← Blog

The shadow portfolio: your company runs 300 internal apps and can name the usage of three

Ask a CPO how the checkout funnel converted last week and you'll get an answer to two decimal places. Ask how many of the company's internal applications were opened by more than ten people last week, and you'll get a guess — usually off by an order of magnitude.

July 21, 2026 · 6 min read

The most expensive software you never measure

Every enterprise has two software portfolios. The first is the one on the website: the products customers pay for, instrumented to the tooth, reviewed weekly, A/B tested to death. The second is the shadow portfolio: the HR self-service portal, the expense tool, the procurement workflow, the reporting layer someone built in 2019, the six internal dashboards that all answer slightly different versions of the same question — and lately, a fleet of AI-agent pilots. Hundreds of applications, thousands of employee-hours a day, and in most companies: zero shared telemetry.

The asymmetry is absurd when you say it out loud. The customer product gets a growth team; the tools your entire workforce lives in get an annual survey — if that. Yet the shadow portfolio often consumes more engineering headcount than the flagship, and its friction is a tax on literally every employee.

What the darkness costs

• Zombie apps survive. Without usage data, no one can say which apps are dead — so nothing is ever retired, and the maintenance budget compounds. Every app kept alive "just in case" is an engineer not building something useful.
• The loudest stakeholder wins. Roadmap decisions between internal products get made on anecdote and seniority instead of usage and sentiment.
• Duplicates multiply. Teams build a new tool because they can't see that another team's tool already does the job — or that it exists at all.
• Frustration is invisible until it's attrition. Employees don't file tickets about a bad internal tool; they route around it in email and spreadsheets, and the failure never reaches whoever owns the roadmap. (More on this: the captive user fallacy.)
• AI pilots repeat the pattern at higher cost. The newest wing of the shadow portfolio burns real inference money, and its usage is measured even less than the apps'. (See: the evidence problem.)

Why it stays dark

Nobody decided not to measure the internal portfolio. It's structural: analytics tooling is priced per user or per event, so instrumenting 300 apps is a budget item nobody will sign (the MAU tax). Instrumentation is a per-app project, so it competes with feature work and always loses. And tracking employees raises privacy questions nobody wants to own, so legal review becomes the place initiatives go to die.

Fix those three constraints — flat pricing, one-line install, anonymous by design — and the shadow portfolio lights up in weeks, not quarters.

What "lit up" looks like

One dashboard, every product on it. Most-used and least-used, top-rated and bottom-rated, trends against last quarter. A usage × sentiment quadrant that sorts the estate into fund, fix, niche, and retire candidates. In-context feedback flowing from every app into one triage queue, and from there into Jira as evidence-backed backlog items. Suddenly the questions that used to be unanswerable become Tuesday-morning routine: What should we sunset this year? Which team's tool deserves the next hire? Is the new expense app actually better than the old one?