The great user acquisition debate: short-term conversion events vs. long-term LTV optimization
When it comes to running online (or mobile) ad campaigns, the current status quo is almost always short-term optimization. Growth marketers often optimize campaigns based on signals/events that take place soon after the initial engagement with the ad. It could be a signup event, a purchase event, a content view event or Day7 login, but some events are tracked in the short term.
The big problem, as we all know, is that LTV (or profitability) is almost always an afterthought, even though it should be the main focus.
SaaS subscriptions, freemium, repeat customers or high-involvement purchases are just a few examples of business models that mostly don’t show an immediate ROAS.
Not that there was much choice in that matter (at least not until now). Current practices of user acquisition dictate that growth marketers focus on super fast optimization in just a few short days. Why? Because of technical and organizational reasons.
Media channels only offer optimization for short term events, and campaign go/no-go decisions are mostly focused on ROAS within the first 7 days from acquisition. While this may be a valuable effective strategy, it is suffering from the “streetlight effect”. Which is why it limits scale.Only several short term events are sent to the media channels from the limited client side, and that’s part of the problem.
Besides, marketers are under constant pressure to show quick wins and justify their spend before large budgets are spent; organizations don’t usually have the patience, and foresight needed for long-term measurements.
Marketer’s focus on short-term ROAS is most likely related to the media channels’ need to quickly produce data that can feed their hungry Machine Learning (ML) engines with conversion data.
But no matter why things evolved that way, we can agree that we are blinding ourselves to the existence of huge profits just around the corner in our rush to generate quick proof of ROI.
Early Birds vs. Late Bloomers
Those who have the potential to optimize for purchase events (as they have enough of them in the conversion window) are lucky that they are at all able to do that — as many products simply do not facilitate enough frequency. Even though they focus on Early Birds, they have a huge potential to scale up.
Add to that the fact that we are all optimizing for the same audiences — those who tend to engage in the same short-term events — so our bid prices are constantly rising, while we’re not necessarily reaping the rewards in the form of profitable, scalable acquired audiences. Houston, we have a problem.
The recent ‘Not Another State of Marketing Report‘ by HubSpot shows that over 60% of marketers surveyed say that their customer acquisition costs have increased in the past 3 years. Why am I not surprised?
What this means is that the cost of acquiring those “Early Birds” — those who convert quickly and generate conversion data for ML engines and network conversion windows — is potentially much higher than the cost of targeting and acquiring “Late Bloomers” — those who may take their time but are much more profitable in the long run.
Focusing on LTV or profitability measurements that often take time to manifest makes much more sense for optimization purposes. In other words, if we could accurately predict the user’s true ROAS we wouldn’t be worried about money not coming in immediately, and we would double down on acquiring similar users.
But we need something that would speed up the process of manifesting the future, because growth marketers can’t take such risks with no prior visibility, and the network optimization engines run out of patience.
Users who don’t take immediate actions represent entire audiences currently excluded by advertisers. The natural behaviour of users is that they explore the product, try it out over an extended period, build interest and engage with it on different levels before committing to buying a premium or signing up for a subscription.
How do you make your purchasing decisions? Do you usually check out a product or service or do you immediately buy it?
Let’s take a closer through one specific example:
Freemium mobile games and apps: Industry leaders know what to look for
We’ve all experienced freemium. A user is given the game or app for free with the option of making purchases to enhance their experience with access to additional features or power-ups.
Let’s focus on games, as they are the industry leaders.
Mobile (casual) games have higher retention and users that tend to pay over longer periods than eCommerce. They therefore tend to optimize towards 1 or 7 day ROAS, but with a target that is much lower (1–15%); they assume that users who purchase at the beginning will continue to do so beyond that conversion window, with some drop-off. They sometimes have an ad monetization strategy on top of that — but that is not too common these days.
A lot of the companies use a 7 days conversion window and set purchase, or revenue, as their user acquisition optimization signal for Facebook or Google (“App Events Optimization on Purchase”).
Unfortunately, this short-term optimization goal is a profoundly askew measurement tool in this vertical. Most gamers usually need some time before they buy into the game mechanics and reach a point where they want to accelerate their experience by purchasing level-ups/credits/energy. This can often take much more than just seven days, meaning that the 7-day Early Bird optimization falls short at identifying (the entirety of) the desirable audience.
