Daytrip: from one mega-campaign to a scored route portfolio Google wrote up.
Daytrip sells private car transfers on more than 60,000 routes. Google Ads treated them as one pile and kept feeding the same few winners. We scored every route, gave each segment its own campaign, widened the set of routes that get budget, and then put all of it under one Search Ads 360 bid strategy with a target per segment. Google Marketing Platform published a case study on that last step.
Monthly ad spend came down 40% and revenue did not drop. It kept growing year over year at the same rate, and absolute gross margin held or grew, so the removed spend had never been buying anything. Multi-Target Portfolios then ran bidding across the whole structure, and Google reported ROAS up 47% year over year.
- Monthly ad spend
- -40%
- Revenue, on 40% less spend
- Held
- Absolute gross margin, or more
- Held
- ROAS, per Google
- +47%
“Daytrip achieves 47% ROAS uplift streamlining bid management using MTP”
Google wrote and published a case study on the bidding part of this work. It names Daytrip, credits eCommvert as the agency, and quotes Daytrip’s Head of Performance. The Multi-Target Portfolio numbers on this page are Google’s. Here is the original document.
Read the original Google case studyThe starting point
Daytrip is a global travel platform founded in 2015. It sells private, door-to-door car transfers with professional drivers, in more than 100 countries.
When I first logged into the Google Ads account, I saw archaeological layers: six years of campaigns built on top of each other, with nothing ever removed.
- Keywords
- 17.3M
- Campaigns
- 63,846
- Ad groups
- 2.8M
- Ads
- 4M
The size was a symptom. The real problem was that all routes competed for budget in the same pool, and Google’s algorithm decided who won. It does what any machine-learning system does: it pushes most of the spend to whatever is already converting.
That left three things broken, and they fed each other. Routes with real potential never got enough budget to prove themselves. Nothing connected ad spend to backend profit, so a high-margin route got the same treatment as a low-margin one. And about 90% of revenue ended up riding on a very small share of routes, which meant one competitor or one cheaper train connection on those routes could take a large slice of revenue with it.
Step 1: score every route
Google Ads data alone could not fix this, because Google Ads only knows about the routes it already spends on. So we built a route score from three sources: backend data, Google Ads data and SEO data.
Backend data
Revenue, margins, booking volume, LTV
Google Ads
POAS, conversion rate, CPA, efficiency
SEO data
Search volume, demand, seasonality
Cash Cows
Budget: Highest priority- Criteria
- Proven winners. High revenue and margin. Consistent POAS above 150%.
- Strategy
- Maximize investment. Never miss traffic opportunities.
High Potentials
Budget: High when scaling- Criteria
- Strong margins. Good signals. Limited data but promising.
- Strategy
- Scaling candidates. Tomorrow’s Cash Cows.
Low Score
Budget: Baseline only- Criteria
- Underperformers. Low margins or poor POAS. Limited demand.
- Strategy
- Minimal maintenance. Test occasionally.
Sleepers
Budget: When diversifying- Criteria
- Never advertised. Zero historical data. Unknown potential.
- Strategy
- Controlled exploration. Discovery pipeline.
3 · Re-score continuously
Each source answers a different question. The backend says which routes make the business money, whether or not ads had anything to do with it. Google Ads says what paid traffic returns. Search volume says where demand exists and when the season starts, before the account has a single conversion to show for it.
Every one of the 60,000+ routes lands in one of four segments: Cash Cows, High Potentials, Low Score and Sleepers. High POAS but low margin? Not a Cash Cow. No ad data but strong search demand and good margin? High Potential.
Routes are re-scored on a schedule, not daily. A Sleeper that performs gets promoted to High Potential, and a High Potential that proves out becomes a Cash Cow. We re-label weekly or every two weeks. If routes jumped between campaigns every day, the campaigns would never stabilise and the bidding could not learn.
Step 2: one campaign per segment
The score only matters if the account structure follows it. We rebuilt the account so each segment runs in its own campaigns, with its own budget.
One mega-campaign with mixed routes
- Algorithm pushes budget to the same proven winners
- High-potential routes never get meaningful budget
- No control over strategic priorities
- The algorithm makes business decisions
One campaign per segment
- Algorithm optimizes within strategic segments
- We control where the 80/20 distribution happens
- Budget aligns with business priorities
- Human strategy + algorithmic execution
The rebuild collapsed 63,846 campaigns into 127 and 17.3 million keywords into 1,564. Of the old keywords, 99.85% had never converted.
Monthly ad spend came down 40%. Revenue kept growing year over year at the same rate, and the business earned as much gross margin in absolute terms as before, or more. Average margins didn’t move: the same margin came in on far less spend. Nothing was traded away to get the saving, which is the clearest evidence that the removed spend had never been buying anything.
