Connecting Marketing Investment to Real Revenue


Pneumatic Vacuum Elevators manufactures residential elevators in North America and sells both directly and through a dealer network. The company invests heavily in advertising for a considered, high-value purchase, but every system told a different story about what that spend delivered. We connected them into one view, from the first ad click to the signed order.
About Pneumatic Vacuum Elevators
Pneumatic Vacuum Elevators designs and manufactures residential elevators for homes in North America. The company sells directly to homeowners and through a network of dealers, and invests heavily in Google, Meta and Microsoft advertising to reach buyers who are planning a considered, high-value purchase.
A typical buyer does not convert on the first visit. They might click an ad, call the company, speak with an AI receptionist, get passed to a dealer and sign an order months later. The median time from lead to order is 98 days, and one in five orders arrives more than a year after the lead. Every step of that journey leaves a trace in a different system.
The challenge
Marketing wanted to answer a simple question: which spend produces orders? Nine systems held part of the answer: the CRM, three ad platforms, GA4, Search Console, an operations order file on SharePoint, the AI voice receptionist and dealer spend sheets. None of them agreed.
Leadership wanted those answers during meetings, in plain language, with the same region filters and caveats an experienced analyst would apply. Instead, every question meant a request to an analyst and a wait.
Why the numbers did not add up
One campaign could carry three unrelated names across ad platforms, CRM tracking codes and spreadsheets, so results were split or double counted.
Ad platforms reported conversions, not leads. In one month Google Ads reported 773 conversions while the CRM recorded 306 leads from it, a gap of about 2.5 times.
Measured on a single date, recent months always looked like a collapse because their orders had not arrived yet. A young cohort could read 0.20% against 2.80% for an older one.
Phone calls had no web click attached, so leads that came from an ad but arrived by phone were not credited to advertising.
Our approach
We started by agreeing on three definitions with the marketing and operations leads: what a conversion is, what a lead is and what a sale is. Those definitions are enforced in every report, dashboard and AI answer. From there, every new metric had to match the source system's own numbers before it went live. If a total did not match the ad platform or the order file, we found out why before anyone used it in a decision.
What we put in place
A cross-channel leadership report for North America with spend, clicks, platform conversions, CRM leads and orders for every channel, printable as a PDF.
156 dashboards in 16 categories, from executive KPIs to attribution, dealers, the sales funnel, spend, geography and products.
Two date axes on every dashboard, lead created and order date, plus static and dynamic close ratios, so long sales cycles no longer distort the picture.
A Google Ads deep dive from campaign down to keyword and ad, with CRM leads and orders tied to the exact click.
Reporting for 16 co-op dealers, with leads, orders and spend per dealer.
An AI analyst that turns plain-English questions into answers, charts and follow-ups, using the company's business rules.
Weekly emailed reports, threshold alerts, exports to Google Sheets and Excel, and admin screens to reconnect sources, map campaigns and upload spend.
Under the hood: one warehouse, one set of definitions
All nine sources land in one cloud data warehouse, each in its own raw layer kept exactly as delivered so every number can be traced back. A modelled layer then resolves customer identity, joins spend to leads and orders, and applies the business rules once. The CRM history behind it covers about 115,000 contacts going back to 2019.
Between the warehouse and every screen sits a semantic layer of 37 tables, with their columns and join patterns, which the AI analyst reads as its context. Every query, whether it comes from a dashboard, the AI, an export or an alert, is checked before it runs: only read queries, only on approved tables. The platform has run in production since its first month, with more than 150 releases and 319 automated tests.
Why not Power BI or Microsoft Fabric?
A packaged BI tool would have shown each source's numbers side by side. It would not have reconciled them. The value here came from linking calls to the ads that produced them, keeping one campaign identity across three naming systems, and reading the same cohort two ways. Those are business rules, not chart settings, and they need to live in one place that every report, AI answer and alert uses.
Running the pipelines on the company's own infrastructure also removed per-connector SaaS fees, and the client's admins handle routine fixes, such as an expired token or a new campaign name, themselves.
What we found along the way
Phone calls were invisible to paid attribution. Once each call was traced to its originating ad, the share of leads credited to paid media rose from under half to 63%, matching what the business had always felt.
Organic search totals were about three times too low for six months after a connector change, and pre-2023 North America lead history was undercounted roughly eightfold because location data was incomplete. Both were restored and validated.
Microsoft Ads, about 16% of North America paid spend, was missing from executive totals. It is now included and joined to leads per campaign.
A weekly report caught campaigns targeting Egypt, the Gulf and Thailand that were being counted as North America. About 70% of paid-search leads are now tied to the exact ad click.
Built for the team to keep growing
Business users get answers in plain English and can save any answer as a dashboard.
Admins reconnect sources, map new campaigns, upload spend and set report recipients themselves.
The platform keeps evolving through regular releases and an ongoing working relationship with the Solverix team, with every change documented in plain language so the client's team understands why a figure moved.
The result
Leadership now reviews one trusted North America view of spend, leads and orders across Google, Meta, Microsoft and organic. Platform conversions sit next to the CRM leads they produced, never against sales, so inflated numbers no longer drive budget decisions. Organic, historical and Microsoft figures have been restored and checked against their sources, and the marketing team answers most questions on its own.
Closing perspective
For a business with a long, multi-step sales process, the value is not in more dashboards. It is in following one customer from the first ad click to the signed order and trusting every number along the way. With shared definitions and every source in one warehouse, PVE now invests in marketing based on orders, not clicks.
What really impressed me about Solverix was how fast they got our business. No endless explanations, they just got it! They brought fresh ideas, smart technology, and real efficiency to the table. These people know what they're doing, and it shows in the results.

Manuel Delgado
Chief Marketing
