Voltera Sees True Margin per Deal with BigQuery, dbt, and Hex Voltera sells and installs solar systems across the Netherlands. Its data sat in four separate tools, so management could see every sale but not what each one earned. Vero by Datum Labs brought all five sources into BigQuery, modeled true margin per deal in dbt, and built live Hex dashboards for management and every closer. Margin is now visible the day a deal closes, refreshed every 15 minutes, and 20 to 25 hours of manual reporting are gone each cycle.
About Voltera Voltera handles the full solar sales cycle. Customers come in through Meta Ads, move through planning and a signed quote with a closer, and finish with an installation at their property. Each stage is handled by a different team and recorded in a different system.
Each deal brings in revenue, but it also carries planner and closer commission, procurement cost, installation effort, back-office time before and after installation, and the ad spend that produced the lead. The real profit of a deal can be very different from its quote value.
The Challenge Every system did its job. None could show what a deal actually earned.
CRM: quotes and closed dealsSQL Server ERP: installation and procurementMeta Ads: acquisition campaignsMoneybird: invoicesCalculating margin meant pulling four exports by hand and hoping the timing matched. It rarely happened, so management made decisions on gross quote value.
Before: four manual exports. After: one automated model and one margin number per deal.
Why It Mattered Wrong performance signal. Closers, campaigns, and deals were judged on close count or lead volume, not profit.Ad spend not linked to revenue. Meta Ads spend could not be tied to installed deals, so budget could not follow what converted.Late visibility. Top performers and cancellation patterns showed up only after they became problems.Heavy manual work. Reporting took 20 to 25 hours per cycle.
What We Built Vero by Datum Labs delivered a full-funnel margin model on a fixed scope, deployed in Voltera's own GCP environment in under two months.
Ingestion with dlt: all five sources (CRM, ERP, Meta Ads, Moneybird, Manatal), schema evolution built in, synced every 15 minutes.Warehouse in BigQuery: sales, operations, and marketing data in one place, stored as raw and structured layers.Margin model in dbt: closed deals joined to installation, procurement, and ad spend, with every cost attributed per deal.Orchestration with Dagster on GKE: scheduled runs and sensors, with Slack alerts on failure.Two Hex dashboards: a KPIs Dashboard for management and an Agent Sales Dashboard for each closer.Reverse ETL: modeled segments sent into an AI Voice Agent SaaS, so the sales team can act on data without waiting on Vero.
How It Works Margin does not live in any one system, so the architecture is built to join them. Each layer has one job.
1. Ingestion dlt pulls data from the PostgreSQL CRM, SQL Server ERP, Meta Ads, Moneybird, and Manatal. Schema evolution keeps pipelines running when sources change.
2. Warehouse Data lands in BigQuery inside Voltera's own GCP environment, so the business owns its data and infrastructure.
3. Modeling in dbt Staging: cleans each raw source.Intermediate: joins deals with installation, procurement, commission, and acquisition data.Marts: Deal Margin, Cost Attribution, Installation Performance, and Agent Reporting.Every cost is attributed to its deal. Conceptual view, not to scale.
4. Orchestration Dagster on Google Kubernetes Engine (GKE) schedules every run and sends Slack alerts on failure, so issues are caught before they reach a dashboard.
5. Dashboards and activation Hex reads from the marts to power both dashboards. Reverse ETL sends modeled segments back into the sales workflow.
Tech Stack Voltera architecture, orchestrated by Dagster on GKE.
Layer
Technology
Role
Sources
PostgreSQL (CRM), SQL Server (ERP), Meta Ads, Moneybird, Manatal
Deals, installation, procurement, campaigns, invoices
Ingestion
dlt
5 sources, schema evolution, 15-minute sync
Warehouse
Google BigQuery
Central storage in the client's GCP
Transformation
dbt
Staging, intermediate, and mart layers
Orchestration
Dagster on GKE
Schedules, sensors, Slack alerts
Analytics
Hex
KPIs and Agent Sales dashboards
Activation
Reverse ETL
Segments pushed into an AI Voice Agent SaaS
The Dashboards The KPIs Dashboard gives management one view of sales, spend, and margin:
Pipeline: leads, proposals, closed deals, funnel completion, revenue, gross margin, estimated margin percentage, cost per client.Paid acquisition: platform-generated leads, ad spend, cost per lead, cost per click, cost per converted lead, ROAS, clicks, impressions.Trends: monthly revenue and margin, ad spend, and platform-generated leads.Closers: top closers by revenue, split by new business, renewal or expansion, and upsell.Deal flow: monthly volume across closed won, pipeline, and expansion.Each of the 10+ closers also has their own Agent Sales Dashboard.
Growth and Revenue KPIs dashboard in Hex (sample data)
Before and After
Area
Before
After
Margin
Not available; decisions on quote value
True margin per deal
Reporting
Four manual exports, 20 to 25 hours per cycle
Auto refresh every 15 minutes
Closers
Judged on close count
Judged on profit contribution
Campaigns
Judged on lead volume
Judged on installed revenue
Agent issues
Seen after the fact
Seen early
New questions
Manual pulls
Segments available via reverse ETL
Results True margin, for the first time: revenue, commission, procurement, installation, and acquisition modeled together per deal.No more manual reporting: 20 to 25 hours per cycle replaced by dashboards that refresh every 15 minutes.Ranked by profit: closers, campaigns, and deals ranked by what they earn, not volume.Smarter ad spend: Meta Ads budget tied to installed revenue, so it shifts toward what converts.Earlier agent insight: 10+ agents on their own dashboards, with top performers and cancellations visible early."For the first time, we could see what a deal actually earned us, not just what it quoted. And we finally knew which closers were making the business work." Voltera team
Why This Matters Quote value is easy to report but a poor number to run a business on. When every sale triggers commissions, procurement, and installation, real profit only appears once those costs are joined back to the deal. Voltera now works from one foundation where every deal has one margin number and every agent's performance is visible before it needs correcting.
.png)
.png)
.png)
.png)
