The Nottingham Agency That Spent £47,000 on Cloud Bills They Didn't Need
Sarah Mitchell, Managing Director of CloudBridge Digital, remembers the exact moment she realized something was catastrophically wrong.
March 2025. End-of-quarter financial review. Their accountant highlighted a line item that had been growing steadily for 18 months: "IT Infrastructure & Software."
£4,100 in January 2024. £6,847 in December 2024. £7,215 in March 2025.
"I thought it was an accounting error," Sarah told me over coffee in their Nottingham office last week. "We're a 23-person agency. We build websites and mobile apps for regional clients. How could we possibly be spending seven grand a month on cloud infrastructure?"
Spoiler: They could. And they were.
The Setup: How a Smart Agency Made Stupid Decisions
CloudBridge Digital isn't some tech-illiterate dinosaur limping along with legacy systems. They're a modern, successful digital agency that wins contracts against much larger competitors.
Their services:
Custom web development (Laravel, React)
Mobile app development (native iOS/Android)
Cloud hosting for clients
Digital marketing automation
Their team:
9 developers
4 designers
3 project managers
4 digital marketers
3 admin/operations staff
Their mistake: Trusting that modern cloud infrastructure "just works" and consultants have your best interests at heart.
Let me show you how this unfolded.
2023: The Migration Everyone Recommended
In late 2023, CloudBridge decided to modernize their infrastructure. They'd been running a mix of on-premises servers and basic hosting. Clients were asking for more sophisticated solutions. Developers wanted better tools.
The consultant pitch: "Move everything to AWS. Implement infrastructure as code. Use managed services. Scale automatically. Pay only for what you use."
The reality: Pay for everything the consultant configured, whether you use it or not.
They hired a "cloud transformation consultant" through a personal recommendation. Three months. £18,000. Nice guy, impressive credentials, zero ongoing accountability.
By February 2024, CloudBridge had:
AWS account with "production-ready" architecture
Microsoft 365 E3 licenses for entire team
Seven SaaS tools for project management, analytics, and collaboration
Auto-scaling configurations for client applications
"Redundant" databases across multiple regions
Three development/staging environments
Monthly cloud spend: £5,200 Previous monthly infrastructure spend: £1,800
"The consultant said the extra cost was 'enterprise-grade reliability' and we'd grow into it," Sarah explained. "We believed him because he had 'AWS Certified Solutions Architect' on his LinkedIn."
The Silent Bleed: 18 Months of Waste
Here's what nobody tells you about cloud infrastructure: It never scales down as aggressively as it scales up.
Waste Vector 1: The Development Environments Nobody Turned Off
The consultant set up three full environments:
Production (necessary)
Staging (arguably necessary)
Development (completely unnecessary duplication)
Each environment was configured identically to production. Full database replication, redundant load balancers, the works.
Cost per environment: £1,200/month Number of environments actually needed: 1 (production) 18-month waste: £43,200
"Developers were supposed to spin down the dev environment when not in use," Sarah said. "Nobody ever did. It ran 24/7 for 18 months."
Waste Vector 2: The Multi-Region Architecture for Regional Clients
The consultant implemented multi-region database replication "for disaster recovery."
CloudBridge's clients: Small to mid-sized UK businesses, mostly within 50 miles of Nottingham.
Number of clients requiring sub-second global latency: Zero.
AWS multi-region data transfer costs: £340/month RDS Multi-AZ deployment premium: £680/month 18-month waste: £18,360
Waste Vector 3: The SaaS Subscription Avalanche
Between January 2024 and March 2025, CloudBridge accumulated SaaS subscriptions like a tech hoarder:
Asana (project management): £299/month for 25 users
Monday.com (also project management, added because one client preferred it): £199/month
Mixpanel (analytics): £449/month for features nobody understood
Intercom (customer support): £359/month for 4 conversations per month
HubSpot Marketing (automation): £680/month, 90% features unused
Notion (knowledge base): £80/month
Miro (collaboration): £96/month
Total SaaS monthly cost: £2,162 Number of tools actively used by more than 3 people: 2 (Asana and Notion) 18-month waste: £27,000 (on subscriptions that duplicated functionality or went unused)
Waste Vector 4: The Microsoft 365 Over-Licensing
The consultant recommended Microsoft 365 E3 for "enterprise security."
What E3 includes: Advanced threat protection, data loss prevention, compliance tools, Azure AD Premium P1.
What CloudBridge actually needed: Email and file storage.
Cost difference:
E3: £23/user/month × 23 users = £529/month
Business Basic: £4.60/user/month × 23 users = £106/month
18-month waste: £7,614
"We were paying for enterprise compliance features designed for banks and healthcare," Sarah laughed. "We're a digital agency. We don't need HIPAA compliance tools."
