Web-to-app journeys and app install campaigns at a glance
A web-to-app journey starts from people the business already has: website visitors and customers for whom the app does a familiar job better. The web-to-app guide plans that move route by route; here it matters only that the people it reaches know the business, and many have an order history the app can carry.
An install campaign starts from people the business does not have yet. The install campaigns guide covers how networks, the event a campaign optimises towardsThe conversion event an ad set or campaign is told to optimise delivery towards, such as a purchase, lead or install., creative and budgets are set; for this choice what matters is that the network, not the business, finds the audience, and the ad and the store listing have to explain the app from nothing.
Neither guide weighs one route against the other, and the app marketing hub sets store presence, paid installs and launch side by side as the ways to get installs without putting the website in that table. So the choice is easy to miss. It turns on one question: do the people the app needs already know the business? Where they do, the website reaches them at no media cost; where they do not, only a channel that reaches strangers will find them. The split echoes the one in the comparison of customer acquisition and retention: a web-to-app journey deepens the relationship with customers already won, an install campaign wins new ones, and the balance between them should follow the economics rather than habit.
Compare table
Web-to-app journeys vs App install campaigns
| Row label | Web-to-app journeys | App install campaigns |
|---|---|---|
| Who arrives | People already on the website or in the business's emails, ideally signed-in repeat customers, account holders and subscribers | People the network finds for the campaign's optimisation goal, most of whom have not used the business before |
| What they already know | The business, its products and often their own order history; the app only has to be better for a job they already do | Possibly nothing; the ad and the store listing have to explain the app and earn the install |
| What it costs | Building prompts, page-to-screen links and sign-in across the switch; no media per install, but each prompt can cost the web visit it interrupts | Media spend, creative produced in volume and measurement setup, paid on every install the network delivers |
| Where the user lands | The matching app screen, or the store and then that screen through a deferred deep link | The store listing, default or custom, and then the app's first session |
| Speed and scale | Bounded by the website's existing customers; routes keep working once built | Starts when the campaign launches and scales with budget once the network has enough optimisation events to learn from; quality falls as volume is pushed |
| Measurement | Routes tagged with campaign parameters, and a shared user identifier set at sign-in on web and app | A measurement partner as the system of record across networks; on iOS, networks receive only limited postbacks |
| Quality signal | Orders or the main goal action across web and app combined, against similar customers who stayed on the web | Activation and early retention by cohort, and cost per activated user |
| Effect on the website | Prompts sit on web pages and can lose, or merely move, the orders those pages produce | None directly; the ads open the store, not the site |
| Main risk | Crediting the app with orders loyal customers would have placed on the web anyway | Buying installs that never activate, including fraudulent ones, or installs that would have happened anyway |
| Typical misuse | A full-screen app overlay shown to every mobile visitor | Judging the campaign on cost per install |
The rows that decide most cases are the first two. If the people the app is for are already on the website, a route that moves them reaches the right audience without paying a network to find it. If they are not, no amount of work on prompts will produce them, and the business has to buy reach. The cost and risk rows then explain why neither side is free: the web route spends the visits it interrupts, and the paid route spends money on every install, whether or not that install becomes a user.
When to choose web-to-app journeys
Choose a web-to-app journey when the app serves people who already use the website and does a job better for them: customers who reorder, people who manage an account often, and members who use the service every week. The web-to-app guide asks for that group to be written down as a definition before anything is built, because it decides where prompts appear, which emails carry a link and which comparison measures success. First-time readers from search and one-off buyers are left on the website, which serves them better.
It is also the right first step when the app is new. The guide to choosing between an app and the mobile website recommends making the website good for everyone and offering the app to customers who come back, and gives the practical reason: offering the app to returning customers first shows whether they keep it before any money goes on finding new users. An app that loses the people who already trust the business is unlikely to keep people who have never heard of it.
The mechanics belong to the web-to-app guide and its neighbours, and this page does not repeat them. Links from each web page to the matching screen, and deferred deep linksA link that opens a specific screen inside a mobile app rather than its home screen. that carry a new installer through the store to that screen, are explained in the deep links and app analytics guide, which also covers the events that show what people did once they arrived. For reading the routes, the web-to-app guide points to Google's documentation on URL builders, which sets out the campaign parameters that go into a tagged link, so that each route can be read by source. The growth loops framework applied to app installs shows the same journey as a content loop that carries web readers into the app, and names the link as the joint that most often breaks.
A web-to-app journey is the weaker choice when the website has few repeat customers, when most visitors arrive once from search, or when the app has to grow beyond the people who already know the business. It can only move the audience the website already has.
When to choose app install campaigns
Choose install campaigns when the app has to reach people who do not know the business yet: a new region, a new audience, an app whose intended users mostly never visit the website. The app marketing hub describes paid installs as the lever that buys reach on ad networks and scales with budget, and the install campaigns guide separates networks that meet people already searching for an app from networks that have to make them want one.
Two things should be in place before the money goes out. The store listing has to convert, because every paid click lands on it: the app store optimisation guide covers the listing, the work known as ASOImproving an app's visibility in app store search and the share of store visitors who install it., and the tutorial on custom store listings for campaigns builds variants that repeat each ad's promise. And the activation event has to be defined and firing. The install campaigns guide lists buying installs before that as its first common mistake, because a campaign with no activation to read falls back on install cost, and the cheapest sources then look best. The tutorial on finding the app's activation event does that analysis from the app's own data, and the Activation RateThe share of new users or sign-ups who complete the activation event, the action marking first real value, within a set window. it produces is the measure both routes can share.
Set the money from the economics rather than from what a network suggests. The install campaigns guide starts from what a user who stays or pays is worth and derives the most a campaign may pay per install; the CPA calculator converts spend, installs and the share reaching each event into the cost at the event that matters, and the marketing budget calculator sizes the total against the users the plan needs. Who gets credit for each install is the subject of the app attribution guide, which explains why an app buying from several networks relies on a measurement partner.
