What an app adds that a mobile website does not
The mobile marketing pillar treats the app as the centre of the discipline because of what an installed app unlocks. Four additions matter for the decision.
An owned messaging channel. Push notifications and in-app messages reach installed users without paying a platform for each contact. The push notifications and in-app messaging guide is clear about the condition: push depends on an operating-system permission the person can refuse or withdraw, and the share who grant it (the push opt-in rateShare of app users shown the notification permission prompt who allowed notifications.) caps everything that follows. The same guide notes that Apple's App Store Review Guidelines say push must not be required for an app to function and must not be used for marketing unless the person has opted in. The channel is owned, but it has to be earned person by person.
A place on the home screen. An installed app sits where the customer sees it every time they unlock the phone, and it opens without typing an address or searching. For a service used weekly or daily, that saved step is the whole point. For a service used twice a year, it is an icon that gets deleted during the next tidy-up.
A second discovery route. The stores are search engines in their own right. The app store optimisation guide explains that people find apps by typing what they want into store search, and that a listing's words, history and ratings decide where it appears. An app can be found by people who would never have searched the web for the business. That route only helps if people search the stores for the kind of thing the business offers.
The device itself. Apps can work with the camera, the phone's sensors, stored data and the operating system's own features more directly than a web page usually can, and they can be designed to keep working with a poor connection. Whether the business's customers need any of that is a question about the job, covered below, not a reason in itself.
What an app costs to run, beyond building it
Build cost is the figure everyone asks for and the one this guide leaves to the business's own quotes. The running costs are the ones that decide whether the app pays back, and they are mostly marketing and measurement work.
| Cost | What it involves | Where the site covers it |
|---|---|---|
| Two store listings | A listing on the App Store and on Google Play, each with its own fields, screenshots, ratings and review rules; the store controls the page and its guidelines bound what may be said | App store optimisation |
| Releases | Every change, including every tracking change, ships inside a release that goes through store review and then waits for people to update | Measurement plan for app installs |
| An install step | Every ad, email or search result has to persuade the person to install before they reach the product, where a web link would have opened it | App installs objective |
| Earning permission | The messaging channel exists only for people who grant notification permission | Push and in-app messaging |
| Attribution | The store breaks the link between ad and app, and privacy rules limit how installs are matched to campaigns | App attribution |
| Two audiences to serve | The mobile website still serves everyone without the app, so it cannot be abandoned | Mobile conversion troubleshooting |
The attribution cost surprises most first-time app owners. On the web a tagged link carries the source into the session. For an app, the app attribution guide explains why a measurement partner is usually needed to decide which campaign earned an install, why iOS reports arrive delayed and aggregated, and why some of the paid results will be modelled rather than counted. A business that runs paid acquisition into an app takes on that measurement work for as long as it runs campaigns.
The last row is the one most often forgotten. An app does not replace the mobile website. People who arrive from search, from a shared link or from an ad for the first time will still land on the web, and the mobile conversion rate troubleshooting guide is still the place to start when they do not buy.
Jobs that justify an app and jobs that do not
The useful test is the customer's job, not the business's wish for a channel. An app tends to be justified when the job is repeated often, when the person benefits from something kept on the phone, or when the job needs the device.
| Customer job | Usually suits | Why |
|---|---|---|
| Reordering the same things regularly | App | Saved baskets, a signed-in account and a home-screen icon remove steps each time |
| Tracking something over time (a habit, a balance, a delivery) | App | The value is in returning, and a timely notification is part of the service |
| Using the phone's camera, location or offline storage as part of the task | App | The device is part of the job |
| A one-off or occasional purchase, often as a gift | Mobile website | The customer will not keep an app for a purchase they make rarely |
| Finding an answer, a price or opening hours | Mobile website | The visit starts in web search and ends in seconds |
| Booking or buying for the first time | Mobile website | An install step stands between a new customer and the first order |
| Reading content shared by others | Mobile website, with links into the app for people who have it | Shared links must work for everyone |
Two warnings apply. First, the business's most loyal customers are rarely typical, so asking them whether they would use an app overstates demand. Look at behaviour instead: how often customers return, what share of revenue comes from repeat buyers, and which tasks people do on the website again and again. Second, an app that exists mainly to send promotions is the case the push guide warns against. If the only reason to build it is a channel for offers, the review guidelines and the opt-in requirement make that a weak reason, and an SMS or push comparison may show that a consented text programme reaches the same customers with no install at all.
The web-first route with an app for repeat customers
Many businesses do not choose one or the other. They make the mobile website good for everyone and add an app for the customers who come back often enough to want one.
