Making datasets easy to find

A practical guide for dataset administrators to improve the discoverability of published datasets

data.gov.sg publishes dataset usage and search metrics every quarter - use these datasets to get an understanding of the discoverability and usage of your agency's datasets.

1. Dataset usage metrics

2. Top clicked pages

3. Top search terms

Introduction

This guide helps you make your datasets easier to find, understand and use. Well-written dataset descriptions appear prominently in search results and clearly show users why they're useful.

Part 1: Writing dataset titles and descriptions

A clear title and description is critical for search

Your title is the first thing users see in their search results. A good title should be clear, specific, and contain the main words people search for.

Datasets without descriptions are less understandable to users

In addition, the lack of a description means the dataset will appear with a generic description on search

A good dataset description will be reflected in search as well as the dataset page, which will help users better understand if the dataset is relevant to them. Many of the best practices for titles also apply to descriptions, although you can now write in more detail.

1. Keep titles under 60 characters and descriptions within 160 characters

• Titles cuts off after 60 characters in the search result

  

  Long titles beyond 60 characters get cut off in search

  • Keep titles succinct - "HDB resale prices by town and flat type"

  • Don't write "A comprehensive dataset containing information about..."

• The first sentence of the description should explain what the dataset offers

  

  State what the dataset contains in the first sentence

  • Descriptions can go up to 160 characters, but will still get cut off on search, so make sure the first sentence captures the key details.

  • Example: "Access monthly rainfall data from 63 weather stations across Singapore."

2. Use human readable search terms that people are using

• When thinking about the title, use terms that users are familiar with, based on their search terms. Avoid jargon as much as possible.

Top 10 search queries related to the elections dataset

We optimised the title and description to include the top search terms

• For example - we looked at the top 10 search queries related to elections and saw that many users were searching for GE2025, grc and electoral boundaries, which prompted us to include this in the title and description, which improved traffic dramatically

3. Be specific about the data inside the dataset to help users decide if it's relevant

Detailed titles help users understand what is in the dataset

When browsing for datasets, users will need to look at the title to see what might be relevant

• Clearly state the time period and geographic area covered

  • For example: "Covers all 26 planning areas in Singapore from January 2010 to present, updated every three months."

• Include key details like time period, location if possible.

  • For example: "Rainfall (Monthly, 2010-Present)"

Part 2: Understanding your agency's dataset metrics

The data.gov.sg has published the following datasets to help you better understand the impact of your agency's open datasets and identify gaps and bottlenecks that might be preventing your dataset from landing more impact.

• Dataset usage metrics

  • How often people see, click and use your datasets

• Top clicked pages

  • Which pages get seen and clicked on the most

• Top search terms

  • What people are searching for

How to interpret the metrics

1. Low impressions

   1. This means your dataset is hard to find and isn't showing up in searches

   2. Check the top search terms dataset for related search terms, try to match your title and description with what users are searching for

   3. Improve your dataset title to be more searchable, reference similar datasets with higher impressions

2. Low click-through-rate (many impressions, few clicks)

   1. This means your dataset is showing up in search, but people are not clicking on it, most likely due to a vague description

   2. You can rewrite the description of the dataset to make it clearer and more specific

   3. Reference other similar datasets that have higher click-through-rate

3. Many clicks and page views, but low consumption (e.g. downloads, API calls, subscriptions)

   1. This could be expected if the dataset already has good understandability and users are looking at the charts or maps available on the dataset

   2. However, it could also mean that the description was misleading, or might not accurately describe the data since users are interested about the data, but are not consuming it

   3. Make sure that your description accurately describes the file and be precise about key details like time period or location coverage

   4. Ensure that the file format is easy to use (e.g. CSV, GeoJSON, API)

Using search data

Compare popular searches with your datasets to find improvement opportunities.

Example: If many people search for "hdb resale prices" but your Housing resale transactions dataset doesn't get many clicks, add these exact words to your description.

Finding missing data: Searches that lead nowhere may show what people want but can't find. Consider whether your agency could provide this data.

Using popular pages data

Learn from the most visited and heavily used datasets to learn best practices:

• Compare your datasets to similar popular ones

• Read their title and descriptions to see what works

• Understand the type of dataset and potential use cases it can serve - e.g. analytical vs transaction

Quick reference:

Symptoms	Likely problem	What to try
Few impressions overall	Wrong search words	Add words people actually search for
Lots of impressions, few clicks	Description unclear	Rewrite the first paragraph to be clearer
Many clicks, few downloads	Description doesn't match content	Check that your description is accurate
Many searches, no results	Data doesn't exist yet	Consider publishing this data

Conclusion

Creating public impact with open datasets starts with making the dataset easy to find. This means understanding what users are searching for and improving your dataset title and descriptions. By monitoring dataset usage and search data, we can help more people discover our datasets. Every improvement makes your agency's public data more valuable and useful to everyone. Thank you for making Singapore's government data universally accessible and helping the public make data-driven decisions!