710 Memorial · UX/UI Designer

WhatsApp archive

A secure internal archive of conversations from October 7, 2023: searchable without losing their context.

Data visualization SearchUX Built with AI In progress · 2026
Messages from a WhatsApp conversation of 7 October 2023, preserved in the archive as they were originally sent.

About the organisation

The system was built for a memorial non-profit dedicated to preserving, researching and documenting WhatsApp conversations collected during the events of October 7–8, 2023. 710memorial.org

Goal

Create a secure internal archive that enables searching, retrieving and exploring those conversations, while preserving context, relationships and permissions.

My role
UX/UI Designer
Team
Product Manager, Content Analyst, CEO
Tools
Lovable, ChatGPT, Supabase, GitHub
Status
In progress

On the material. This project preserves messages sent during the events of 7–8 October 2023. The conversation above is real, published as part of the memorial organisation's documentation work. The interface clips further down run on the anonymised dummy dataset built for development: the production archive itself is encrypted and access-controlled.

The challenge

4 constraints

What made this hard

Four constraints, none of them about the interface.

  • Extremely sensitive and traumatic content.
  • Large-scale, unstructured WhatsApp data.
  • Complex permissions and encrypted fields.
  • Multiple views over the same underlying data.

Working without the data

I built a fake database so I could design against the real one.

Real production data is encrypted and unavailable during development. So I designed a fully anonymised dummy database that mirrors the production schema. That allowed realistic UX design, safe exploration of the system, and readiness for a secure connection to the encrypted data later, without a single real message being handled to make a layout decision.

The anonymised dummy database mirroring the production schema.

Discovery

3 inputs

01. Target users · two roles

One needs breadth across everything. The other needs depth inside one conversation.

That tension is the reason the archive has more than one view: the same dataset has to serve someone sorting all of it and someone reading a single thread closely.

Content analyst
/ content manager

Responsible for organizing, validating and managing large volumes of WhatsApp conversations and related content.

Pain points
  • Information is scattered across multiple conversations and sources.
  • Manual tagging and validation are extremely time-consuming.
Needs
  • Broad visibility across all content.
  • Ability to quickly locate relevant conversations, people and timeframes.
  • Confidence for long-term use.
Primary use cases
  • Participant search: find all chats uploaded by a specific participant.
  • Storytelling & curation: identify messages with a specific tone or tags for storytelling.
  • Cross-referencing: cross-reference chats from a single residence.
  • Media evaluation: preview media inline to evaluate relevance quickly.
  • Data export: export filtered datasets for further analysis or workflows.

Researcher
(external / internal)

Uses the archive to examine conversations deeply, identify patterns and extract insights for research and documentation.

Pain points
  • Difficulty working at scale across many conversations.
  • Losing narrative continuity while analysing large datasets.
Needs
  • Advanced search and filtering capabilities.
  • Easy switching between chat-based and table-based views.
  • Ability to export and reference findings for research.
Primary use cases
  • Verification: compare anonymised content with the original before publishing.
  • Exhibition selection: select audio, video or images for exhibitions.
  • Quality assurance: validate summaries against raw chat content.

02. Best practices · HCI literature

With material like this, restraint is a design requirement.

  • Context preservation: a WhatsApp-style chat view and persistent metadata keep the narrative intact.
  • Sensemaking at scale: multiple views (chats, messages, people) support gradual exploration.
  • Cognitive load reduction: clear separation between narrative browsing and analytical table views.
  • Ethical and emotional sensitivity: a calm UI, no gamification, explicit permission handling.

03. Benchmarking

Three reference sets: data tables, chat tools, and memorial archives.

Tables (Airtable, Notion, Google Sheets, DataGrip) for analysis at scale. Chat tools (Slack, WhatsApp, Gmail) for how conversation is read. And the archives that do this work properly: the USC Shoah Foundation, Yad Vashem, the Smithsonian, and the 9/11 Memorial Museum.

Benchmark of archive and research tools handling large conversational datasets.

Design

3 views, one dataset

System flow

Every route through the archive, including the ones permissions close.

From VPN check and login through role permissions, into the three views, and out again through filtered exports, with anonymised and identified exports as separate paths, because who may see what is part of the flow rather than a setting.

The full system flow: VPN check, login, role permissions, the chats, people and messages views, filtering, media preview, and anonymised or identified export paths.

Solution & features

Three ways into the same material, so nobody has to work the wrong way.

01

Chats view

A WhatsApp-style conversation view showing messages as they originally appeared, enriched with metadata: participants, tags, a conversation summary, in-conversation search, download, and quick navigation into a table view filtered to that chat.

The chats view: a conversation shown as it originally appeared, with the message list alongside. · 9.5 MB

02

Contributors (people)

A directory of everyone who shared conversations, with the details they provided, the conversations they contributed, access to extended metadata, and quick navigation to either the chat view or a filtered message table.

The contributors directory: everyone who shared conversations, with their stated relationship. · 4.1 MB

03

Messages

A unified table of all messages across conversations, with advanced filtering, search and tag-based exploration, and export either anonymised or with identifiers, depending on permission.

The unified message table with filters, tags, inline media and CSV export. · 5.2 MB

04

Global capabilities

Search across the whole system, favourites, role-based permissions, encrypted fields, and multiple views over the same underlying data.

Global search across the archive, with tag-based filtering. · 1.5 MB

Trade-offs

  • Context vs. scale: preserving full conversations improves understanding but slows large-scale analysis.
  • Flexibility vs. complexity: multiple views serve different users, at the cost of more complex navigation.
  • Security vs. accessibility: strong permissions protect sensitive data while limiting visibility and export.

Future exploration

  • Controlled GPT API integration.
  • Chat-based conversation views.
  • Secure connection to encrypted production data.

Key learnings

AI is most effective as a thinking partner, not an automated solution. Data modelling is a core UX decision; the schema decided what the interface could ever offer. And dummy data unlocks progress in sensitive environments: building an anonymised mirror of the production schema was what made it possible to design at all, without touching a single real conversation.