Can Unlimited Pricing Be Profitable? Backblaze's $5 Storage Bet
Short answer: Unlimited pricing can work when most customers use modest capacity, heavy users remain a bounded minority and your marginal cost falls below the contribution from the whole cohort. Model the 90th and 99th percentile, not only average use, and fund the capacity peak before launch. If one legitimate customer can create an unbounded loss, the offer is not ready to be unlimited.
Customers like a fixed price because it removes calculation and fear. Founders like recurring revenue. The risk sits between them: the customer knows the bill, while the business accepts uncertainty about use.
Calling something unlimited does not remove limits. It makes the company responsible for engineering, pricing and explaining them honestly.
Backblaze started with the price constraint
Five founders pooled US$75,000 to launch Backblaze, according to the company's history of its storage hardware. Their intended offer was simple: unlimited computer backup for US$5 per computer each month.
Available infrastructure made that price look impossible. Co-founder Gleb Budman later wrote that Amazon S3 then cost US$0.15 per gigabyte monthly. A customer storing 33GB would consume the entire US$5 charge before data retrieval, support, engineering or marketing. He estimated that such a customer would cost at least US$10 monthly under that route.
Commercial storage systems also demanded large upfront spending and cost around US$1,000 per terabyte when raw drives were about US$100 per terabyte. Backblaze's choices were to remove unlimited use, raise the price, fund losses or change the cost structure.
It chose the fourth. A 2009 technical account priced its 67-terabyte Storage Pod at US$7,867 in materials and said that, after electricity, bandwidth, space and administration, its storage cost was one-tenth of several commercial alternatives. These are historical figures, not current purchasing guidance.
My view is that unlimited pricing should be a cost-engineering decision, not a marketing phrase. If you cannot explain why the distribution of legitimate use is affordable, do not offer it.
Use the Tail-Cost Pricing Test
The Tail-Cost Pricing Test checks the customers at the expensive end before approving a flat price.
| Layer | Calculation | Decision question | |---|---|---| | Core | Revenue less ordinary use cost | Does the typical customer contribute? | | Tail | Cost of the heaviest legitimate users | Can the cohort carry the tail? | | Timing | Cash received versus capacity purchased | Can you fund growth before payback? | | Failure | Support, replacement, abuse and recovery | Is the service still viable when things go wrong? |
Unlimited works at cohort level. A heavy customer may be loss-making while the cohort remains sound. That is acceptable only when the heavy use is legitimate, predictable in aggregate and not growing faster than price or cost improvements.
Build a usage distribution
Start with observed use from a limited paid service, pilot or comparable operation. Divide customers into bands rather than relying on one average. At minimum, calculate median, 90th percentile, 99th percentile and maximum legitimate use.
For each band, price variable infrastructure, third-party charges, support and replacement. Fixed assets still matter because you must buy enough capacity for the whole distribution. Allocate them over a cautious useful life and expected occupancy.
If you have no usage evidence, the responsible offer is metered, capped or explicitly subject to a tested allowance. “Unlimited” is not a substitute for not knowing.
Separate marginal cost from capacity cash
A service can have low long-run cost and still run out of cash. Servers, warehouse space, labour or supplier commitments may be purchased before monthly subscriptions repay them.
Build a capacity step schedule. State how many additional customers trigger the next purchase, how much it costs and how many months of contribution repay it. Include spare capacity and failure replacement rather than assuming every asset operates continuously.
The guide on [how much money you need to start a business](/business-finance/how-much-money-do-you-need/) shows why peak cash need differs from eventual profitability.
Define legitimate use without hiding a cap
An unlimited consumer backup service can still prohibit resale, automated commercial storage or use unrelated to personal backup. Those are scope rules, not secret quantity caps.
Write the intended job in plain language. Detect abuse from behaviour connected to that job, and give customers a fair review route. Slowing, cancelling or refusing heavy customers merely because they use the advertised service is not honest unlimited pricing.
Consumer, subscription, data-protection and fair-terms requirements vary by country and sector. Have qualified local professionals review claims, renewal, termination and data handling.
Worked example: VaultCloud
VaultCloud offers unlimited photo backup to UK households for £8 a month. These figures are illustrative.
It tests 1,000 paying accounts for three months. Net revenue after payment fees is £7.70 per account, or 1,000 x £7.70 = £7,700 monthly.
| Usage band | Accounts | Storage per account | Monthly storage and transfer cost | |---|---:|---:|---:| | Light | 700 | 100GB | £0.65 | | Regular | 250 | 500GB | £2.10 | | Heavy | 45 | 2TB | £6.80 | | Extreme legitimate | 5 | 8TB | £24.00 |
Monthly infrastructure cost is (700 x £0.65) + (250 x £2.10) + (45 x £6.80) + (5 x £24) = £1,406. Support costs £1,150, hardware replacement and monitoring cost £820, and customer acquisition amortised over expected life costs £1,600.
