Codrva Digital

Almost Everyone Built a Developer Portal. Around a Tenth Use It.

The tool is installed, the catalogue is populated, and the tab is closed. Adoption failures at this scale are not a training problem, and treating them as one is why they persist.

C
Codrva Team
Published Nov 13, 2026
7 min read
Dark green cover with a enterprise tools icon, the headline in bold condensed type and a panel listing four findings: The maintenance eats the roadmap; Nobody opens a catalogue for its own sake; Measure opens, not coverage; Product discipline is the actual variable

Dark green cover with a enterprise tools icon, the headline in bold condensed type and a panel listing four findings: The maintenance eats the roadmap; Nobody opens a catalogue for its own sake; Measure opens, not coverage; Product discipline is the actual variable

Backstage is used by more than 3,400 organisations and holds roughly 89 percent share among internal developer portal frameworks. Gartner expects 80 percent of large engineering organisations to run platform teams this year, up from 45 percent in 2022. By any measure of adoption, this category has won.

Then there is the other number. Average internal adoption of these portals sits around 10 percent. The organisations that bought in overwhelmingly have a portal that most of their engineers do not open, and the reported cause is not resistance. It is that platform teams exhaust themselves maintaining the thing before they ship anything developers actually wanted.

The maintenance eats the roadmap

The failure has a recognisable shape and it is not about the software being bad.

A portal of this kind is assembled from plugins, and those plugins have their own release cycles, their own dependencies and their own breaking changes. Keeping the installation current is genuine ongoing engineering, and it competes directly with building anything on top of it.

So a small team spends its first quarter standing the thing up, its second quarter keeping it running, and never reaches the work that would have made anyone open it. From outside it looks like a project that stalled. From inside it was fully occupied the whole time.

That is why the platform teams we see succeeding are the ones who reduced scope hard at the start rather than the ones who staffed up. A portal doing one thing well gets opened; one doing nine things adequately gets bookmarked.

Nobody opens a catalogue for its own sake

The second cause is that most portals launch with a service catalogue and little else.

A catalogue is genuinely useful to the platform team, who need to know what exists. It is close to useless to an engineer who already knows which service they work on, which is why the launch lands quietly.

What makes people open a portal is a task they have to perform anyway being faster there than anywhere else. Creating a new service with the standards already applied. Requesting access without finding out who approves it. Seeing whether the thing they just deployed is healthy. Those are errands, and an errand done in thirty seconds beats one done in twenty minutes through three other systems.

Pick the single most tedious recurring errand in your organisation and make the portal the fastest route to it. That is the whole adoption strategy, and it is more effective than any amount of documentation about the catalogue.

Measure opens, not coverage

Platform teams almost universally report the wrong number, and it flatters them.

Percentage of services registered in the catalogue is a measure of the platform team's own diligence. It says nothing about whether anybody else benefits, and it can reach one hundred percent while adoption stays at ten.

Report weekly active engineers instead, and the count of tasks completed through the portal rather than around it. Those numbers are uncomfortable at first and they are the only ones that describe whether the investment is working. It is the same substitution we argued for in where your throughput actually went: report the constraint, not the activity.

We ask for weekly actives in the first review and it is frequently the first time anyone has looked. The chart usually settles a disagreement that had been running for two quarters about whether the rollout was going well.

Product discipline is the actual variable

The reported divide in this space is not between tools. It is between organisations treating platform engineering as a product discipline and those that renamed their DevOps team and changed nothing else.

Product discipline here means specific things. Somebody owns the roadmap and can say no. Users are asked what they need before it is built rather than after it is ignored. Adoption is a target somebody is accountable for. Features that nobody uses get removed rather than accumulating.

A team without those will produce a technically competent platform that nobody adopts, and will reasonably conclude that engineers are difficult. The engineers will reasonably conclude that the platform does not help. Both are describing the same missing function.

This is the same pattern as any internal tool that arrives without an owner, which we went through in when a personal script becomes business critical, arriving from the opposite direction.

Buy or build is now a real question

Given that maintenance is the constraint, the self-hosted default deserves examining.

The open framework gives maximum flexibility and requires dedicated headcount indefinitely. A managed option costs money and removes the upgrade treadmill that consumed the roadmap. Which is right depends entirely on whether you have a team whose job this genuinely is.

The honest test is whether you can name the person responsible for portal upgrades next quarter. If the answer is that it gets absorbed by whoever has time, self-hosting will fail, and it will fail slowly enough that nobody identifies the cause.

Our enterprise tools work starts from the errand rather than the catalogue, because a portal that saves an engineer twenty minutes a week gets opened and one that documents your architecture does not.

What to do if yours is already quiet

Assume the portal exists and adoption is poor. Adding features is the instinct and usually the wrong move.

Ask ten engineers what they did this week that was tedious, and check whether the portal helps with any of it. Then remove everything nobody uses, which is faster to maintain and makes what remains findable.

Then rebuild around one errand and measure weekly actives on it alone. A portal with one genuinely useful function and 60 percent adoption is worth considerably more than one with nine functions and ten.

Frequently Asked Questions

Why is adoption so low when the tools are good?

Because platform teams spend their capacity maintaining the installation rather than building what developers wanted. The tooling works; the roadmap gets consumed by upgrades and plugin breakage before anything useful ships.

What should a portal launch with?

The fastest route to one tedious recurring errand, not a service catalogue. A catalogue serves the platform team. An errand done in thirty seconds instead of twenty minutes serves everyone else.

What metric should platform teams report?

Weekly active engineers and tasks completed through the portal. Catalogue coverage measures the platform team's own diligence and can hit 100 percent while adoption stays near 10.

Should we self-host or buy managed?

Ask who owns upgrades next quarter. If the answer is whoever has time, self-hosting will fail slowly. Flexibility is only worth its price when somebody's actual job is maintaining it.

Our portal is already built and quiet. What now?

Do not add features. Ask ten engineers what was tedious this week, remove everything nobody uses, and rebuild around one errand. One function at 60 percent beats nine at ten.

Is this a training problem?

Almost never. Adoption failures at this scale are a product problem, and treating them as training produces sessions people attend once and a portal they still do not open.

How big should a platform team be?

Small enough to be forced to prioritise. The teams that fail are frequently the ones large enough to attempt everything, because attempting everything is what consumes the roadmap.

What separates the organisations that succeed?

Product discipline: a named roadmap owner who can say no, users consulted before building, adoption owned as a target, and unused features removed. Not tooling choice.

Talk to our team if you have a developer portal and cannot say how many engineers opened it last week.

Share this post:
← Back to Blog

Comments (0)

Leave a Comment

Minimum 10 characters

No comments yet. Be the first to comment!

Chat with us