While it is true that Early Birds are more likely to purchase later on, it is also true that users that are highly engaged, play many sessions, invite friends, or signup with their facebook account are ALSO more likely to purchase later on.
There are actually a plethora of signs indicating long-term purchasing users, and the trick is leveraging all of them. Potentially, one could use historical data to model the probable LTV of each user, based on *all* of their product engagement data. This means that there’s a way to use data and create a short-term event that is actually an LTV score prediction. And if you could send this score to the network for optimization purposes, that would change the entire picture. While this is a gaming example — the principle holds for any business model and vertical that cultivates long-term relationships with customers and expects to provide value over time.
But how is that possible? Let’s take a look…
Finding the right balance — Optimizing UA by LTV predictions
Giant brands are already moving away from optimizing for short-term ROI. They’ve realized that letting media platforms optimize on the explicit revenue signal will not maximize profits at scale. They therefore model LTV internally and make keep or kill campaign (or ad) decisions accordingly. They can do that because they use their internal BI, LTV predictive models based on their internal data lakes, and attribution data for each new campaign cohort. While they couldn’t (until recently) send media server side LTV predictions for optimization (now, with Facebook conversion API and Google Server Side Tagging, they can), they could make keep or kill decisions offline.
Today, the most advanced companies switch to user-level prediction (which is much harder than cohort level predictions) and send them to the networks’ server-side APIs.
But what about all of us, non-giants? Most media buyers that spend millions on UA, but do not possess advanced in-house data science capabilities, therefore are unable to reap the huge scale and ROAS potential associated with the Late Bloomers.
Building and maintaining such user-level models is super hard:
- A “continuous” real-time value model is needed, constantly checking the prediction as the user advances within the product experience.
- Models need to be monitored, retrained, and analyzed frequently.
- Optimizing towards a network’s optimization engine is very much different from today’s cohort-level analysis best practices. It’s a whole new area of algorithmic expertise.
Well, it’s changing.
New AI-driven technologies are empowering end-to-end plug and play solutions that can accurately predict long-term profitability using third-party data alongside historical internal data.
Brands can now leverage these solutions to not only identify users that are more profitable over time, but to optimize for new, similar users.
They can then leverage the media platforms’ new server-to-server APIs to send them signals that represent LTV predictions (or profitability or virality or anything similar, for that matter), in order to optimize their campaigns and this time (for a change) target users that have the biggest chance of contributing the most to the business, thereby enabling segmentation down to the individual level.
There may be a few interesting solutions out there, but I can only speak of mine. Voyantis’ end-to-end, zero coding solution leverages AI, data science and a few other goodies to let growth managers and UA teams effectively target users based on long-term LTV predictions. With Voyantis’ Signal Optimization solution, marketers can target much greater-potential users, probably at a significantly lower price than the Early Birds who may or may not offer LTV profitability down the road, without them needing to beg for R&D resources.
It’s not an either/or decision.There’s a whole continuum between “target Early Birds” and “target Late Bloomers”. You should probably consider targeting any user on the profitability spectrum, depending on multiple interests that any business may have. You can go for either of those or choose a ‘mix and match’ strategy. This allows growth marketers to find the right balance. If I may use an analogy, you wouldn’t invest in bonds while ignoring the stock market, right? One should treat their holistic UA strategy as managing a portfolio of stocks, each with their own expected risk (how quickly can we observe success) and reward (potential ROAS and scaleup) — which needs diversifying, and restructuring as your strategic company goals shift.
For years, media platforms’ algorithms offer immediate rewards and satisfaction for Early Birds strategies — and that was the low-hanging fruit everybody went for. This means that fast-converting audiences have become significantly more in demand and hence much more expensive to acquire If you want to scale up, you need to step up your game.
This game was rigged to all but the few iconic brands; they were the only ones that invested what it takes to apply predictive models according to historical data and long-term metrics that enable them to identify high potential users.
The rest of the players had no choice but to use the metrics readily available for optimization, but now things are shifting rapidly.
New AI-driven solutions are disrupting this status quo by enabling UA teams to optimize their users for long-term profitability without needing significant BI, data expertise and R&D resources. Using such new solutions, growth marketers can manage and diversify their strategies portfolio on a sliding scale between short-term ROI and higher LTV risk-reward.
I welcome your comments. Feel free to reach out if you have any.