Enterprise Android App Campaign ROI for MY Retailers

The Firebase-GA4 First-Open Baseline

Malaysia’s Android share hovers around 70% of the active smartphone fleet, so the install event is near-worthless unless you timestamp it. Google Ads reports an install the moment Play Services confirms it, but it doesn’t tell you the device’s quality or whether it came from a pre-warmed app ID. The fix is a match table: left side `install_referrer` fields (`utm_source`, `utm_campaign`, `install_begin_timestamp`), right side `firebase_first_open` timestamp from the BigQuery export. When the gap between `install_begin_timestamp` and `first_open` exceeds 180 minutes, flag that record as a probable click-injection loop — a real problem in Malaysian programmatic traffic routed through low-tier ad exchanges from outside the Klang Valley.

Time-to-conversion from install to actual purchase also differs by network. Typical 4G users in crowded zones such as LRT Kelana Jaya or Pasar Seni delay buying until Wi-Fi at home — usually between 8 p.m. and 11 p.m. That pattern means a 7-day conversion window hides more than it reveals. Export `first_open` and `purchase` timestamps to BigQuery, compute the median time-to-purchase per campaign, and then shorten or lengthen the Google Ads conversion window accordingly. Get that wrong, and your tROAS bid adjustments get flooded with phantom purchases.

Klang Valley Cost Benchmarks: CPI, IPM, tCPA

Budget-setting for an Android App Campaign targeting Klang Valley today lands in these ranges:

– CPI (retail utility/loyalty app): RM 2.10 – RM 3.20

– CPI (e-commerce app): RM 3.10 – RM 4.40

– IPM: 12–24 with broad targeting; drops to 12–15 when restricted to a 10 km radius around Suria KLCC or The Exchange TRX

– tCPA for purchase (fashion): RM 14 – RM 18, on a RM 195 average basket

– tCPA for purchase (grocery): RM 9 – RM 12, on lower baskets with higher repeat rate

The most controllable leak is geography. If the target radius expands beyond the Klang Valley — especially into Johor Bahru or Pulau Pinang — IPM rises, but verified device age says those users are less likely to hold current Play Services. You can’t ban states easily anymore, but you can apply negative location exclusion at the district level in Google Ads. Also switch the conversion value to the GA4 `purchase` event and override with actual net revenue (GMV minus returns and payment gateway fees). Otherwise, the algorithm optimizes for revenue that later gets reversed in your ERP when refund cycles flush out returns after 30 days.

Payment Loops: FPX, GrabPay and E-Wallet MDR

The shopping funnel dies at the payment page, not the ad click. In Malaysian integrations, most FPX gateways launch a sandboxed web view. If your Android app does not declare Android App Links via `assetlinks.json` on the merchant domain, the user gets dropped back into Chrome after authenticating on their bank’s page, losing the in-app session. Across local retail tests, this single oversight removes 15–20% of legitimate checkout completions. That’s a campaign-ROI leak invisible to any attribution tool because the install already happened.

The cost side is equally concrete. E-wallet merchant discount rates in Malaysia — Touch ’n Go eWallet, GrabPay, Boost — sit roughly between 0.5% and 2.0%. FPX’s per-transaction fee is near flat RM 0.30–RM 1.60 depending on volume. For a RM 210 fashion basket, choosing TnG eWallet over FPX leaves RM 3–4 less net revenue for the same conversion value, which maps to a false ROAS improvement if your tROAS denominator uses gross revenue. Track payment rails as a separate dimension in the purchase event parameters, and deduct the MDR in the same SQL view that feeds your dashboard.

Also keep the e-invoice sequence in mind when LHDN MyInvois validation lags after a successful payment. Campaign in-app purchase events fire in real time, but the official invoice’s validity state may arrive hours later. Sync cost fields on `invoice_validated` rather than on payment success, otherwise your first-day ROAS includes orders that need resubmission.

PDPA-Safe Push and Web-to-App Re-Engagement

New installs are expensive; re-engaging the existing daily active user base is where the real ROI materializes. Under Malaysia’s PDPA, asking for notification permission on the first screen creates consent friction and lowers opt-in below 10%. Trigger the request after two completed `home_screen_view` events with a clear value statement (e.g., “Restock alert for your saved items”). This gets opt-in rates around 35–40% for grocery apps and 25–30% for fashion apps — credible numbers for the region.

For the re-engagement workflow, use Google App Campaigns with the “re-engagement” goal rather than installs. Add Web-to-App Link or a Branch/Adjust-generated link to pass a product ID directly into the app; the URL scheme must be tied to a single product detail page. Re-engaged users consistently produce first-week purchase rates three times higher than fresh installs, and the median order value spikes when you pair the deep-linked product with a free-delivery voucher for Klang Valley delivery zones. If voucher minimums are set below RM 49, you turn the entire campaign into a shipping-cost subsidy.

SKAdNetwork-Agnostic Bidding with GA4 Signals

Android’s advantage is full-visibility conversion feedback. No SKAdNetwork delays, no postback truncation. Use this to create conversion values at the item level: send item-level data from your retail app into GA4, then expose the historical revenue cohorts back to Google Ads through predictive LTV signals. With 90-day revenue and 30-day return rate from BigQuery, you can feed the algorithm precise customer quality. The model then shifts budget toward users who browse and checkout via FPX rather than those who live in high-return segments.

Running Android campaigns this way creates a clean operating loop: Play Install Referrer matching for install quality, BigQuery for purchase timing, payment-rail MDR subtraction in the event layer, and deep-linked re-engagement for cohort revival. Revise the tCPA every quarter against Malaysian shipping costs and e-wallet fee changes.