Monthly contribution is therefore £7,700 - £1,406 - £1,150 - £820 - £1,600 = £2,724, or £2.72 per account.
The five extreme accounts each lose money directly: £7.70 net revenue less £24 infrastructure cost equals negative £16.30 before support. Together they lose £81.50 on infrastructure contribution. The cohort can carry that legitimate tail.
VaultCloud then stress-tests growth. If extreme users rise from 0.5% to 3%, 30 accounts cost £720 rather than £120. Contribution falls by £600 to £2,124 but remains positive. At 10%, the model weakens materially. The company sets an operating alarm at 3%, holds three months of planned capacity spending in cash and keeps the offer genuinely unlimited within personal photo backup.
Stress-test what changes together
Heavy users may generate more support, longer restores and greater failure exposure. Do not vary storage cost while leaving every correlated cost fixed.
Run at least four cases: expected use, heavier tail, infrastructure price increase and faster customer growth. Include a combined case. A business that survives each problem separately may fail when drive prices rise during a growth surge.
Also test price sensitivity. A £1 increase across 1,000 customers creates up to £1,000 monthly revenue before tax and fees, but only if retention holds. Do not assume it will.
Make the decision in 30 days
In week one, define the job included by “unlimited” and collect the best available usage data. In week two, build the distribution and allocate all direct costs to each band.
During week three, model capacity purchases, failures and the heavy-use stress case. Write scope and cancellation language that matches the service customers will actually receive.
In week four, choose among unlimited, a visible cap, usage tiers or metering. If you select unlimited, set monthly alarms for tail share, contribution per account and months of capacity cash. Reprice or redesign before the alarm becomes a crisis.
Related guides
Frequently asked questions
Does unlimited pricing depend on inactive customers?
It should not depend on customers forgetting to use the service, although low and varied use can legitimately support a pooled model. Calculate contribution with active customers and a heavier-than-expected tail. Dormant subscribers may improve current margin, but they can cancel together when they notice the charge. Track activation and value received as well as revenue. If the business fails when every customer uses the service for its advertised purpose, the offer is structurally misleading or underpriced. Fix the cost or boundary rather than hoping for neglect.
How much heavy use is too much?
The answer is the share at which cohort contribution, capacity cash or service quality breaches your pre-set floor. Model several tail proportions using observed cost per heavy account. A storage company may tolerate a small group using many times the median, while a labour-intensive service may not. Do not label ordinary high use as abuse simply because it is expensive. If heavy customers perform the promised job, change the price, cost structure or offer for future customers and honour existing legal commitments.
Should I use a fair-use policy?
Use one to define the intended customer activity and prohibit conduct such as resale, automation or unlawful use, not to hide an undisclosed numerical cap. Give concrete examples and explain investigation and appeal in plain language. A fair-use policy cannot rescue pricing that fails under legitimate use. Test whether the service remains profitable when customers use it fully within scope. Requirements for contract terms and termination vary, so obtain qualified local advice and ensure marketing claims match the actual restrictions.
Is unlimited better than usage-based pricing?
Neither is inherently better. Unlimited reduces customer calculation and makes bills predictable, while usage pricing aligns revenue more closely with cost. Choose from the customer's buying anxiety and your cost distribution. If usage is difficult to predict but cheap to supply, a flat price may differentiate the offer. If a small number of customers can generate enormous cost, metering or tiers are safer. Compare conversion, use, support and contribution in a limited test rather than copying a competitor's headline price.
When should I raise the unlimited price?
Raise it before contribution or capacity cash falls below the floor set in your model, not after the account balance forces a rushed decision. Separate temporary cost spikes from a permanent shift in usage. Test the new price with incoming customers and observe conversion and retention. Existing-customer changes require clear notice and may be constrained by contracts and consumer law. A higher price that preserves a simple, honest promise can be better than adding confusing restrictions, but only your customer and cost evidence can decide.
Can a service be unlimited if it excludes business use?
Yes, if the offer plainly says it is unlimited for a defined personal job and the restriction is prominent before purchase. The problem arises when a broad headline is narrowed by hidden terms or arbitrary enforcement. Describe who the service is for, what activity is included and which conduct is excluded. Monitor scope breaches consistently. If small-business customers are common and profitable, consider a separate plan designed around their use rather than forcing them through a consumer policy that does not fit.
Comments (0)