Optimizing SEO, content marketing, and Google Ads with data at Car Next Door
Digital Marketing Manager at Car Next Door
Try Google Cloud
Start building on Google Cloud with $300 in free credits and 20+ always free products.Free trial
As the SEO and SEM (search engine marketing) manager at Car Next Door, a peer-to-peer car rental service, I’ve learned a lot about how to use marketing data in a way that yields remarkable results. In this case, “remarkable results” means a 3x increase in organic search traffic and reducing the CPA (cost per action) from SEM from $100 to $30 — all by leveraging data to optimize SEO, content marketing, and Google Ads.
Those results delivered not just the satisfaction of a job well done, but also the joy of knowing my efforts were helping to do something good for the world by reducing the number of cars on the streets. My easy access to data throughout this journey has made it possible to make better decisions, track progress, and communicate wins internally.
Optimizing SEO to triple non-brand organic search traffic
The first project I worked on was a complete overhaul of the information architecture of the website to improve its SEO performance. This involved a total redesign of the website structure and the creation of thousands of new landing pages. This was a crucial first step because SEO rankings are highly influenced by how easy your website is to navigate as a whole.
Building new pages
Previously, the website had around 15,000 pages, with one page for each variation of a keyword (such as “car hire Sydney”, “car rental sydney”, “sydney car hire”, and so on), and no logical hierarchy — they all sat in a flat structure beneath the website homepage.
With the help of the UX and development teams, I set out a new structure for the website.
I created new landing page templates, implemented improved horizontal and vertical linking to help Google to crawl the site, did keyword research to determine which keywords to target, and spun up some unique introductory paragraphs for each page. My developers built a basic CMS (just a big Google Sheet) that would allow me to control, test and improve the content of the pages once they’d been released.
In order to get SEO work prioritised in the future, it was important that the business could see and understand the impact that the project would make on key metrics such as borrower growth and revenue. I built a Looker dashboard that would help us track performance over time.
It was necessary to split out the traffic by vehicle type for two reasons:
- Firstly, a customer’s lifetime value varies significantly depending upon the type of vehicle they first choose (a person who borrows a car will have a higher LTV as they tend to make more repeat bookings).
- Secondly, it helps us to assess SEO performance: “car hire” keywords are much more competitive than the other types, so increases in conversions from those pages would indicate that our SEO is working really well.
I achieved this by using a custom dimension to extract the vehicle type from the URL of the landing page, and then using this to pivot the table.
The graph below is one of the most important in helping to communicate positive results with my key stakeholders. It shows the number of new member sign-ups per month in the period after completing the website overhaul.
To make these results targeted and meaningful, I created a filter that shows me only new members who reached our site through the recently created landing pages. We get a lot of traffic from our paid search efforts and through branded keywords, too — by excluding people who found us in one of those ways, I can more clearly show the impact of my SEO work.
Using Looker to monitor and enhance the performance of Google Ads campaigns
A few months into my role as SEO manager, the CEO asked if I would mind taking over our Google Ads account. With some hesitation (I had very little experience), I accepted the responsibility — and after some early mistakes (and wasting quite a bit of money — sorry, CEO), we have ultimately been able to reduce the cost per first-time borrower from over $100 to less than $30.
There were two main projects that led to the improvements in results:
- Restructuring the account, since it had become completely disorganised over the years
- Implementing smart bidding strategies to optimise for conversions
Essentially, we needed simplification.
Looker played an important role in the transition, particularly with regards to the use of smart-bidding strategies. By using Looker dashboards to visualise metrics like CPA and ROI, it was possible to demonstrate the effectiveness of automated bidding on bottom line metrics.
Before, getting the full picture from our campaign and revenue data was time-consuming and prone to human error. I used to extract spend data from the Google Ads platform and manually match it up to revenue data that I exported to Google Sheets. Now, we use the Google Ads Analytics Block to pull SEM spend and campaign data into Looker. This has been a real game changer. I have a series of dashboards that allow me to keep track of spend, CPAs and ROI all within a few clicks. It saves so much time when Iproduce reports for the leadership team or investors, and when I have regular catch ups with my manager.
As shown in the graph below, the improvements to CPA have come at the expense of volume (we cut back budgets significantly when COVID-19 hit — see the huge drop in the green line on the graph below, around June), but the Looker reports are making it easy for me to gradually increase budgets again whilst keeping a close eye on CPAs to ensure that ROI doesn’t take a hit. I check the graph below each week and make sure that applicants are trending upwards whilst CPAs stay fairly consistent. If it’s all looking good, I increase the budget by a small amount and let the smart bidding strategy do its work to maximise conversions.
Importing offline marketing spend data into Looker
Another internal challenge I am trying to overcome is a tendency for some decision makers to focus on efficiency metrics like ROI at the expense of effectiveness metrics like reach. This problem is exacerbated by a general suspicion of brand advertising, and a tendency to favour performance campaigns which deliver a return that is more easily measurable.
One way I am overcoming this is to steer internal reporting towards higher-level metrics (such as the “overall PA, calculated by dividing the total marketing budget by the number of new users), and away from channel-by-channel analysis, which always favours activity concentrated at the lower end of the funnel.
To do this, we have recently used Stitch to import spend data for offline channels and those we can’t link into Looker directly (such as outdoor advertising, radio, PR, SEO agency fees, copywriters, and so on). We achieved this by inputting spend data each month into a Google Sheet and importing this into Looker with Stitch. Whilst this still involves some manual input, the ability to access this data in Looker and create high level reports and dashboards is a meaningful addition to our overall understanding of our marketing efforts.
What do we have planned next?
I’ve been learning more about how to use Python for advanced data analysis and machine learning over the past year, using the Looker API SDK to import Looks directly into Google Colab or Jupyter Notebook. It really speeds up the workflow, especially when working with very large datasets.
We’re just getting started with our predictive analytics, but here are some projects we’re excited to explore:
- Building a model that can predict if a first time borrower is likely to take a subsequent trip (we’ll use this to improve our Adwords bidding and increase ROI)
- Analysing how the lifetime value of different cohorts of borrowers has changed over time
- Finding out if COVID-19 has led to more people taking trips in the middle of the day
Multi touch attribution
All our marketing reporting is currently done based on a last click attribution model. This isn’t ideal as it significantly undervalues the role that some channels (notably SEO and SEM) play in helping people to discover our brand. Fortunately, we are already capturing all user interactions with the website using Segment, a customer data platform.
I’ve recently discovered a Looker Block that will allow us to surface the attribution data within Looker. I’m just waiting for our developers to set it up. This will be another huge leap forward for us as a company, helping us to get a more accurate picture of our users’ conversion paths and making better marketing decisions as a result.
Overall, it’s fair to say that I’m a big Looker fan! I have to give credit to our CTO, Dave, for choosing such a great platform and investing the time into getting all our data flowing into it. Having data so readily available creates endless possibilities to uncover insights, improve performance, and ultimately expand our skillsets as marketers.
If you’d like to learn more about how other teams at Car Next Door are using Looker to optimize and enhance our customer experience, you can read our case study here.