Constructing an Android App: Why the Proper Growth Associate Issues

As of October 2026, 43.6% of energetic Android units nonetheless run Android 13 or older, whereas Android 16, now the most typical model, sits at 25.4% and Android 17 has crawled to 1% since its June launch. Even Android 8.1 nonetheless has a much bigger slice than the latest launch. That unfold is the job. On Android, launch day is the simple half, and the companion you select exhibits its high quality in month six, not week one.

I’ve shipped and maintained Android apps lengthy sufficient to cease judging builds by their first week within the Play Retailer. Week one seems nice virtually each time. The set up curve climbs, the screenshots look sharp, the founder posts a launch thread on LinkedIn. Then the true customers arrive, on telephones no person on the company owned, operating OEM skins no person examined, and the app begins amassing one-star critiques that every one say some model of “it simply stopped working.”

So I take advantage of a easy filter now. I name it the month-six check: ignore the demo, ignore the launch, and ask what the app seems like half a 12 months later. By then it has lived by way of at the least one Android launch, one goal API deadline, a few Play coverage emails and some million background duties killed by battery savers. A weak companion passes week one and fails that. An excellent one deliberate for it earlier than anybody drew the primary display.

Porting the iOS app is how Android initiatives get into bother

“An Android app is simply the iOS app ported over.”

You hear some model of this in almost each scoping name, normally from somebody attempting to chop the estimate in half. It’s half true for the UI layer, and that’s the damaging half. Shared enterprise logic by way of Kotlin Multiplatform or Flutter can save actual cash, and I’ve advisable each. The issue is all the pieces beneath the UI, which is the place Android apps truly dwell or die, and the place the estimate quietly assumes nothing will occur.

Apple ships its OS to some dozen machine fashions from one producer, and most of them replace inside months. Android ships to 1000’s of fashions from Samsung, Xiaomi, Oppo, vivo, Motorola, Google and an extended tail of manufacturers you’ve by no means heard of, every with its personal pores and skin, its personal battery supervisor and its personal opinion about what “background” means. Notifications. Background sync. Runtime permissions. Predictive again. Foldable hinges and split-screen. None of that ports. It must be designed for, after which examined on {hardware} that behaves badly. An iPhone construct that passes QA on three units has coated a significant share of its viewers. An Android construct that passes QA on three units has coated three units.

That’s why I don’t begin companion evaluations with the portfolio. Portfolios present launches. After I overview proposals, from solo freelancers and from corporations that promote full Android app improvement providers alike, I ask one query first: what broke in your final Android app after launch, and the way did you discover out? A workforce that has maintained apps for years has a selected, barely embarrassing reply prepared. A workforce that solely builds them talks concerning the launch social gathering.

Right here’s what the primary 12 months truly seems like, month by month.

Month 1: Android vitals grades you on telephones your companion by no means touched

The primary month is when Play Console begins filling Android vitals with actual knowledge, and Google doesn’t grade on a curve. The printed bad-behavior thresholds are a 1.09% user-perceived crash price and a 0.47% user-perceived ANR price throughout all units, plus 8% for both metric on any single cellphone mannequin. Cross them and Google Play could cut back the app’s visibility and present customers a warning on the shop itemizing. Play judges the trailing 28 days, so one unhealthy launch lingers for a month.

The per-model threshold is the one which catches weak companions. An app can look wholesome total and nonetheless be falling over on one common price range Galaxy or a Redmi operating HyperOS, as a result of the company examined on two Pixels and the founder’s personal cellphone. I’ve watched precisely this play out. The general crash price was clear. The ANRs clustered on one chipset household, and the trigger turned out to be a database learn on the primary thread that solely received sluggish on cheaper storage. No person on the workforce owned a cellphone with that storage.

An excellent companion dealt with month one earlier than launch. They pulled the Play Console machine catalog to see which fashions the goal market actually makes use of. They ran the pre-launch report and Firebase Take a look at Lab throughout a diffusion of API ranges, not simply the newest two. They turned on StrictMode in debug builds, so disk and community calls on the primary thread fail loudly in improvement as a substitute of quietly in manufacturing. And so they wired up crash reporting with mapping information uploaded, so a stack hint from a Xiaomi in Jakarta factors to an actual line of code as a substitute of obfuscated noise.

The weak companion finds out about ANRs from the critiques.

Month 3: OEM battery savers are quietly killing your background work

By month three, the assist inbox has a theme. Reminders arrive late. Step counts reset in a single day. Chat messages seem solely when the app is opened. Customers blame the app, and actually they’re proper to, though the trigger lives within the cellphone’s firmware.

Don’t kill my app, the community-run tracker for this precise downside, places it bluntly: many producers “want battery life over correct performance of your apps.” Its rating offers Huawei, Xiaomi, OnePlus and Samsung the worst attainable rating, with Meizu, Asus and Oppo one notch higher. Pixel and Android One units rating zero. Your customers don’t purchase Pixels within the proportions your builders do.

