What Owning Your Ad Data Changed, by Industry

Each BigQuery marketing case study on this page is a brand that joined its ad data to real sales and published what changed: 5 cases across insurance, banking, B2B SaaS and retail, each with its source and date. The short answer to “does BigQuery improve marketing results?” is yes, when the sale data goes back into the ads.

Verti, an insurer, reported 20% sales growth at 15% lower CPA after it bid on predicted sales. Millennium bcp, a Portuguese bank, cut paid-media CPA by 36% with audiences its in-house team built in BigQuery ML. We also list the industries where we found no proof yet, online education included.

Plays from YouTube, which may set cookies.

Does owning your ad data actually change results?

In the 5 public cases we checked, yes. Each brand joined ad data with sales or CRM data and used it to steer targeting, bidding or budget. Each reports a gain: NTUC Income cut customer acquisition cost by 40%, and Hoff raised online ad ROI by 17%.

Each brand’s own published result on a Google page. None of these brands are nn.partners clients.

The move is close in every row: sales or CRM data is joined to ad data in BigQuery, and the result steers bidding, targeting or budget. We cover what BigQuery is and why ad data is rented in the first piece of this series.

Why is the average BigQuery marketing case study hard to trust?

Many data stories report clicks, speed or dashboards instead of revenue, and many come from the vendor that sells the tool. Google sells BigQuery too, so read every case below as a best case. We kept only results with a named brand, on a Google page, and we opened each one on 19 Oct 2026.

• Vendor wins: the company that sells the tool publishes the number.
• Commissioned studies: some big figures describe a composite company in a study the vendor paid for. We leave them out of this page.
• Moving baselines: Millennium bcp compared its new audiences with its other first-party audiences. That is a harder test than no targeting at all.
• Age: Verti’s story is from 2018 and Hoff’s work began in 2016. We say so in each section.

The fastest proof is your own account. Request a Data Audit: we check which of your ads bring closed deals, from click to CRM, in 48 hours, then walk you through it on a 20-minute call. It is for advertisers spending $30K or more a month. You can also book 45 minutes with us.

What changed for lead-gen businesses with a long sales cycle?

The strongest BigQuery results in lead generation we found come from insurance. Verti, a direct insurer in the Mapfre Group, fed a sale prediction built from BigQuery data into its bidding and reported 20% sales growth at 15% lower CPA. NTUC Income, a Singapore insurer, joined CRM, email and web data and reported a 40% lower customer acquisition cost.

Verti, insurance

The problem, in Verti’s words: the purchase cycle “takes several weeks to finalize an acquisition, and our Target CPA algorithms can only guarantee a control over the cost per lead“. The customer journey starts online and ends offline.

What it built:

• GA360 user data exported to BigQuery.
• A machine-learning model that predicts the probability of a final sale just after the lead happens.
• Offline conversion import that feeds the predicted sales into SA360, the bid management tool.

The published result: “20% sales growth at -15% CPA” (the PDF’s results list words it “+20 incremental sales at -15% CPA”), in a campaign the PDF calls previously maxed out, with open budgets and flat query growth. A second model grouped customers and found a target audience that reports “+50% conversion rate from lead to sales”. The PDF states no baseline for that figure. It also reports “spotting of 30% incremental sales undetectable by conventional offline conversion tracking”. Of the cases we found, it is the clearest offline conversion import case study: the predicted sale went into bidding as an offline conversion.

The story is from 2018. The mechanism still holds, and the tools have moved. Google’s API page now sends new users to the Data Manager API for offline conversions: “Use the Data Manager API instead”. Its help page says “you might upload data from BigQuery for Enhanced Conversions for Leads, then reuse the same connection to bring data into Customer Match”. We walk through the full path in how the sale gets back into bidding.

NTUC Income, insurance

NTUC Income launched 10 use cases between 2018 and 2020 with a Google Marketing Platform partner. It joined CRM, email and web data in BigQuery and scored people on their likelihood to buy. Its head of digital business says: “We achieved a 40% reduction in customer acquisition cost and an increase in click-through rate by 92% by connecting our audience with what they are interested in.”

One more pattern for businesses with a sales floor: Assurance IQ, an insurance marketplace with agents, uses Looker to time TV ads to when agents are free. Google’s page gives no cost number, so it counts as a pattern only.

What changed in banking and financial services?

Millennium bcp, a Portuguese bank, built BigQuery ML audiences of past personal-loan buyers for a 2023 campaign. In paid media, conversion volume doubled, conversion rates were 1.9x higher and CPA dropped by 36%, compared with other first-party audiences. The bank’s own in-house digital marketing team ran it.

Google’s page says it plainly: “In paid media, conversion volume doubled, conversion rates were 1.9x higher, and the CPA dropped by 36%.” For owned media such as email, it reports conversion rates 2.6x higher and conversion volumes 2.4x greater than other first-party audiences.

