Periodically, we write about this. SaaS and downtime. Downtime and churn. What is changing? What’re people doing to combat it? How do people feel about products they use crashing and what impact will it have.
Not a good one. As you can imagine.
A 2023 survey found that 68% of SaaS customers would consider switching after just one major outage.
And when we consider that 3.5% is the benchmark of churn for B2B (B2C being a couple times higher) we can see that a pondering 68% is no good thing.
But you probably didn’t need us to tell you this. In 2025, we can report, there have been no earth shattering pieces of research pointing out anymore of a relationship between downtime and churn than previously known.
Only many, earth-ruffling bits of research confirming your intuition that your customers don’t like it when your app goes down and that they might try another instead.
So, what do you need us to tell you?
Is the frequency of downtime or its impact getting better or worse (for SaaS applications and in general) 2025 edition
This is more of a mixed bag. And the more interesting for it.
A recent report from the uptime institute has this to say:
- Overall, outages are falling (that’s good!)
- However, while the major building blocks of the internet report fewer outages (AWS, Meta)…digital service providers (mid-to-large level SaaS businesses and other such providers) have reported in increase in outages (that’s bad)
- Over two thirds of publicly reported outages come from failures which reporters DON’T CONTROL (i.e. a third-party service like Twilio) – although you can see how it being someone else’s fault would lead to one being happy to get good and public with it (that’s probably bad unless you’re reading from Amazon HQ just to gloat)
- Simple human error has increased it’s hand-share in whom wields the scythe, the report states: “Nearly 40 % of organizations have suffered a major outage caused by human error … 85 % stem from staff failing to follow procedures or from flaws in the processes and procedures themselves.” (that’s… interesting)
Add to all this that fact that outages are actually getting more expensive and you have a potentially bleak prognosis. The type of apps you own are failing more often. It’s costing more and it’s outside your control.
So what can you do, SaaS-offering, churn-fearing reader?
Many pieces of advice remain timeless. Partner with a great 24/7 support provider well-versed in preventing SaaS downtime. Ahem. Rehearse failure. Prioritise stability.
However, there are some 2025-y things you could be doing:
Implement or build on strategies to protect your SaaS app from upstream, third-party outages
In general, you or your support provider should be monitoring first-party metrics for your providers, building graceful degradation paths and working on ways to minimise blast radius once an upstream provider goes super-nova.
A new technique for 2025
New CPaaS services like Kompose CPaaS Mesh allow you to switch communications providers on the fly. So, if you rely on, let’s say, Twilio for your OTPs, you can begin using a backup once the mesh detects X delivery failures – without touching your code or paging engineers.
Put AI-guard rails on your runbooks to keep your humans in check
In general, your runbooks should be detailed and drilled to prevent fat-finger syndrome. But now there’s extra help.
A new technique for 2025
Services like PagerDuty Copilot, the newer Atlassian Compass AI Runbooks or simply services which allow GTP plugins (other LLMs are available) enable you or your provider to install an AI agent to to walk responders through each runbook step in real time. This can prevent staff making mistakes in the heat of the moment.
Run chaos drills in a safe digital twin to expose hidden failure points
Depending on your size/level of maturity, you should be running regular game days. Chaos engineering can be a life saver. However, you can now delegate to the virtual world.
A new technique for 2025
Tools like ChaosMesh Shadow Cluster or Gremlin Digital Twin can spin up a carbon-copy of your prod straight from CI, auto-inject a failure and email a pass/fail report that shows how fast your recovered.
Hit the Big-Red Freeze the instant prod starts wobbling
In general, you should pause every deploy the moment an incident begins so fresh code can’t make things worse. Now that’s easier than ever.
A new technique for 2025
Universal freeze APIs are now built into GitHub, GitLab, CircleCI and LaunchDarkly. They essentially allow a single Slack message to halt everything.
Switch your SaaS product to a really good 24/7 support provider
You should, of course, already be working with an excellent 24/7 support provider or have an envious internal function.
You may be relying on the support functions of various providers, (your public cloud, the partner tier of your monitoring provider, etc) however, this might be the year to hop.
A new technique for 2025
Look for 24/7 support providers that do a few of the following: offer a buck-stopping service, i.e. they are the single point of responsibility, monitoring your upstream providers, APIs, cloud and code with runbooks in place for all scenarios; offer a follow-the-sun staging model, i.e. there is always a human being working a normal shift somewhere across the world (not a dev getting out of bed or an alert that won’t be looked until Monday, 9am); they compile helpful lists of 2025 downtime-busing tactics.
How we can help
If the various bits of self-promotion in this article went over your head, allow us to aim lower.
We’re Just After Midnight, the uptime experts and downtime scourges. We have delivered 24/7 support, monitoring and remediation to SME and enterprise SaaS products. Read about our churn-bursting work:
- Preventing downtime for App Store No. 1 The BodyCoach App from Joe Wicks
- Supporting insurtech product SIMLIFI with sub-three-minute incident response time
- Keeping the lights on for Five Guys & Tacobell journey planner LineTine
Or to get less downtime, less churn and 24/7 peace of mind, just get in touch.