Waste Vector 5: Auto-Scaling That Never Scaled Down
The consultant configured "intelligent auto-scaling" for client applications.
The configuration:
Minimum instances: 4 (per application)
Scale up threshold: 60% CPU
Scale down threshold: 30% CPU
Scale down cooldown: 10 minutes
The reality:
Average CPU usage: 15-25%
Applications never hit scale-up threshold
Applications kept 4 instances running 24/7 because minimum was hard-coded
Actual requirement: 1 instance per application, with 2-3 during peak hours.
Monthly waste: £840 on compute instances sitting idle 18-month waste: £15,120
The Discovery: March 2025
Sarah's accountant asked a simple question: "Why is your IT infrastructure cost growing faster than your revenue?"
Good question.
Sarah pulled up AWS Cost Explorer for the first time in six months.
"I nearly threw up," she said. "March AWS bill: £4,800. For what? We have 11 clients, 23 employees, and we're not exactly running Netflix."
She called an emergency meeting with their senior developer, operations manager, and the accountant.
The audit revealed:
47 active AWS resources
23 of them had zero traffic in the last 30 days
8 EC2 instances running 24/7 for "testing" environments
4 RDS databases that hadn't been queried in months
Data transfer costs of £680/month for backups nobody ever verified
SaaS audit:
11 active subscriptions
6 hadn't been logged into in 60+ days
3 had fewer than 5 active users despite paying for 25+ licenses
Total monthly SaaS spend: £2,162
Actual value delivered: Debatable
Microsoft 365:
E3 licenses for features 0% of users had enabled
Advanced security alerts going to an inbox nobody monitored
Compliance tools configured but never used
The Recovery Plan: April-September 2025
Sarah made a decision: No more consultants. No more "best practices." No more accepting vendor defaults.
She assigned their senior developer, James, to become "Cloud Cost Czar" (his term, he made himself a badge).
Month 1: Stop the Bleeding
Week 1 Actions:
Terminated staging and development AWS environments (kept production)
Cancelled 6 SaaS subscriptions immediately
Downgraded Microsoft 365 from E3 to Business Basic
Set AWS budget alerts at £2,500/month (50% of current spend)
Week 1 savings: £2,100/month
Week 2-4 Actions:
Eliminated multi-region database replication
Changed auto-scaling minimums from 4 to 1 instance
Implemented aggressive scale-down rules (2-minute cooldown)
Terminated 15 zombie AWS resources
Month 1 total savings: £3,400/month
Month 2-3: Right-Size Everything
James spent 40 hours analyzing actual usage patterns:
Discovery 1: Client applications averaged 8 hours of meaningful traffic per day (business hours). The other 16 hours they were burning money on idle compute.
Solution: Implemented scheduled scaling. Minimum 1 instance overnight, scale to 2-3 during business hours.
Monthly savings: £680
Discovery 2: RDS databases were provisioned for 10,000 IOPS. Actual usage: 400 IOPS average.
Solution: Downgraded from db.m5.large to db.t3.medium, reduced provisioned IOPS.
Monthly savings: £420
Discovery 3: They were paying for AWS data transfer between services in the same region because the consultant had configured VPCs poorly.
Solution: Restructured networking to keep data local.
Monthly savings: £180
Month 4-6: Process Changes
Sarah implemented mandatory "cloud governance":
New resources require approval (£50/month threshold)
Quarterly SaaS audit (usage checks, license optimization)
Monthly cloud cost review (15-minute standing meeting)
Developer training on cost-effective AWS usage
Client billing transparency (show clients exactly what their infrastructure costs)
Cultural shift: Developers started treating cloud resources like company money (because it is).
The Results: September 2025
Original monthly cloud spend (March 2025): £7,215 Optimized monthly cloud spend (September 2025): £2,650
Monthly savings: £4,565 Annual savings: £54,780
Total waste in 18 months (Jan 2024 - June 2025): £47,293
Clawed back through immediate changes: £41,280 (18 months × £2,293 reduced rate had they started in Jan 2024)
Ongoing annual savings: £54,780
"We effectively gave ourselves a £55,000 annual raise by firing our cloud consultant and actually looking at what we were paying for," Sarah said.
The Lessons (That Should Infuriate You)
Lesson 1: Consultants Have No Skin in the Game
The cloud consultant configured everything to "best practices" with zero regard for cost because he wasn't paying the bill.
Enterprise-grade multi-region redundancy? Great for Fortune 500. Stupid for a 23-person agency in Nottingham.
The fix: Any consultant engagement should include 6-month cost accountability. If your bills exceed their projections by >20%, they refund the difference.