Step 3: a broader portfolio
Cutting waste was half of it. The other half was growth, and the segments showed where it would come from.
Joining backend revenue to ad spend surfaced a group nobody had looked at: routes with real bookings from organic and other non-paid channels, and close to zero ad spend. Google Ads had never funded them, so it had no conversions on them, so it kept ignoring them. We gave these routes their own label and their own budget.
The Cash Cows did not shrink. The base underneath them got wider. Around 300 previously unfunded routes went from nothing to about 14% of revenue, and they performed much like the Cash Cows we had been obsessed with. There were just far more of them.
That wider base is also why total ad spend could grow again, this time on purpose.
Chart of contribution margin versus ad spend for two scenarios. A small product portfolio reaches its peak margin at low spend and then declines as the budget overspends a finite set of products. A large portfolio peaks at far higher spend and far higher margin, showing that a bigger catalog raises the ceiling on profitable scale.
Step 4: one bid strategy, many targets
Segmenting the account created a new problem. Every campaign now bid off its own conversion data only. The Cash Cows had plenty. The smaller segments, the ones we most wanted to grow, never gathered enough signal to bid well. And someone had to keep every target straight by hand.
A plain portfolio bid strategy would pool the data, but it wants one target, and segments with different economics should not share one number. Search Ads 360 Multi-Target Portfolios remove that catch. The portfolio is carved into subgroups and each subgroup gets its own target. Because Search Ads 360 sits above the ad platforms, one portfolio can also span campaign types and publishers, not only Google Ads.
That matters because the structure did not stay in one platform. It was built in Google Ads first, and the same segments and campaigns were then set up in Microsoft Ads (Bing). The labels live with the routes, not with the ad account, so they travel. Every route carries its segment in the catalog, which means the same labels can drive targeting in Meta Ads as well.
One bid strategy per campaign
- Each campaign learns from its own conversions only
- Smaller segments never gather enough signal to bid well
- Every target is kept straight by hand
One portfolio, a target per subgroup
- Every subgroup bids off the whole portfolio’s conversions
- Targets still differ where the economics differ
- One strategy to manage instead of many
eCommvert handled the integration of the Multi-Target Portfolio setup into Search Ads 360, so cross-channel conversion data lands in one place. That plumbing is the part people skip, and it decides whether any of this works. Automated bidding is only as good as the conversion data it optimizes against.
“By consolidating cross-campaign and cross-publisher efforts into subgroups with tailored targets [through Multi-Target Portfolios (MTP)], we significantly simplified our campaign management.”
Marek Lacina, Head of Performance, Daytrip
This is the step Google Marketing Platform wrote up and published, with Daytrip’s approval.
The results
The numbers below come from different stages and different periods, so they should not be added together. Each one measures the step it belongs to.
| Stage | Metric | Change | What it means |
|---|---|---|---|
| Scoring and restructure | Monthly ad spend | -40% | Two fifths of the monthly budget came out and stayed out. Year-over-year revenue growth held at the same rate. |
| Scoring and restructure | Absolute gross margin | Held | The business kept as much gross margin as before, or more, on 40% less spend. Average margins didn’t change; the wasted spend did. |
| Broader portfolio | Revenue from the top bucket | 90% → 70% | Not because the Cash Cows shrank. Around 300 proven routes that ads had never reached went from 0% to about 14% of revenue. |
| Multi-Target Portfolios | Return on ad spend | +47% | Year over year, as published by Google, alongside conversions +118% and cost per action -33%. |
Google’s write-up is direct about where its numbers came from, and we are keeping that framing. The year-over-year growth was the result of a holistic approach, driven by internal ad strategy work and favorable external factors. Multi-Target Portfolios were the bidding piece of it. The scoring, the segments and the broader portfolio described above are the strategy work it sat on top of.
What to take away
- 1
You can’t stop the 80/20. You can choose where it happens.
Any campaign pushes most of its budget to its top performers. In one mixed campaign the algorithm picks the winners. With one campaign per segment, the business decides which routes compete for which budget, and the algorithm does the rest.
- 2
Your backend sees revenue the ad platform never will.
Google Ads only knows what it spent on. The booking system knows every route that makes money, including the ones that were never advertised. Joining the two is how we found both the 40% of spend that bought nothing and the routes worth funding next.
- 3
A broad proven portfolio scales further than a narrow one.
A small set of routes hits diminishing returns fast. Past a point, more budget only buys more expensive clicks on the same winners. Hundreds of proven routes give the budget somewhere profitable to go.
- 4
Segment the campaigns, pool the bidding.
Splitting an account into segments gives you control and costs you data, because each campaign learns alone. A multi-target portfolio gives the data back: one bidding model across all segments, with a separate target for each.