AI for UA: chatting with AppSamurai’s Faheem Saiyad
Where can AI help in mobile user acquisition? I recently chatted with Faheem Saiyad, a director at mobile advertising platform AppSamurai, about AI for UA.
The big question: where can AI for UA help marketers the most? Hit play to start the conversation, and keep scrolling …
First: building your app
Most UA managers won’t be helping to also build their apps and games unless you’re part of a very small team. But if you are an indie developer and that’s your current reality, AI can help.
“Back in the day designing a game was quite a complicated task,” Saiyad says. “With AI in place, what it has done has taken a lot of workload off. So, let’s say you are an indie game developer, back in the day, probably you might need a couple of freelancing outsourced people that will help you build a certain game out. But now with these resources from AI you can actually build out a certain project, do a quick soft launch, use A/B testing, even the creatives can be designed by an AI.”
I can certainly see building some of your creatives with generative AI tools like Creative Diffusion or MidJourney. And GPT-4 isn’t bad at coming up with new ideas for games that you can probably modify or riff on to improve.
AI for UA: data analysis and campaign optimization
You don’t have to be part of a small team to want extra help with data analysis and campaign optimization. BI analysts are always busy, and manually messing with CSV files, Excel spreadsheets, or Google Sheets is annoying and time-consuming.
Good news: AI is pretty good at data analysis.
“With AI what we have seen is a lot of well-run analysis that you can basically summarize, especially when you are at a soft launch stage,” says Saiyad. “So this basically helps you save a lot of time and money.”
One note: double-check a few of the calculations that your AI tool is making for you.
I have literally caught GPT-4 making some silly assumptions and mistakes. You can save a lot of time, but if you uncritically accept what a generative AI system has invented, you can also end up with some egg on your face.
Of course, sometimes the error is yours: you just didn’t prompt the generative system appropriately. Kinda like that one time I wanted a CSV file but accidentally asked for a CSS file …
There are also AI-driven tools that will find, name, and segment audiences. In some cases they’ll identify actions you can take or offers you can make to maximize conversions and spend; in other cases they’ll even take that action for you.
AI for UA: campaign creation and launching
Using AI for UA in ideation and creative development is one thing. And using AI for UA purposes in data analysis for campaign optimization is another. But using AI to actually design and launch ad campaigns autonomously within certain parameters is pushing the bar significantly higher.
“In a realm where you have an advertiser that looks to automate a certain campaign creation process, our APIs are able to plug into their backend platforms,” Saiyad says. “The idea here is that they can automatically launch campaigns, they can set bids … the optimization — although it’s not perfect — we are still getting real close.”
Right now, that’s probably a bit like Tesla’s Full Self Driving: totally fine for some who trust it and use it and swear by it, while interesting but dangerous for others who see the potential but are wary of trusting their lives to it.
Which Saiyad totally understands, adding that AI for marketing is not quite set and forget it at this point: you generally want a human in the loop.
One thing we know for sure, however, is that while AI is demonstrably imperfect, it’s also improving at breakneck speeds. So betting against automated campaigns is probably not wise, at least in the long term. (And let’s be honest: is it really that different from giving Meta or Google a stack of cash and having their algorithms figure it out?)
Not for fraud detection just yet
While AI for UA managers kinda looks perfect for fraud detection, since it should be exactly what you need to munch through mounds of data, find patterns, and either identify those correlated with poor down-funnel performance or those which are associated with shady or duplicitous means of taking credit for good ROAS … that’s not the case.
It’s not quite there yet, Saiyad says.
“We still do have manual human resources that actually keep a check on AI, especially in terms of ad fraud,” Saiyad says. “A lot of ad fraud has been going on, and as of yet, at least in my experience, I haven’t seen AI being used in the realm of preventing ad fraud at scale.”
AI for UA: much more in the full episode
If you haven’t subscribed to Growth Masterminds, now is a good time. You can watch it on YouTube or listen on your favorite podcasting platform.
In the full episode, we talk about:
- 00:00 Introduction and welcome
- 01:19 The role of AI in game development
- 03:35 The Impact of AI on user acquisition
- 06:22 Challenges of campaign optimization
- 10:25 Introducing rewarded playtime: a new ad unit
- 13:44 The future of AI in marketing and game development
- 16:38 Wrapping up the discussion
And, as usual, if Singular can help with anything, connect with us today.