My work with Branch

Almost never where it was tested.

A tap can land a person in one of four states, and most teams only ever check the first: the app is installed and the person is signed in; the app is installed and they are not; the app is not installed at all; or there is no app involved and the destination is the web. The third state is the one that costs money, because the link has to survive a trip through an app store, where the tap and the install are two separate events on two separate surfaces.

The failures that come up again and again:

• The link opens the store listing instead of the content, so a person who wanted one recipe, one shoe or one article gets a home screen and leaves.

• The link is correct and the route does not exist. The destination is a real page on the web and has no equivalent screen in the app, so the app opens at the top and nobody can tell from a dashboard that anything is wrong.

• The link works from a browser and not from inside a social app's own browser, which is where a large share of paid traffic actually sits.

• A redirect in the middle of the chain strips the parameters, so the page loads, the person is happy, and the campaign gets no credit.

• A printed QR code points at a URL that someone changed six months later.

The order I work in: inventory every live destination before touching anything, because the list is always longer than the team thinks. Then test the four states against the five destinations that carry the most spend. Then the parameters. Then the reporting. Fixing the reporting first is the common instinct and it wastes a month, because you end up building an accurate picture of a broken route.

How do you check an attribution number before you trust it?

Ask three questions, in this order, and do not read the dashboard until you have the answers.

What event is being counted? An install, a first open and a registration are three different numbers, and teams quote whichever one is highest without noticing they have changed the subject.

When is it counted? A click window and a view window include different people, and a window that stays open improves every report that runs later. Pick one of each, write them down, and keep them when they flatter you less.

Who is counted twice? A person on a phone and a laptop is one customer in the brand's own database and two in an ad platform. Every blended number inherits that gap.

Then reconcile. Not the whole account: one campaign, one week, against the brand's own backend, line by line until you can explain the difference. When the difference has a reason, the number is usable even if it is wrong, because everyone is wrong in the same direction. When nobody can explain it, no amount of extra reporting helps.

What usually turns out to be the problem is not the tool. It is that the definition was never written down, so two honest reports disagree and the hour goes on arithmetic instead of the funnel. The same argument, and the one-page way out of it, is in Attribution without the arguments.

What is different about running growth inside a large brand?

The constraint is not knowledge. Large brands have more of it than anyone. The constraint is the number of people who have to agree before anything moves, and the number of tools that each hold a piece of the answer.

So the work splits in two. There is the change, which is usually small. And there is the document that lets someone who was not in the room keep the change working after a reorg. The second one is what decides whether the first one lasts.

One practical shortcut: find the person whose number the executive quotes in meetings, and start with them. It is rarely the person who owns the channel.

Who built Branch, and who backed it?

Four people who met at Stanford: Alex Austin, its CEO until 2023, who studied engineering and then took an MBA there and worked at NASA Ames before this; Mada Seghete, a Stanford MBA who was at Siemens and Deloitte, and who has been on the board since 2024; Mike Molinet, a Stanford MBA who was at 3M; and Dmitri Gaskin, the CTO, who was on Forbes 30 Under 30 in 2015.

NEA led Branch's first round and its $300M round in 2022. NEA has also backed Databricks, Cloudflare, MongoDB and Robinhood. Founders Fund led Branch's $35M round in January 2016 and invested again in 2017; it has also backed SpaceX, Palantir, Stripe, Airbnb and Spotify. Playground Global led the $60M round in 2017, and Pear VC, Madrona, Cowboy Ventures and Samsung Ventures also invested.

What are Branch's own numbers?

Two, and they are Branch's, with their dates.

Branch was valued at $4B when it raised $300M in February 2022. More than 100,000 apps and brands have used it since 2014.

Branch names 16 of its customers publicly: Reddit, Instacart, Shopify, Nordstrom, Credit Karma, Under Armour, T-Mobile, Warner Bros, BuzzFeed, Strava, Chipotle, Headspace, Puma, Grubhub, NBC and Philips.