Technology stack fragmentation cost for a growing company is unowned joins. Okta 2025: 101 apps on average. Flexera: complete visibility at 43 percent. The outage lives between contracts.
Technology stack fragmentation cost for a growing company is the hours and errors created when website, hosting, security, and support sit in four contracts. Okta's Businesses at Work 2025 report put the average customer at 101 apps. App sprawl is not the same as four vendors, but both produce the same failure mode: no one owns the join.
I am Amelia S. Gagne, CEO of Kief Studio. We build across SMB to mid-market operations, including regulated stacks. This URL owns fragmentation cost. The invoice-versus-claim version lives on IT platform consolidation cost savings vs vendor claims.
It is the sum of idle software, delayed changes, and incidents that have no owner. Flexera's 2025 State of ITAM Report found complete visibility across the technology stack at 43 percent, down from 47 percent. When you cannot see the estate, you pay for copies of the same job in different tools.
A typical growing stack looks reasonable on paper: an agency built the site, a host runs the VPS, a scanner watches the edge, a freelancer ships code, a helpdesk takes tickets. Each contract is narrow. The outage is wide. The person who can restart nginx is not the person who can patch the app, and neither of them sees the DNS change.
Two vendors have one join. Four vendors have six joins. Eight have twenty-eight. Each join is an API, a shared secret, a person who remembers the exception, and a test you skip under deadline. Disconnected data is the same graph with CSVs instead of servers.
BCG, in analysis cited by CIO Dive in 2026, put SaaS at 21 percent of total tech budgets in 2024, up from 13 percent in 2019. More of the budget is now a subscription that can be bought by a department without a shared architecture review. That is how a growing company collects parallel CRMs.
NPI's 2025 sourcing study found 82 percent of enterprises pursuing supplier reduction, while 66 percent already concentrate 80 percent of spend with 25 or fewer vendors. The pain is in the long tail: the last 20 percent of spend spread across the rest of the list. That tail is where fragmentation cost lives for a company that is still hiring its first full-time IT person.
When the contact form dies, hosting says the box is up. The agency is offline until Monday. The scanner fired an alert that landed in a mailbox nobody triages. Each vendor performed their statement of work. The weekend of missed inquiries is yours. That is not bad luck. It is the contract design.
A single team that owns the stack cannot export the join. Either it works or they repair it. That is the argument in vendor consolidation for regulated industries, applied here to a growing company that is not yet a fund administrator. Security questions for vendors get sharper when you know who holds the logs.
Monitoring that means anything needs a picture of normal. Security rules that do not page you at 2am need the same picture. Deploys that do not break mail need to know which hostname still points at the old box. That knowledge is not in any one vendor wiki. It sits in the middle.
Security architecture fails the same way: controls on each slice, no one tracing a request from DNS to database. kief.dev is the public engineering surface for how we keep that path short. The cost of keeping it long is this page.
Do not confuse this with "buy one suite." A payroll system can stay separate. A lab LIMS can stay separate. Fragmentation cost appears when functions share runtime: host, app, identity, backup, and who is on call. Those should share a brain.
If the unknown arrows outnumber the documented ones, you are already paying fragmentation cost. You just are not booking it. LTFI is one way to hire a department that holds the joins. It is not the only way. The requirement is ownership, not a brand.
Okta Japan's 2025 cut of the same Businesses at Work dataset put organizations under 2,000 employees at 71 apps on average and organizations at 2,000 or more at 247. A 20-person firm that copies the 101 headline will overbuy. A 20-person firm that already has 40 overlapping tools is rehearsing the 71. Count unique jobs, not logos: site, identity, mail, files, books, CRM, support. If two logos share a job, you are paying fragmentation cost on purpose.
Flexera's visibility drop to 43 percent is the companion number. You cannot retire a duplicate if you cannot see it. Start with SSO or card statements. Those two lists are usually more honest than the "official stack" slide. Then kill the duplicate that has the older last-login date. That is cheaper than a transformation program.
The security review on briansgagne.com is the posture of those remaining tools. Fragmentation cost is the hours before that review even starts.
No. Clear APIs and documented interfaces can keep independent vendors honest. Consolidation is the stronger fix when the functions share runtime and on-call. Specialists with a clean boundary can stay.
Count idle seats, count hours spent routing tickets between vendors, and count incidents with two "not us" replies. Flexera's 43 percent complete-visibility figure is the industry version of that blindness.
Okta's 101 is an average across its customers, including large estates. A 30-person company does not need that count. If you are adding apps faster than you are retiring them, you are rehearsing the 101.
When the work is independent (payroll vs website) or the specialist gap is material and the interface is an API you control. If two vendors must share a database or a night, they are already one system wearing two logos.
Over 92% of the Western world's data sits on U.S.-owned servers, and the CLOUD Act lets authorities demand access regardless of location. Owning the stack is not ideological. It is jurisdictional, operational, and the difference between answering 'where is the data' with a street address or a vendor FAQ.
We run our own servers, our own git, our own identity provider, our own monitoring. Not because we're paranoid — because the math works out and the control matters.
Cloud hosting isn't inherently better than self-hosted — and self-hosted isn't inherently more secure. The right answer depends on variables most cloud comparisons don't bother to address.
Work With Us
Kief Studio builds, protects, automates, and supports full-stack systems for businesses up to $50M ARR.
Newsletter
Strategy, psychology, AI adoption, and the patterns that actually compound. No spam, easy to leave.
Subscribe