Install campaigns are the weaker choice when the app has not yet held the customers who know the business, when paid visitors convert on the listing far below organic ones, or when most people a campaign would reach are already website customers. In each case the business pays for something cheaper to fix or free to reach.
Decision table
| Situation | Choose | Why |
|---|---|---|
| Repeat customers order on the website and the app does their job better | Web-to-app journeys | They already know the business, and the app only has to be better for a job they already do |
| The app is new and nobody knows yet whether customers keep it | Web-to-app journeys first | The business learns whether the app holds people who know it before paying for strangers |
| The app has to reach a region or audience where few people know the business | App install campaigns | The website has nobody there to move |
| Most website visitors are one-off buyers or readers from search | App install campaigns, with the website left to serve its visitors | Prompting them delays the visit, and few would keep the app |
| Paid visitors convert on the store listing far below organic ones | App install campaigns paused until the listing is fixed; web routes can continue | A campaign is judged through the listing it sends people to |
| The activation event is not yet defined or firing | Neither at scale yet: define the event first | Without it, neither route can be judged on active users |
| Web prompts are live and nobody knows whether they add orders or only move them | Web-to-app journeys, tested | A randomised switch test reads orders across web and app combined |
| Both run, and credited installs rise where the business is already well known | Install campaigns read against a holdout | Credited installs show which ad came last; only a holdout shows how many people would have installed regardless |
Running both without paying for customers you already have
Most apps that grow end up with both, and the order matters. The growth loops page for app installs advises starting with the loop built on something users already do, and adding paid reinvestment only once each acquired user is known to pay back. For an app with a website behind it, that means the web routes first, because they reach people who already know the business, and install campaigns second, aimed at the people the website cannot reach.
The overlap is where money leaks. An install campaign that reaches the business's own customers buys an install the website could have produced without media, and the attribution guide warns that installs from people who were coming anyway are often credited to the last ad shown. Three habits keep the two apart. Tag every web and email route, so that the app records which route each person came by and installs from the website can be counted beside those the partner credits to ads. Read how many newly credited app users sign in with an existing website account: the shared user identifier the web-to-app guide recommends joins web and app records for the same person, which gives a working estimate of how many paid installs came from customers the business already had. And point install campaigns at places and audiences where the business is not known, so that the reach being bought is reach the website does not have.
Credit is not cause. The attribution guide describes incrementality tests that set an exposed group against a comparable Holdout GroupA randomly chosen part of an audience kept from exposure to measure what exposure caused. left unexposed, split by region, by period or, where a network allows it, by audience. The comparison of iROAS and ROAS explains the same gap for revenue: the figure a platform credits is good for steering a campaign, while the figure measured against a holdout is the one that shows whether the budget line is worth funding. The web side has its own version of the test: the web-to-app switch test holds part of the website's visitors back from the prompt and counts orders on web and app combined, so that an order moved into the app is not mistaken for an order added.
Finally, read both on the same unit. A CPIAverage ad cost of each app install. on one side and a count of switches on the other compare nothing. Count active users added by each route, using the one activation event the app has agreed, and read the web routes beside the visits they interrupt and the paid routes beside the money they spend. The comparison library holds the other pairs that are easy to confuse when one choice is really two.
A regional grocer splits its app's first year between its website and paid installs
Skarloe Grocers, an invented regional grocery chain with 14 shops and a busy click-and-collect website, launches an ordering app in March. The website has 58,000 signed-in customers, and the team counts as worth moving those with at least three click-and-collect orders in the previous six months: 16,400 people. For ten weeks half of them, chosen at random, see a dismissible banner on their order-history and order-confirmation pages and a link in the "order ready" email that opens the order in the app; every route is tagged. An active ordering user is anyone who orders in the app within 28 days. By the end, 2,300 of the 8,200 prompted customers have ordered in the app and 1,800 of them are still active ordering users, while orders per customer across web and app combined are 3% higher than in the unprompted half, so the routes added orders rather than only moving them. The prompts go live for the whole group.
A two-week install trial across the home region, at £200 a day, is credited by the measurement partner with 2,700 installs at about £1.04 each, but 1,850 of the people who install sign in with an existing website account. Skarloe stops buying installs where its shops are well known and moves the campaign to four of the six towns where it opened shops that year, optimised to a first app order rather than the install, with the other two new-shop towns held back. Over six weeks at £150 a day, £6,300 in all, installs in the campaign towns rise by 3,300 against the held-back towns and active ordering users by 540, about £11.67 each. Skarloe keeps the website routes for every new repeat customer, buys installs only in towns where few people know it, and reports both each month as active ordering users added.
Common mistakes
- Buying installs from your own customers. A campaign aimed where the business is well known pays for people the website could have moved for nothing.
- Running install campaigns before the app holds existing customers. If people who know the business do not keep the app, strangers will keep it even less.
- Prompting every visitor. Someone reading one article from search loses the page to a prompt; offer the app only to customers it serves better.
- Treating the web routes as free. A prompt can lose the order on the page it sits on, so read the interrupted visit beside the switch.
- Untagged web and email routes. Without campaign parameters nobody can tell which installs the website produced, so ad-credited installs cannot be checked against them.
- Judging the campaign on cost per install. A cheap install says nothing about whether anyone kept the app.
- Crediting the app with loyal customers' habits. Early switchers are the keenest customers, so judge the app against similar customers who stayed on the web.
- Sending paid traffic to a listing that does not convert. Every paid click lands on the listing, so fix it before scaling spend.
- Reading the two routes on different units. Count active users added on both sides, against one agreed activation event.
