A network diagram with many shared dependencies emanating from a single, central node.

Will The All-in-One Solution Actually Cost More?

Every now and then, a major technology outage reminds millions of people that many seemingly separate workflows depend on a surprisingly small number of underlying components. Even if just one part of a system fails, many functions can break: messaging, scheduling, accounting, storage, security, and [gasp] precious AI tools.

One intuitive response is to minimize opportunities to fail by reducing the number of technologies we rely on: fewer vendors or platforms mean fewer systems to secure, train, maintain, integrate, and troubleshoot, right? Actually, that’s only half right.

Reliability depends not only on the number of things that can fail, but also on the number of things that can fail together.

I’m reminded of this every time core technological infrastructure goes down. But I continue to see organizations chase the intuitive-yet-incomplete all-in-one “solution”. And I have yet to find a clear, cogent, concise objection to the underlying fewer-is-better intuition. So I thought I’d take a crack at it in this post.

1. From Efficient to Resilient

The argument starts from a fairly uncontroversial goal:

Premise 0. All else equal, we should minimize opportunities for technological failure.

This failure-reducing premise implies a need to reduce the number of vendors, platforms, etc. After all, embracing more systems means embracing more opportunities for failure. So perhaps we should enable as few systems as possible — just one, if it can handle everything. That last part is the premature part. It ignores other realities.

Premise 1. Even well-designed and well-managed technologies sometimes fail.

If there were one system that could both (a) do everything we need and (b) never fail, then the one-system approach might work. Alas, no system is immune to failure. And this is where logic begins to turn against the one-system-is-best intuition.

Premise 2. A reliable organization needs critical functions to continue during technological failures.

If each component of our system were as independent as they feel from the user’s perspective, you may think one component can fail without taking down all the other components. If a vendor’s email client fails, their chat will still work, right? And if their chat breaks, we could fallback to their voice or video services, right? Not necessarily!

Seemingly separate components in a system often share the same underlying dependencies and infrastructure. So even if only one vertebra in a tech backbone breaks, everything that depends on it may stop working: email, chat, voice, video, browser, cloud, security, etc.

Premise 3. System components with shared infrastructure can fail together.

So supporting multiple systems does not necessarily increase risk; adding a second system may introduce another opportunity for failure, but it may preserve some critical functions during the failure. Of course, if the primary and backup systems depend on the same system infrastructure, they can fail together. So having a backup is necessary but insufficient.

Conclusion 1. Every critical function needs a fallback that survives when its primary option fails.

There may not be any all-in-one solutions that can credibly claim to fulfill this need for independent backup functionality. Why?

Premise 4. Even the best all-in-one solutions contain components that share some dependencies.

Sure, when one vendor’s infrastructure is working as intended, their all-in-one ecosystem can meet many or even all of our needs. But the shared dependencies in their all-in-one systems create shared failure risks. When a shared dependency fails, the all-in-one system can become an all-encompassing problem. So we need at lease one more system.

Conclusion 2. Some critical fallbacks need to be independent of the primary ecosystem.

You might be thinking, “Ok. We will enable an independent backup system when the primary system fails, but until then we need to support only one system to minimize risk, right?” No. A dormant backup system is not enough.

Think of how long it took people to learn how to work from home during the pandemic. Learning a new system takes a lot of time — and habit formation is hard, even for the tech savvy among us. (Years after videoconferencing went mainstream, people still have to say, “Nick, you’re muted”.)

Conclusion 3. Critical fallbacks need to be seeded before the primary system fails.

For a backup system to be truly ready, people need to be familiar with it before the primary system fails; they need the backup to fit their preferred settings, workflows, habits, etc. Otherwise, we get lawyers-with-cat-faces every time our primary system fails.

The infamous Zoom call of a legal proceeding in Texas, with one of the lawyers unwittingly appearing as a kitten.
“I am not a cat,” says a lawyer who does not know how to turn off a face filter in a video conferencing app. Source: 394th District Court of Texas.

After considering the premises above, it may seem obvious that an all-in-one solution is not enough. And yet it may not be obvious to the decision-makers in charge of technology policy or procurement!

