How Mavendo 3x monthly lead volume

How Mavendo tripled monthly lead volume

and cut cost per lead by more than 50%

When someone passes away, the family doesn’t research options for weeks. They pick up the phone. That single fact makes Google Ads for funeral services structurally different from almost any other lead generation account, and it was the starting point for everything Adverge built with Mavendo.

Mavendo Uitvaartzorg is an independent funeral provider in the Netherlands. Their entire intake process runs through one channel: inbound phone calls from bereaved families. The business operates across multiple cities and competes on local relevance – “funeral Amsterdam” and “funeral Groningen” are different markets, with different competitors, different prices and different search behaviour.

The situation

Mavendo came to Adverge managing their own Google Ads. The account had grown organically: a separate campaign for each location, bids managed manually based on availability, no call tracking in place. It was functional, but it had no foundation underneath it.

Three structural problems made optimisation almost impossible:

1. No phone tracking meant no real conversion data.
   A funeral provider that lives on phone calls but counts website form fills as conversions is giving the algorithm the wrong signal entirely. Google’s bidding can only optimise against what it can see. Without call tracking, it was learning nothing useful.
2. Not every incoming call is a real intake request.
   Wrong numbers, price comparisons, callbacks from the same family. If all of these are treated as conversions, the algorithm is rewarded for reaching the wrong people. The account had no way to separate quality calls from noise.
3. Demand is hyper-local and unpredictable.
   A passing away does not come with upstream digital signals. There is no research phase, no comparison shopping. The signal window is short and intent is binary. Generic national ads don’t convert in this category, but building and managing separate local campaigns for every city creates an account that quickly becomes unmanageable.

What we did

Adverge has been working with Mavendo for over two years. The account has been built in three distinct phases, each with the same underlying goal: more confirmed intake requests at a lower cost per request.

Phase 1 – Tracking as the foundation

The first intervention was rebuilding conversion tracking from scratch using Qooqie call tracking. Every call placed through one of Mavendo’s ad-driven phone numbers is now logged with full metadata: which campaign, which keyword, which landing page, and how long the call lasted.

Call duration became the first quality filter. Wrong numbers and short orientation calls almost always end within a minute. Real intake conversations don’t. By counting only calls above a minimum duration threshold as conversions, Adverge gave Google’s algorithm a clean signal to learn from. Non-qualifying calls stopped being rewarded with more budget. Genuine intake calls did.

That single change was the foundation for every improvement that followed. Without it, none of the later phases would have worked.

Phase 2 – Local segmentation

With tracking in place, the account was restructured around local relevance. Each city got its own ad copy, its own keywords, its own bids. Enquiry volume grew and cost per lead came down.

But the structure had a ceiling. Hundreds of ad groups per campaign, each requiring its own optimisation cycle. Scaling further would have meant either slowing iteration or adding headcount. Neither was the right answer.

Phase 3 – AI-driven scaling (ongoing)

With a clean foundation and a proven local structure, the account moved into its current phase: using AI to scale without adding complexity.

4,000+ ad customisers at keyword level. Scripts generate a custom ad text variable for every location-keyword combination. The result: 300 location-specific ad groups collapsed into a single consolidated ad group per campaign, while every impression remains location-specific in the ad text the searcher sees. The account is simpler to operate and gives Google more data per ad group to learn from, without losing the local message.

Dynamic landing pages with location parameters. A single page per service accepts a location parameter in the URL. Adding new cities means adding entries to a lookup, not building new pages. The page renders the right city, the right phone number and the right local copy on the fly.

AI iteration on ads and landing pages. Weekly AI-driven tests run on headlines, descriptions and on-page elements. What performs gets promoted, what doesn’t gets retired. The loop runs continuously at a scale no manual workflow could match.

AI call scoring at transcript level. Duration is a good proxy for call quality. But not a perfect one. Some short calls are genuine intake requests. Some long calls are not. The only reliable way to know is to read the transcript. In a market where every call represents a family at a difficult moment, getting that classification right matters beyond just signal quality. Mavendo doesn’t have the capacity to manually classify every call, so Adverge built a script that does it for them. The script reads each call summary and classifies the conversation as a confirmed request, not a request, or uncertain, with approximately 97% accuracy. That classification is fed back into Google Ads as the conversion event. The algorithm now optimises against actual confirmed intake requests, not a duration proxy.

The results

Cost per lead (CPL) came down by more than half while intake requests roughly tripled on a fixed monthly budget. The consolidation in Phase 3 has made the account easier to manage and faster to iterate on. The AI-driven phase is ongoing and early signals across all four metrics continue to move in the right direction.

Monthly lead volume	3x
Cost per lead (CPL)	Reduced by more than 55%
Account structure	From: 300 location-specific ad groups To: 1 consolidated ad group per campaign
Conversion signal quality	From: duration proxy To: AI transcript classification at ~97% accuracy