A weak companion writes background code that works on inventory Android, ships it, and payments the repair as a change request in month three. An excellent companion treats OEM conduct as a requirement from day one. Which means WorkManager for deferrable jobs, high-priority FCM messages just for issues the consumer is definitely ready on, foreground providers declared with the fitting foregroundServiceType (necessary for apps focusing on Android 14 and up), and an onboarding display that walks Samsung and Xiaomi homeowners by way of the battery exemption with brand-specific steps. That final display seems unglamorous in a design overview. It additionally saves a assist particular person chunk of each week.

Not a bug. Not a crash. A silent failure, which is worse, as a result of vitals gained’t flag it and your dashboards will say all the pieces is ok.

Month 6: the goal API deadline finds each shortcut

Month six is the place the check earns its identify, as a result of for lots of apps it overlaps with Google Play’s yearly goal API bump. This 12 months’s guidelines sit on Google’s goal API degree necessities web page, and so they’re stricter than most founders understand.

The dated numbers price taping to a monitor:

• Since August 31, 2026, new apps and app updates should goal Android 16 (API degree 36) or greater to be submitted to Google Play.
• Present apps should goal at the least API degree 35 to remain accessible to new customers on units operating a more moderen Android model than the app targets.
• November 1, 2026 was the newest an authorised extension may push the deadline, so there is no such thing as a extra room to attend.

Elevating targetSdkVersion isn’t simply altering a quantity in Gradle. Every degree adjustments platform conduct to your app, and API 36 carries an enormous one for big screens. In accordance with Google’s adaptive apps steerage, on any show with a smallest width of 600dp or extra, which covers tablets and the interior screens of book-style foldables, Android 16 ignores an app’s portrait lock, its resizeableActivity flag and its side ratio limits. There’s a manifest opt-out for now. The identical steerage says it disappears at API degree 37.

Image the portrait-only app an affordable companion constructed. It regarded nice on telephones. After the goal bump, it opens on a Galaxy Z Fold’s interior show stretched throughout a virtually sq. display, with a backside sheet floating within the center and a digital camera preview rotated the unsuitable manner. No person wrote new code. The platform simply stopped masking for the outdated shortcut.

An excellent companion constructed with window dimension courses and adaptive layouts from the beginning, so the bump prices every week of testing as a substitute of a redesign. Additionally they saved dependencies present by way of the 12 months. That issues greater than individuals anticipate, since an advert SDK or funds library pinned to a two-year-old model is usually the factor that blocks the improve, and you’ll’t ship your repair till the seller ships theirs.

Month 12: coverage emails and possession gaps present up collectively

On the one-year mark, the issues cease being technical and begin being administrative, which is someway worse. Play coverage deadlines maintain arriving on their very own schedule. On September 30, 2026, Android developer verification started rolling out on licensed units for apps from taking part shops in Brazil, Indonesia, Singapore and Thailand, and Play Console began asking builders to register any apps that weren’t registered robotically. On January 27, 2027, Play’s SMS and Name Log coverage stops accepting phone-call account verification as a cause to request READ_CALL_LOG. Every change is small. Each wants somebody who reads the e-mail, understands what it means for this particular codebase, and ships earlier than the date.

Then there’s possession. I’ve helped founders get well apps the place the company held the add key, registered the Play Console account beneath its personal identify and saved the Firebase venture inside its personal group. No person acted in unhealthy religion. It’s simply what occurs when no person writes the setup down. When the contract ended, the founder owned a retailer itemizing they couldn’t replace.

An excellent companion arrange Play App Signing on day one, created each account beneath the consumer’s group, documented the discharge pipeline and left a runbook a brand new developer can observe on a Monday morning. Additionally they plan the subsequent bump earlier than anybody asks. Google has raised the goal requirement yearly thus far, so I anticipate API 37 to turn into necessary for updates someday in 2027, which might additionally finish the large-screen opt-out for good. That final half is my inference, not Google’s printed schedule.

The distinction between a companion and a vendor suits in a single desk

Discover that not one row asks concerning the launch.

Rating your shortlist earlier than you signal something

Use this as a 0 to 2 scorecard for every candidate, 14 factors in complete. Something beneath 10 means you’ll find yourself paying for month six twice.

1. They reply the post-launch query with specifics. Rating 2 in the event that they identify an actual bug, the machine it hit and the way they caught it. Rating 0 for “we check completely.”
2. Their machine matrix comes from knowledge. Full marks if it was constructed from the Play Console machine catalog or analytics to your market. Zero if it’s no matter telephones are on the workplace desk.
3. They know the vitals thresholds chilly. A companion who quotes 1.09% and 0.47% from reminiscence has been burned by them at the least as soon as. Give 1 level in the event that they know vitals exist however not the numbers.
4. They’ve shipped a goal API bump lately. Ask concerning the transfer to API 36 and what it broke. A transparent story about particular conduct adjustments scores 2.
5. Background work comes with an OEM plan. Rating 2 if they create up dontkillmyapp.com or brand-specific battery settings with out prompting.
6. Massive screens are within the estimate. Foldables and tablets ought to seem as line gadgets with hours connected, not as “part two.”
7. You personal all the pieces from day one. Play Console account, add key, Firebase venture, repository, CI. Something held of their identify scores 0, irrespective of how pleasant the gross sales name was.

Launch day tells you who can construct an Android app, and month six tells you who can maintain one alive.