The website comes first because it serves every visitor. Google's page experience guidance, as the site's page experience guide explains, asks whether pages display well on mobile devices and avoid intrusive interstitials, alongside the Core Web VitalsGoogle's user experience metrics. that the speed and mobile guide for landing pages covers in plain terms. A site that is slow, cramped or covered by overlays on a phone loses customers whatever the app strategy is, and fixing it raises the mobile conversion rate (CVRShare of visits or clicks that convert.) for first-time and repeat visitors alike.
The app then becomes an offer to a defined group: signed-in customers, frequent visitors, subscribers. They are the people for whom saved details, notifications and a home-screen icon are worth an install, and they are the customers whose behaviour the retention and payback guide shows decides whether acquisition pays. The mobile website links them into the app with deep linksA link that opens a specific screen inside a mobile app rather than its home screen. that open the matching screen when the app is installed and the web page when it is not, as the deep links and app analytics guide describes. The work of moving people across belongs to the web-to-app journeys guide.
The route has a practical advantage: the business learns whether repeat customers want the app before it spends on acquiring strangers into it. If the app does not hold the customers who already know the business, paid installs from people who do not will hold even fewer. Whether a prompt on the mobile website adds orders or only moves them into the app is a question of its own, and the web-to-app switch test is a design for answering it.
Once the app holds those customers and the next question is whether to buy installs from people who do not yet know the business, the comparison of web-to-app journeys and app install campaigns weighs the two routes side by side.
Scoring the decision with your own criteria
When the answer is not obvious, a weighted scoring matrix keeps the decision on criteria the team agreed before anyone argued for an option. Score at least three options rather than two: improve the mobile website only, website plus an app for repeat customers, and app-led. Set the weights first, then score each option against evidence from the business's own data.
| Criterion | Question to answer from your data | Evidence |
|---|---|---|
| Frequency of use | How often does a typical customer return? | Repeat visits and repeat orders per customer |
| Share of repeat customers | How much revenue comes from people who have bought before? | Customer and order records |
| Device features or offline use | Does the job need the camera, location, storage or a poor connection? | The task itself; support questions |
| What the website does badly | Which mobile tasks fail today, and could the website fix them? | Mobile conversion by step; the troubleshooting guide's checks |
| Messaging value | Would timely notifications be a service, not just promotion? | The jobs table above; what customers ask to be told |
| Discovery | Do people search the stores for this kind of thing? | Store autocomplete and rival listings, as the ASO guide describes |
| Capacity to run it | Who will own two listings, releases and app measurement? | Named people and time, not intentions |
Run the sensitivity check the framework describes: change the two heaviest weights and see whether the winner changes. If it does, the decision depends on a judgement the team should make explicitly rather than hide in the weights. The app installs objective covers the planning that follows if the app wins, and the app marketing hub shows where the store and paid work sit.
A florist weighs an app (illustrative)
Crinklebud Florists scores an app against its mobile checkout
Crinklebud Florists, an invented florist with a delivery website, is pitched an app with push offers. Before deciding, the owners pull a year of orders: 7,400 orders from 5,900 customers, so most people buy once or twice a year, typically for birthdays, anniversaries and Mother's Day. Phones bring 72% of visits but only 49% of orders, and the checkout loses most phone shoppers at the delivery-date and card-message step. They score three options out of 5 on five criteria weighted to 100: frequency of use (25), share of repeat customers (20), what the website does badly (25), messaging value (15) and capacity to run it (15). Fixing the website scores 4, 4, 5, 3 and 5, a weighted 4.25; the website plus an app for repeat customers scores 3, 3, 4, 3 and 2, a weighted 3.1; the app-led option scores 1, 2, 2, 3 and 1, a weighted 1.75. The order holds when the two heaviest weights are changed. Crinklebud rebuilds the mobile checkout, adds delivery-update texts for customers who opt in at checkout, and parks the app with a note to revisit it if repeat buyers ever pass a quarter of orders.
Common mistakes
- Deciding on enthusiasm rather than behaviour. Loyal customers say they would use an app; order frequency shows whether they would.
- Treating push as free reach. It reaches only people who installed, kept the app and granted permission, and the review guidelines limit marketing use to people who opted in.
- Budgeting the build and nothing else. Two listings, releases, permission and attribution are running costs for as long as the app exists.
- Neglecting the mobile website once an app exists. First-time visitors and shared links still land on the web.
- Building an app to fix a broken mobile checkout. If the website fails on phones, fix the website; an app adds an install step in front of the same problem.
- Offering the app to every visitor. The people who benefit are repeat customers; a new visitor wants the page they came for.
- Scoring two options instead of three. The hybrid of a better website plus an app for repeat customers is often the winner and is missed when the choice is framed as either-or.
- Ignoring who will run it. An app without a named owner for listings, releases and measurement decays quickly.