2. What does this mean in practice?

An organization can designate one well-supported system as its default, but it should continuously preserve enough independent alternatives to keep critical work flowing when that primary system fails.

Notice that I have not argued that we need an independent, redundant alternative for everything. The more we need some function to survive technological failures, the more we need an independent redundant backup of that function. So we can consolidate expendable functions, but we need redundancy for critical functions.

With the conclusions and caveats on the table, we can begin to answer questions that many information security officers face:

A. “Should we consolidate down to one vendor or system?”

I get the appeal of all-in-one ecosystems. Volume pricing, easier integration, and fewer contracts, and consistent experience may reduce financial or administrative costs.

We can all relate to this. Suppose my spouse and I decide to subscribe to only one streaming service per month. That would save us not only money, but also time (when choosing what to watch). But streaming is not a crucial part of life. So when a streaming service is down, we can just go outside, read a book, or maybe even talk to each other — ok, maybe streaming should be the fallback for those better activities.

Anyway, when it comes to critical functions, removing redundancy can reduce resilience.

So the ultimate question is not whether fewer vendors minimizes cost during the good times. The ultimate question is, “How many vendors, platforms, etc. minimize cost during both the good and the bad times?”

A table of tech companies with columns for "lost hours" and "lost wages".
The cost of internet down time by Fortune 100 company via Forbes, according to technofaq.org, which cited TollFreeForwarding.com.

Proper cost projections need to account for the times when a shared failure can take down critical functions. In the good times, a solid all-in-one solution seems ideal. But that all-in-one solution may become extraordinarily expensive in the bad times: loss of productivity, loss of revenue, loss of reputation, and — in certain high-stakes settings — loss of life.

B. “Should we support everything?”

Of course not. Nor should you try. Some systems are genuinely better for an organization than competing systems. And resources are scarce. So we have to prioritize. This is precisely why it can be wise to default to one system, platform, browser, etc.

But complete cost consciousness can also override the impulse to support only one system. I’ll say more about that in a moment. For now the point is that we cannot (logically) jump from, “We shouldn’t support everything” to the conclusion that, “We should support one thing.”

We have to include both the benefits and the costs of supporting only one system. And we have to compare that to the potential net benefit of maintaining a limited, strategic number of independent systems. As soon as you account for inevitable outages and other failures, you begin to realize how much you can lose while you wait for your all-in-one savior to be resurrected.

C. “Can we reserve the backup for emergencies?”

No. Because a fallback system is only useful if it is usable.

Suppose your organization blocks every platform except the one it prefers. That’s seductively streamlined, but at what cost? When your default fails, your tech team can theoretically enable the backup. But an emergency is a terrible time to stress test your backup system, its training materials, new dependencies, etc. So you cannot afford to deploy the backup systems only during emergencies. You need to allow people to use the backup before it’s your only working system. Otherwise your people will be slowly and poorly using the backup system precisely when you need them to be fluently using it.

Ideally, your organization would have a broad distribution of people who already use the independent alternative systems enough that they can sustain their critical work during emergencies while helping neighbors make the transition — your tech teams can’t simultaneously troubleshoot your primary system while training your organization on a new system.

Blocking all but one system may simplify a typical day in your tech support office, but transition costs mean this seemingly streamlined approach can actually reduce overall resilience.

So resilience means keeping critical fallbacks tested, accessible, and familiar even in the good times.

3. Summing Up

Finding optimal equilibria between thrift and resilience is not easy.

  • Too much redundancy creates unnecessary complexity.
  • Too little redundancy creates unnecessary dependence.

For unimportant functions, maintaining redundant backups may not be worth the cost. But for critical functions, the cost-benefit equilibrium changes: redundancy becomes worthwhile.

So consolidation can protect against complexity, but redundancy protects against correlated risks — and resilient organizations need both! So go ahead and prioritize, standardize, and default to your preferred system. Consolidate low-risk functions into as few systems as you like (if doing so actually saves resources). But don’t pretend you can get by with just one system for the functions you can’t afford to lose. Maintain independent fallbacks that can keep critical work alive when your preferred system cannot.

Published by

Nick Byrd

Nick is a cognitive scientist at Florida State University studying reasoning, wellbeing, and willpower. Check out his blog at byrdnick.com/blog