Last week we hosted our inaugural Serpico event at Google’s HQ in London, where we explored some of the opportunities offered by Google Marketing Platform, as well as broader challenges and opportunities for digital marketers.
As part of the event, Croud’s very own Kevin Joyner, Director of Planning and Insight, explored the importance of lifetime value tracking (LTV), and some approaches that marketers can adopt. Here is a round-up of his talk.
Why LTV is important
Kevin started out by referring to one of his favourite writers and marketing experts, Seth Godin, who hosts a podcast called Akimbo. Godin promotes the theory that, whilst marketing might have loftier ambitions later, it should first attempt to connect with the smallest possible relevant audience – the people who really connect as early adopters with your product. As such, he places emphasis on the quality of customer experience, as well as measuring LTV.
When talking about LTV, Godin says, “Torrents are made of drips.” Kevin gave Godin’s examples of this:
- If an Apple upgrade breaks your phone and you switch to Android, it costs Apple more than $10,000.
- If you switch supermarkets because a clerk was snide to you, it removes $50,000 from the store’s ongoing revenue.
These examples focus on the potential impact on churn, on the loss of customer value. But to think of it in a positive way: If you connect with the perfect customer, and treat them right, they’re worth (let’s say) 10x as much as the others.
All of this demonstrates that marketing is a long-term game, and that not all acquisitions are equal. As such, cost-per-acquisition misrepresents return on investment. And that’s why LTV is vitally important.
It’s also about an opportunity to invest in first-party data – the data that describes your unique relationship with your customers – which is an asset; unlike third-party data, which is the same commodity that your competitors can buy.
Getting started with LTV
First things first, it’s important to establish the functional requirement for an LTV data technology project, which essentially fall into three buckets:
- Analysis: of marketing channels, products and pages.
- Audiences: either remarketing or lookalike audiences.
- Bidding: e.g. automated bidding strategies, or the Smart Bidding feature in Google Ads, for search conversation optimisation.
And when it comes to methodology, there are a couple of key options – about how you calculate LTV (horizontal axis below), and how you record it (vertical axis).
Marketers essentially have a choice between predicted and actual LTV. Predicted LTV is based on acquisition signals, with the benefit of being quicky available, as you can choose when to calculate it after the initial customer acquisition. However, with predicted LTV, it is of course only an estimate and can be more tricky to set up from a technical perspective.
Actual LTV, on the other hand, is simpler to set up, as it is essentially a cumulative record of customer behaviour, which means it’s also much more precise. However, this approach means you’re limited by the pace of repeat purchase; it accumulates only as fast as repeat purchase happens.
With predicted LTV, you can choose to predict a specific metric or choose to bucket users into groups of high, medium and low, for instance. Predicting LTV is a process of revising a model over time, meaning it will get progressively better. In the example below, showing two iterations of a confusion matrix for LTV, the green cells show the correct predictions; the white cells show error.
When it comes to actually recording LTV, the two key options are:
- Conversion data integration
- Google Analytics (GA) data import
- Capture the gclid for the last session before the first customer transaction.
- Deliver your LTV data to the ad platform (e.g. Search Ads 360) using the offline conversion API
- Send the “conversion” you record post-dated so that it coincides with the time of the customer transaction, and match it against the gclid.
When it comes to the GA data import approach, you’re adding a new data table to Google Analytics, to extend the data that’s already tracked in. It’s great for analysis, and audiences, but it’s not conversion data, so it’s not good for bidding.
GA360 is even better, because Query Time data import allows you to extend historical data, meaning you can analyse going back in time, and it opens up the possibility of customer data being important against click or event IDs, instead of user IDs.
So there you have it – a whistle-stop tour of how to get started with lifetime value tracking. In summary, it’s best to use a couple of combinations: predicted LTV with conversion data integration for Bidding; an actual LTC with GA data import for Analysis. All combinations outlined above will work for Audiences.
If you’d like to find out more about lifetime value tracking, or have any questions on getting everything set up for your business, contact us.