Where the Data Actually Lives (and Why It Doesn’t Help You Today)

If you map out how a modern HR or People Analytics team operates, the stack is fragmented in a very specific way:

• Workday / BambooHR holds structured data like comp, tenure, role changes, and performance ratings

• Lattice / CultureAmp contains review cycles, feedback, engagement surveys

• Gmail / Outlook holds manager conversations, escalation threads, exit discussions

• Slack / Teams reflects day-to-day communication patterns and team dynamics

• ATS systems show hiring pipeline, backfill urgency, and team growth pressure

Each system captures a different slice of the same story: how someone is actually experiencing their job.

But when teams try to make this data useful, they run into a hard boundary. The analytics layer, usually Workday dashboards or a People Analytics warehouse in Snowflake, only sees what is clean, structured, and intentionally piped in.

Everything else, especially the messy, high-signal data, stays locked away.

Not because it is inaccessible in theory, but because extracting it is painful, slow, and often politically sensitive. So most teams default to what is easy to measure rather than what is actually predictive.

Meanwhile, the full history of this data is already being retained. Every email, every review, every document, every system snapshot is sitting in backup systems for compliance and recovery.

That dataset is far more complete than anything your analytics dashboard sees.

It is just never used.

The Missed Opportunity: HR’s Most Valuable Dataset Is Its Own History

The most predictive patterns in attrition are not found in a single system. They emerge across time.

You see it when:

• Performance reviews shift from specific to vague

• Manager feedback becomes less frequent or more generic

• Internal communication volume drops or changes tone

• Employees stop participating in cross-team conversations

• Compensation discussions start appearing more often in private threads

No single tool captures this end to end. But your backups do.

Backups are not just copies of systems. They are a time-indexed record of how your organization actually operates. They contain the sequence of decisions, conversations, and behaviors that led to outcomes like attrition, promotion, or burnout.

This is exactly the kind of longitudinal data that machine learning models need to be useful.

And it already exists inside your company.

Duplicati’s core idea is simple: instead of treating that data as something you only touch during a disaster, treat it as the foundation for building internal intelligence.

What This Looks Like in Practice

Imagine a People Analytics team trying to answer a very specific question:

“Which employees are at risk of leaving in the next 90 days, and why?”

Today, this usually turns into a combination of:

• Static dashboards in Workday

• Periodic engagement surveys

• Manual manager input

• Maybe a basic regression model built on structured HR data

The result is directionally useful, but often late and shallow.

Now consider a different approach.

Instead of starting from curated HR tables, you start from the full historical record captured in backups:

• Performance reviews across multiple cycles

• Manager 1:1 notes stored in documents

• Email threads around role changes or compensation

• Slack activity patterns within teams

• Historical org structure changes

Duplicati extracts and structures this data into formats that can actually be used, not just stored. That means converting raw backup data into something like Parquet or Delta tables, indexing it, and making it queryable alongside your existing analytics systems.

From there, your existing tools still apply.

• Your data team can pull this into Snowflake or Databricks

• Models can be trained using familiar workflows in Python, MLflow, or Weights & Biases

• You can build internal dashboards or apps with Streamlit or Tableau

The difference is not the tools. It is the dataset.

Instead of training on a narrow slice of structured HR data, you are training on the actual behavioral history of your organization.

Why This Replaces “HR Analytics Add-Ons”

Most HR analytics products try to layer intelligence on top of systems like Workday.

They are constrained by what those systems expose:

• Clean tables

• Limited history

• Predefined schemas

• Survey-based inputs

They cannot see across systems. They cannot reconstruct context. They cannot learn from how decisions actually unfolded over time.

So they end up optimizing for reporting, not prediction.

Duplicati takes a different path. It does not try to be another dashboard. It turns the underlying data, already captured and retained, into something that your existing data stack can actually learn from.

This shifts the role of People Analytics from:

“Reporting on what happened last quarter”

to:

“Understanding how behavior evolves and predicting what happens next”

The Practical Path to Getting This Working

This is not a rip-and-replace of your HR stack. It is a data layer that sits underneath it.

A typical starting point looks like:

1. Identify high-value use cases
   Attrition prediction, promotion readiness, manager effectiveness

2. Map where the data lives
   Workday, Lattice, email, Slack, internal docs

3. Use Duplicati to extract historical data from backups
   Instead of building one-off pipelines across each system

4. Structure and store it in your existing data platform
   S3 + Parquet, Snowflake, or Databricks

5. Train models or build analytics on top
   Using the same tools your team already knows

This approach avoids the usual bottleneck of building and maintaining dozens of fragile integrations. The data is already there. The challenge is making it usable.

The Shift

HR teams have spent the last decade building systems to capture data.

The next phase is about actually learning from it.

The irony is that the most complete version of that data is not in your dashboards. It is in your backups, quietly accumulating a detailed history of how your organization works.

Once that history becomes accessible, the role of People Analytics changes. It stops being a reporting function and becomes a source of real operational insight.

And for the first time, predicting attrition is not about guessing earlier. It is about finally using the data that already knew.