Mobile user acquisition has quietly become a harder discipline than it was three years ago. Signal loss has made attribution probabilistic in places it used to be deterministic, competition for the same inventory has pushed rates up, and the easy wins from broad targeting are gone. What replaced them is less glamorous and more durable: better measurement, tighter funnels, and channel mixes chosen deliberately rather than by default.
This is a practical map of what works now, and what each channel is actually good for.
Start with the event, not the channel
The most common failure in UA planning is picking channels first. Channels are delivery mechanisms; the strategy is the event you optimise toward. Before any budget moves, you need one number: the earliest in-app event that reliably predicts a user's value. Everything else — bids, creative, source selection, which partners you even talk to — is downstream of that decision.
If you cannot name that event, the honest first project is instrumentation, not acquisition. Buying users you cannot evaluate is how budgets disappear without anyone being able to explain where.
Programmatic and DSPs: reach, at the cost of transparency
Programmatic buying gives you scale and machine-speed optimisation across enormous inventory pools. For apps with clean measurement and a healthy budget, it remains the fastest path to volume in competitive categories.
The trade-off is visibility. Algorithmic optimisation is only as good as the signal you feed it, and a DSP optimising toward installs will happily find you the cheapest installs in the world. Feed it your post-install event instead, insist on placement-level reporting, and treat any partner who will not break out sub-publisher data as a partner you cannot evaluate.
Affiliate and CPA networks: risk-shifted volume
Working with a performance network moves the risk of non-converting users onto the supply side, which changes the incentives in your favour. Instead of bidding for impressions and hoping, you define the event and pay when it happens.
This is where a well-run network earns its position. Across AppGro's 120+ vetted publishers and 65+ active brands, the value is not access to inventory — anyone can buy inventory. It is the vetting layer, the fraud filtering that runs before you are ever billed, and the sub-publisher transparency that lets you tell a good source from a lucky one.
The failure mode to watch for is a network that will not show you what sits underneath a campaign. Blended reporting is where poor traffic hides.
OEM and alternative app stores: underrated in emerging markets
Preload and on-device placements through handset manufacturers have become a serious channel in markets where a large share of users discover apps outside the mainstream stores. Volume is high, the traffic is genuinely human, and competition is lighter than on mainstream inventory.
Retention curves differ from store-driven installs, so it needs to be measured as its own cohort rather than blended into the campaign average. Judged on its own terms, particularly across South and Southeast Asia, it frequently outperforms the channels that get far more attention.
Influencer-driven installs: growing up
Influencer UA has moved from brand experiment to measurable performance channel, largely because tracking finally caught up. Where it works, it works because the recommendation carries trust that no ad unit reproduces — which shows up as unusually strong retention rather than unusually cheap installs.
Treat it as a performance channel with a longer feedback loop: individual creator deals behave like small campaigns, and the winners deserve renewal budgets rather than one-off fees. Judge on cohort quality at day 7 and day 30, not on install cost in week one.
Retention is an acquisition strategy
The cheapest user is the one you already paid for. In a market where acquisition costs rise every year, the highest-leverage UA work is often not in the ad account at all — it is in the first session, the onboarding flow, and the point where new users decide whether the app is worth a second open.
A ten percent improvement in day-7 retention changes what you can afford to bid. No amount of creative testing does that.
Practically: fix the drop-off between install and first meaningful action before you increase spend. Improving that ratio raises the ceiling on every channel simultaneously.
Fraud: still the largest silent cost
Invalid traffic remains the biggest unmanaged line item in most UA budgets, and it is rarely obvious, because sophisticated fraud is designed to resemble good performance. Click injection, install farms, SDK spoofing and attribution hijacking all present as installs that fail quietly further down the funnel.
Defence is layered, not singular: device fingerprinting, click-to-install time analysis, geo and carrier consistency checks, and behavioural validation on the first session. Run them before billing rather than as a monthly reconciliation exercise, and cut offending sources rather than negotiating with them.
Building a mix that holds up
A durable 2025 mix tends to look like this: a scalable core channel that carries volume, one or two performance-priced channels where risk sits on the supply side, a geo-specific channel that exploits a market advantage others ignore, and a permanent testing allocation of roughly ten to fifteen percent for whatever is next.
Review it on a weekly cadence at the source level, not a monthly one at the campaign level. Campaign-level monthly review is how a bad source runs for four weeks before anyone notices.
None of this is exotic. The teams doing well in mobile UA right now are not the ones with a secret channel — they are the ones measuring the right event, refusing to buy traffic they cannot see inside, and treating retention as part of the acquisition budget rather than someone else's problem.