evidoria

← Back to browse

Good practice Imported

SANDRA — Singapore's Conversational AI Assistant for National Statistics Discovery

Singapore · Singapore · See the Singapore profile · See the Singapore profile

Evidence: Descriptive / self-reported Top 66% 53/100 · Ask Evidence Copilot about this practice

Singapore's Department of Statistics built SANDRA, an AI chatbot letting the public query ~2,400 SingStat data tables from 70 agencies in plain language, using semantic search over ~1,000 vectorised time-series datasets.

2,400 data tables
SingStat Table Builder dataset coverage (as of 2026)
70 agencies
Public-sector agencies providing source data (as of 2026)
1,000 datasets
Time-series datasets vectorised for SANDRA (as of 2026)
SANDRA — Singapore's Conversational AI Assistant for National Statistics Discovery

Details

Maturity
Pilot
Promoter
Singapore Department of Statistics (DOS)
Period
2025–ongoing (Phase 1, beta)
Keywords
statistics, open data, public administration, digital government

Context

Singapore's Department of Statistics (DOS) manages the SingStat Table Builder, a platform hosting roughly 2,400 data tables sourced from 70 public-sector agencies; search-log analysis and stakeholder interviews found that users typically search using natural language and conceptual terms rather than technical table names, making the existing interface hard to navigate for non-specialists.

Activities

Working with delivery partner PebbleRoad through a four-stage design process (discovery, design, detailing, development), DOS built SANDRA (Statistics ANd Data Retrieval A.I. assistant): metadata from roughly 1,000 time-series datasets was converted into vector embeddings so a large language model can match a plain-language question to the most relevant table and return it as an interactive chart or table, with related-dataset suggestions.

Results

SANDRA Phase 1 launched in 2025 and, as of August 2026, is still labelled "beta" on singstat.gov.sg; it won an OpenGov Asia "AI-Powered Chat eXperience" recognition in May 2026. No independently published figures on query volume, answer accuracy, or measured time savings were found, so its demonstrated impact rests on documented reach and design process rather than measured outcomes.

Implementation
Indicative cost
Low (< €50k)
Time to results
Short (< 1 year)
Staffing & skills
Singapore Department of Statistics (DOS) owns the service, built with delivery partner PebbleRoad

Conditions for success

  • Design grounded in search-log analysis and stakeholder interviews showing users search in natural language/conceptual terms rather than technical table names
  • Four-stage design process (discovery, design, detailing, development) preceded build
  • Metadata converted into vector embeddings so an LLM can match plain-language questions to the right table

Common failure modes

  • Still labelled 'beta' as of August 2026, over a year after Phase 1 launch
  • No independently published figures on query volume, answer accuracy, or measured time savings

Where it fits

Governance type
national statistical agency with external delivery partner
Scale
national
Income level
high-income

Commonly funded by

National / regional programmes

Indicative funding routes for practices of this type — always check each programme's current calls and eligibility rules.

Do you run this practice? Claim it — verified implementers get a public contact pathway and can propose corrections.

Data sources

Where this practice's information was retrieved from, and when.

Attachments

Similar practices you may find useful