The baseline is audiences the bank already built from its own data, so the 36% is the extra lift from modeling on top of the bank’s own data.

What changed for B2B SaaS?

Freshworks, a software company, sends keyword spend, ad clicks and visitor-to-lead conversions into BigQuery and joins them with its CRM to judge lead quality by campaign. It runs up to 4,000 campaigns a year, which account for about 70% of revenue. Google’s page reports a 50% increase in return on investment.

In Google’s words: “Integrating BigQuery with customer relationship management systems enables Freshworks to sharpen its focus to the right campaigns, regions, and keywords, driving a 50% increase in return on investment.” The team also scores visitors and pushes that information back into Google Ads. Freshworks judges each campaign by the quality of its leads.

We leave out the page’s fivefold lead growth, because the page ties it to new local-language campaigns.

What changed in retail with call centers and stores?

Hoff, a Russian furniture retailer, started in 2016 to put online, call-center and store sales into BigQuery. It built new attribution models there and exported the results to a bidding platform that changes bids automatically. Google’s case page reports online advertising ROI up 17% and new keywords found.

The problem, in Hoff’s words: “we spend money advertising to them, then they might go to the physical store and buy something but we couldn’t connect these sales to our online data.” And the fix: “Now all our data is stored in Google BigQuery from multiple sources like online activities, call center orders or sales from the physical stores.” If your orders start online and close by phone, this is the closest public example.

What about online education?

We found no online education brand that published a BigQuery ad result on a Google page. Mindvalley, an online courses company, says it was “capturing all user data in BigQuery” by 2016, with no marketing number. One vendor reports a 200% lead-to-conversion lift for an unnamed EdTech company, with no baseline.

That claim sits on EasyInsights’ own page, a vendor writing about its own client.

Our own education result: we helped a California-based online school triple its student count and reach 47% of the California market, up from about 13%. The case page shows the whole program, data work included. It stays out of the table above, which holds only brands’ own results on Google pages. Read the full online school case.

What have we seen in our own work?

A client with a sales team grew the share of its highest-value leads from 10% to 30% after we changed which events are recorded and sent to the ad account. Another client, who once feared Google would misuse their data, had the biggest growth in their history in revenue, efficiency and profit within a year. Neither is a BigQuery result.

Here is how we build the tracking layer.

What is still unproven, and how do I know if my business fits?

We found no quantified public case in legal, home services, real estate, lead-gen call centers, automotive or healthcare. You fit the pattern in these cases if your sales close outside the ad platform and you run a CRM that records them. Our Data Audit is for advertisers spending $30K or more a month.

We have experience with CRM platforms, from big names that offer a lot of out-of-the-box solutions, along with fully custom-made CRMs that teams built inside their organizations.

If you want to know what this would change on your own account, ask for a Data Audit: 48 hours on your own data, then a 20-minute call. Or book 45 minutes. The analytics and data intelligence service page shows the full scope.

Next: what it costs, who you need and the first 30 days.

Frequently asked questions

Are these nn.partners clients?

No. Each case is the brand’s own result, published on a Google page and linked. Our own results are in a separate section and stay unnamed.

Why are some of the cases old?

Verti’s story is from 2018 and Hoff’s work began in 2016. The mechanism, sending sale data back into the ads, is the same today. The upload path moved to Google’s Data Manager.

What does “compared with other first-party audiences” mean?

Millennium bcp measured its BigQuery ML audiences against audiences it had already built from its own data. The gain is the extra lift from better modeling on top of the bank’s own data.

Is there a BigQuery case study for online schools?

We found no online education brand that published a BigQuery ad result on a Google page. Our own result, a California online school that tripled its student count, is on our case page.

What result should I expect?

Nobody can promise a number before seeing your data. What the cases share is that sale or CRM data started to steer the ads and the budget.

Do I need machine learning to get these results?

Not to start. Verti and Millennium bcp used models, but the first step in each was getting the sale data next to the ad data, which is what most accounts are missing.

By Nik Vdovenko, founder of nn.partners, and the nn.partners team. nn.partners is a Google and Meta Partner, in the top 1% of PPC agencies on Upwork, with 9 years in the work and $10M+ in ad spend managed.

Researched by the nn.partners team from Google’s own documentation and the named case studies. Every number links to its source. Last checked: 19 Oct 2026.

Sources

• Verti case study (Google, with Making Science), 2018
• Millennium bcp (Google Cloud)
• NTUC Income (Google Cloud)
• Freshworks (Google Cloud)
• Hoff (Google Cloud)
• Assurance IQ (Google Cloud)
• Mindvalley (Google Cloud)
• EasyInsights EdTech page (vendor)
• About Google Ads Data Manager (Google Ads Help)
• Manage offline conversions (Google Ads API)