Lesson 2: Vendor Defaults Are Designed to Maximize Revenue
AWS auto-scaling defaults to conservative scale-down because they make more money when you run more instances.
Microsoft pushes E3 licenses because it's 5× the revenue of Business Basic.
SaaS tools offer "unlimited" features because nobody uses them but everyone pays for them.
The fix: Question every default. Configure for your business, not vendor revenue targets.
Lesson 3: Cloud Complexity Is a Feature, Not a Bug
The more complex the infrastructure, the less likely you are to understand what you're paying for.
AWS has 200+ services with pricing models that require a PhD in economics to understand. That's not accidental. Complexity obscures cost.
The fix: Simplify ruthlessly. If you can't explain why you need a service, you don't need it.
Lesson 4: SMBs Don't Need "Enterprise-Grade" Anything
CloudBridge served regional UK clients with zero SLA requirements for sub-second response times.
They didn't need:
Multi-region redundancy
Enterprise compliance tools
Advanced threat protection
99.99% uptime guarantees
They needed:
Basic compute and storage
Regular backups
Sensible security
99% uptime (which is 3.65 days of downtime per year, perfectly acceptable)
The fix: Define "good enough" for your business. Ignore what enterprises do.
Lesson 5: The Real Cost Is Vigilance
James (the Cloud Cost Czar) now spends 2 hours per week on cloud cost management:
30 minutes reviewing AWS bills
30 minutes checking SaaS usage
30 minutes hunting zombie resources
30 minutes optimizing based on usage patterns
2 hours per week × 52 weeks = 104 hours annually
At his salary, that's roughly £3,000 in labor cost annually.
To save £54,780 annually.
ROI: 1,726%
Why don't more businesses do this?
Because nobody's assigned to do it. Cloud costs are "IT's problem" and IT assumes finance is watching the bills.
The Nuclear Connection
This week we covered how Microsoft, Google, and Amazon are restarting nuclear reactors and building new ones to power their AI ambitions.
CloudBridge's story shows the other side of that coin: UK SMBs funding Silicon Valley's nuclear renaissance through infrastructure waste.
Microsoft didn't restart Three Mile Island for fun. They're doing it because their AI workloads consume gigawatts of power, and they're passing those infrastructure costs through to customers.
CloudBridge was paying premium cloud pricing while Microsoft uses that revenue to buy nuclear reactors.
FinalSpark's biocomputing platform uses one millionth the energy of silicon processors. But that technology is years away from commercial deployment.
Until then, vigilance is your only defence against becoming an involuntary nuclear power investor.
What CloudBridge Is Doing Now (November 2025)
Monthly cloud spend: £2,650 Monthly SaaS spend: £580 (down from £2,162) Total monthly infrastructure cost: £3,230
Services they're running:
AWS production environment (optimized)
Microsoft 365 Business Basic (appropriate licensing)
Asana (actually used by everyone)
Notion (knowledge base)
GitHub (code repository)
Services they killed:
Monday.com (redundant with Asana)
Mixpanel (Google Analytics does enough)
Intercom (email works fine)
HubSpot Marketing (overkill for their needs)
Miro (Figma has built-in collaboration)
Multiple redundant AWS environments
Client relationships: Stronger. They're now transparent about infrastructure costs, billing clients appropriately, and using savings to reinvest in better development tools.
Profit margins: Healthier. That £55K annual savings went straight to bottom line, funding bonuses and a new junior developer position.
Sarah's advice to other UK SMBs:
"Treat cloud spending like you treat any other vendor. Question everything. Audit regularly. Don't assume technology companies have your interests at heart. They don't. They have shareholders."
Your Turn: The 30-Minute CloudBridge Audit
Want to see if you're making the same mistakes?
Spend 30 minutes this afternoon:
Pull your AWS/Azure bill (actual usage, not estimates)
List every SaaS subscription (include trials you forgot to cancel)
Count resources running 24/7 (how many do you actually need?)
Check Microsoft 365 licensing (are you paying for features you don't use?)
Review auto-scaling settings (are minimums set to vendor defaults?)
If your cloud bill has grown >15% in the last 12 months without corresponding revenue growth, you're probably wasting money.
Not maybe. Probably.
CloudBridge wasted £47,293 over 18 months before Sarah's accountant asked one simple question.
What question haven't you asked about your cloud spending?
| Source | Reference |
| AWS Pricing | EC2, RDS, and Data Transfer Pricing Documentation |
| Microsoft 365 | Business Plans Comparison and Pricing |
| Flexera | State of the Cloud Report 2025 (SMB Waste Statistics) |
| Gartner | Cloud Financial Management Report 2025 |