Outsourcing is a loan.

Outsourcing a capability is a loan. You get the thing now, and you pay interest on it for as long as you don't own it.

Sometimes that's a great deal. Early on, when you don't know what the product is yet, renting a team is cheaper than building one you'll have to unwind. Nobody should feel bad about borrowing to get a roof over their head.

The trouble is that nobody sends you a statement when the interest goes up. The vendor that was a bridge in year one quietly becomes the foundation, and you're still paying as if it were temporary. And the invoice is only part of the interest. The rest is everything you never learned how to do yourself.

What I walked into

When I was VP, Digital & Innovation at PenFed, the mobile app had been built by a very large digital agency. The trouble was everything around it.

There was no standing team on it at the agency, only standing support. Nobody was working a backlog, because there wasn't one. When I arrived, PenFed had no backlog for its own app. Every update meant going around to all the stakeholder teams inside PenFed, cobbling together what each one needed and why, then sitting down with the agency to work out what it meant and write the requirements. So the gap between "we want this" and "members have it" was enormous.

And the app was a black box. A lot happened inside it that PenFed couldn't see or understand. That became the agency's moat. It's hard to build something new, or even change something small, when you don't know how the code works.

Because someone else had built it, PenFed had never built the skill. No internal mobile developers had been hired. That's the part that compounds. The longer you rent, the less you know, and the less you know, the more you have to rent.

The instinct in a lot of rooms is to renegotiate at that point. Squeeze the rate, add a service-level agreement, call it a win. That's refinancing. It lowers the monthly payment and leaves the balance where it was.

Pay down the principal, in order

We brought it in-house, and the order mattered.

First, mobile engineers. Their first job was to look at everything happening outside the black box and work out what was happening inside it. You can't take back what you don't understand.

Second, we started building PenFed's own version of the app, internally.

Third, we built out more of the mobile team, added a product manager and started a backlog.

I gave the team three goals. A mobile release every month. A backlog a year out of the major things that needed doing. And a meeting with stakeholders every month, so we stayed aligned on what they needed and why. That last one was the answer to the old scramble: the conversation with stakeholders had a standing slot, instead of starting from zero every time someone wanted a change.

Moving members without stranding them

When the new app was ready, we had the agency do the first switchover. Members needed a new app, and we wanted the on-ramp to be easy. We ran both apps in parallel and encouraged members to move over a three-month ramp, instead of announcing that the old app was about to turn off.

That's the part of a move like this that people underestimate. You can build the better thing and still lose people on the day you make them switch. Give them a bridge and time to cross it.

What it paid back

Release cadence went from about once every 14 months to every month. We could have gone faster. We chose monthly on purpose, because we didn't want members constantly seeing the update screen and starting to feel the app was unstable. Development cost came down 80%, with a much smaller team, and we made some key architectural decisions along the way.

The cost line is the headline. The bigger change was what became possible once the team was inside.

We connected the app to Adobe Experience Cloud, so the language, imagery and campaigns running in marketing and on the website could show up in the app too. We set up shared text blocks so the app and the website said the same things in the same words. That opened up personalization. Little by little, how products showed up in the app started to match the person instead of being the same for everyone. If you loved to travel, the credit card offers you saw were travel cards. The language might echo an ad you'd already responded to. (The segments behind that came from the marketing analytics work in "The brand is the product's first screen", combined with UX research on how people actually used the app.) We also added the ability to resume an application from the app, which wasn't possible before.

By pulling it inside, all the connections, content and knowledge we had as an organization could come over to the app in real time. None of that was possible while the app lived in someone else's black box.

People ask what I'd do differently. Nothing.

When to bring it in

Not every capability should come home. Three questions help:

Does it touch the customer every day? Back-office plumbing can stay rented a long time. The thing your customer holds in their hand usually can't.

Does it change often? Interest compounds on change. A capability that's finished and stable is a cheap loan. One you'd like to update every month is an expensive one.

Do you know what "good" looks like better than the vendor does? If you don't, you're borrowing their judgment as well as their hands, and that's worth paying for.

If one of those is missing, keep renting. If all three are there, you're paying interest on something you could own.