The workflow before#
TechSmith makes Snagit and Camtasia and sells them through trials, e-commerce and renewals. Product usage data, e-commerce transactions, support tickets and marketing data each lived in a separate system. Analysts stitched them together by hand, spending 30+ hours a week on reports, and dashboards ran on daily batches that could be stale by the time anyone read them. Pipelines failed often enough that people double-checked the numbers. No one could follow a customer from first trial to renewal in one place, so customer success usually learned an account was at risk when it didn't renew.
What we built#
We built one data layer on Azure and Databricks that brings TechSmith's product, sales, support and marketing data together, then put a customer 360 view, self-service dashboards and churn scoring on top of it. Analysts stopped assembling weekly reports by hand. Customer success started seeing at-risk accounts well before renewal.
One home for every customer signal
Automated pipelines bring product telemetry, e-commerce, support and marketing data into a data lake laid out in bronze, silver and gold layers. Each load runs quality checks and raises an alert when something's off, and pipeline reliability went from 92% to 99.5%. We put checks on every load rather than only at the end, so a bad feed gets caught before it reaches a dashboard.
A customer 360 and self-service dashboards
On the gold layer we built a customer 360 view that follows each account from trial to renewal, plus Power BI dashboards for executives, product and marketing. More than 50 business users now answer their own questions, and report generation dropped from hours to minutes. Row-level security means each team sees only the data it should, which is what made opening up access safe.
Churn and cross-sell models
We trained churn prediction, segmentation and cross-sell recommendation models in Databricks and tracked them in MLflow. Customer success sees which accounts are likely to churn ahead of renewal and which are good candidates for an upgrade. The models run on the same governed data as the dashboards, so a score and a report never disagree about who a customer is.
How we delivered it
We built the foundation first: the data lake, pipelines and quality checks. The customer 360 and dashboards came next, so analysts got their 30 hours a week back early. The models came last, once the data under them was trusted. Infrastructure is defined in Terraform and changes ship through GitHub CI/CD, so TechSmith's own team can run and extend it.
Before and after#
| Measure | Before | After |
|---|---|---|
| Manual reporting time | 30+ hrs a week | Automated |
| Report generation | Hours | Minutes |
| Pipeline reliability | 92% | 99.5% |
| Customer churn | Baseline | 15% lower |
What's next#
With churn scores in production, the next step is an agent that turns each at-risk account into a drafted outreach plan for customer success to review. The customer success manager approves or edits the plan, so the relationship stays with a person and the prep work moves to the agent.
