme, for 10 November 2026. Katadata reported on 21 October that the National Economic Council (DEN) said President Prabowo Subianto would open it personally, and that Luhut Pandjaitan expects about 50 million recipients, all drawn from deciles 1-5 of the national registry. At the time of writing we have found no report confirming the launch took place, so this analysis covers the design and the evidence available beforehand.
The strongest case for the programme
The case for reform is serious. When Social Affairs Minister Saifullah Yusuf explained the single-data instruction in June 2025, he said that about 45 percent of Program Keluarga Harapan and Sembako recipients were suspected of not being on target. A Social Affairs Ministry trial that quarter found more than 1.9 million recipients who did not meet the criteria, plus eligible people missing from the lists. Money that reaches the wrong households is money withheld from poor ones. A country of Indonesia's size, with thousands of islands, cannot fix that with paper registers and local-office discretion.
The plumbing is a legitimate public-infrastructure play. Presidential Instruction No. 4 of 2025, issued on 5 February 2025, merged three databases (DTKS, P3KE and Regsosek) into a single NIK-based registry, DTSEN, drawing on 15 ministries and agencies. The state secretariat's account of the instruction says it is mandatory guidance for all ministries, agencies and regional governments that distribute aid. Perlinsos Digital layers digital identity, payments and registration on top of it.
What the early evidence does and does not show
The pilot results are encouraging but thin. GovInsider reported that the first pilot in Banyuwangi did not yet cover the disbursement stage. Katadata's launch report cites a Bali study in which exclusion error fell by 24.1%, and it also cites 11.5 million registered households and some 775,000 registration agents. These are officials' figures. We found no independent audit of them.
The reporting is also inconsistent on basic scale. Within the same Katadata article, one official says trials ran in 514 districts and cities, while the communications minister later says 215. The Jakarta Post reported in June that testing was under way in 42 cities and regencies before expanding to all 541 nationwide. Biometric Update, in July, said "more than 40." Claims of high targeting accuracy therefore cannot be checked against a stable pilot baseline. Whether the pilot covered 40 districts or 500 changes what it proves.
Where the risks sit
A DPI system moves the point of failure from a local official to a data model. Three risks matter, and none is an argument against building it.
- Exclusion by decile. Eligibility rests on a household's DTSEN decile, a relative rank rather than a direct measure of need. A household wrongly ranked in decile 6 is locked out, and a digital system makes that error uniform across the country instead of local and fixable.
- Registration friction. Digital onboarding favours people with IDs, phones and connectivity. The poorest, most remote households are the ones least likely to register themselves, which is why the agent network matters as much as the software.
- Data concentration. A registry linking identity, payments and welfare status is a high-value target. The programme should state which agency is accountable for breaches and for any use of the data beyond welfare.
Proportionate regulation here does not mean slowing the rollout. It means a few cheap, specific safeguards that make the system self-correcting.
What a good launch would include
First, publish error rates. If officials claim accuracy near 90%, they should publish inclusion and exclusion errors per district and per decile, with the method, so that outside researchers can reproduce them. A system that claims to replace discretion with data should be willing to show its data.
Second, make appeals real. Katadata reports a complaint mechanism for people who miss aid because of DTSEN data. Its value depends on turnaround times and on fixing the record, not merely logging the complaint. Government should publish both.
Third, keep non-digital routes open. Agent-assisted registration and cash pick-up points should remain available until exclusion data shows they are no longer needed.
Fourth, scope the data. Limit use of the registry to welfare targeting by default, and require a legal basis, not a ministerial request, for any other use.
Why the stakes extend beyond Indonesia
Indonesia will host the Global DPI Summit in Bali on 22-25 March 2027, per Biometric Update's event listing. It is the summit's first edition in Asia, co-hosted by Co-Develop, the ITU, the UN, UNDP, the World Bank Group and an Indonesian government committee. Indonesia will arrive there with a national social-protection deployment to showcase. If the numbers hold up, it becomes a credible model for governments weighing the same trade-offs. If the claims were never tested, the programme risks becoming evidence for critics who say DPI is a surveillance and exclusion machine with good branding.
The pro-innovation position is neither to cheer the launch nor to wait for perfection. A programme that can cut verification from months to minutes, as GovInsider's reporting suggests Banyuwangi did, deserves to scale. It should scale with its error rates in public view, so that Bali in March discusses measured results rather than slogans.