Summary

  • Standard B2 free egress is tied to three times average monthly storage, calculated from byte hours over the billing cycle—not a current, peak or month-end stock.
  • Reducing genuinely retained bytes lowers their future contribution to stored time. It can reduce the allowance with other conditions unchanged, but does not establish a higher total invoice or justify retaining unneeded data.
  • Partner routes and B2 Overdrive have different free-egress scopes. Lifecycle hiding and deletion are distinct, and the 2023 policy FAQ says unused allowance does not roll over.

A stock is not a month of storage

A team can see a large dataset in a bucket and read Backblaze B2’s three-times free-egress headline as permission to send three copies of that dataset anywhere without an outbound charge. The current pricing explanation is more specific. The allowance follows average monthly storage, measured from byte hours over the billing cycle.

The difference matters even when the headline is generous and accurately advertised. A dataset arriving late in a cycle contributes less stored time than the same dataset present throughout that cycle. A current inventory can show the same amount of data in both cases while the time-weighted average is different.

This is not evidence of a hidden penalty for moving data. It is the denominator of the offer. The buyer needs to distinguish the volume present at a moment from the storage service consumed across time. Neither the highest observed stock nor the final inventory is a substitute for that average.

Backblaze’s current developer page restates the byte-hour basis. Its pricing page separately explains pay-as-you-go storage billing in byte hours. Storage and the standard outbound allowance therefore refer to a temporal measure, rather than simply to the latest set of visible files.

What deletion can and cannot change

If actual retained bytes are reduced, they contribute fewer byte hours from that point onward than if they had remained. The time already accumulated earlier in the cycle does not disappear merely because the stock later falls. This is an analytical consequence of the stated time-average rule, not a measured adjustment in a customer account.

Holding other conditions constant, lower average storage can mean a smaller standard free-egress allowance. It does not mean that reducing storage necessarily increases the total bill. Less stored time can also reduce storage consumption, outbound volume can change, and an eligible route can have another charging scope.

The wrong purchasing response would be to keep unneeded data solely to enlarge the denominator. Retention has useful purposes: recovery, an application’s history and an accepted record of its work. It also has costs and risks. A commercial allowance cannot by itself decide which versions should remain useful.

The relevant comparison is a joint trajectory. How much data actually remains, for how long, and how much non-exempt traffic leaves during the applicable cycle? An end-of-month screenshot answers only one part of that question. A smaller stock is neither proof of economy nor proof of an outbound-cost problem.

No bucket, deletion, download or supplier account was tested for this article. No customer invoice or saving was calculated. The mechanism is the documented relationship between stored time and allowance, with its scope kept visible.

A monthly allowance is not a reserve of unused exits

Backblaze introduced the broader three-times policy in its 2023 product announcement. That announcement’s FAQ says unused free egress does not roll over; it is calculated independently for each month. The current pricing page confirms the monthly time-average basis.

The no-rollover statement belongs to that published FAQ, not to an account implementation inspected here. The older announcement is also not a source for current storage prices. Its historical rates should not be copied into a present-day comparison merely because the explanation of byte hours remains relevant.

The commercial implication is still important. A long quiet history is not, under the FAQ’s described policy, a bank of unused monthly allowances available for a future large transfer. A business may have consumed little outbound traffic in earlier periods and still need to assess the cycle in which its new downloads occur.

Likewise, the three-times relationship is not a guarantee of three complete restores of the current dataset. Restore volume, cycle-average storage and destination scope can differ. Backblaze’s backup and archive page presents free egress as a way to make testing and recovery more accessible. That is a material benefit; it is not a measurement of a buyer’s recovery workload or assurance that every planned transfer is covered.

Identify the route before applying the ratio

The standard relationship is not the only offer. Backblaze’s current pricing page describes unlimited free egress when downloads go to or through named CDN and compute partners. It separately presents B2 Overdrive with unlimited egress to any destination and a different commitment and service scope.

Those alternatives are counterevidence against treating the three-times denominator as a universal obstacle. They can remove the relevant outbound charge under their applicable conditions. They do not justify assuming that every CDN path, every destination or every downstream service is free.

A buyer must identify the product and the qualifying path before applying the standard ratio. A familiar partner name in a proposed architecture is not a tested traffic receipt. Nor does an exemption on Backblaze’s side certify that another service has no cost or that the download meets a business deadline.

There is no actual route test here. The article does not recommend changing a destination or making a commitment. It distinguishes charging scopes so that an ordinary standard-B2 allowance is not mixed with a partner arrangement or an Overdrive offer in the same forecast.

Visible names are not necessarily retained bytes

Retention also needs a precise object of observation. Backblaze’s lifecycle documentation separates hiding a version from deleting it. Replacing a file can hide the earlier version, and a configured lifecycle stage can delete hidden versions after a specified number of days.

A file disappearing from the latest-name view is therefore not, by itself, evidence that its older stored version has been removed. The allowance’s denominator concerns the storage measure; it cannot be inferred solely from how a listing looks to an application.

This does not establish that a lifecycle rule has failed or that a customer is paying for a particular hidden version. No rule or bucket was inspected. It means that visibility and actual retention are different claims, and a forecast should not silently replace one with the other.

The same discipline applies to apparently simple price language. Backblaze currently advertises no minimum file-size or storage-duration fees. The analysis should not import another supplier’s minimum-retention penalty into this offer. Its pay-as-you-go transaction page also distinguishes free Class A, B and C calls from Class D outbound notification charges. “Free transactions” is not a reason to claim every possible API-related activity is universally uncharged.

Accept the useful workload, not just the headline

The buyer is not obliged to distrust a favorable offer. The current policy can make active storage, testing and data movement less punitive than an arrangement in which outbound charges dominate. Its benefit is clearest when the correct measure and route are used.

The acceptance problem is to connect decisions that teams often make separately. The storage owner decides what remains; an application team schedules distribution; a network or platform team determines the path; finance reviews the cycle. Each can make a sensible local choice while the resulting trajectory differs from a forecast based on a point-in-time stock.

A useful commercial review asks which retained bytes serve the work, how long they contribute, what downloads are expected in the same billing cycle and which scope applies to their destinations. It keeps product price, storage consumption and outbound allowance separate from delivery performance and downstream charges.

That is a more durable reading of data freedom than either assuming unlimited standard egress or calling a smaller denominator a trap. A stock tells the buyer what is there. Stored time and the traffic path tell the buyer what relationship the published allowance describes.

Sources