04. Discovery question bank

Ask, then shut up. Write the numbers on the board. Every number in your TCO must come from their mouth.

Format: question, then why you ask in italics. Codes let you reference questions in call notes (Q-GEN-03).

General (Q-GEN)

1. Q-GEN-01: What prompted you to look at storage now? Trigger event. No trigger, no deal.
2. Q-GEN-02: What happens if you do nothing for twelve months? Cost of inaction.
3. Q-GEN-03: How much data do you store in object storage today, and how fast is it growing? Sizing; growth rate drives TCO.
4. Q-GEN-04: Where does it live: which clouds, regions, on-prem systems? Map the estate.
5. Q-GEN-05: Who signs, who influences, who can block? Decision process.
6. Q-GEN-06: Is there a budget line, and when does the fiscal year turn? Timing.
7. Q-GEN-07: What does success look like in 90 days? Defines POC criteria.
8. Q-GEN-08: Which vendors are you also evaluating? Picks the battlecard.
9. Q-GEN-09: Have you tried an alternative cloud before? What happened? Uncovers scar tissue.
10. Q-GEN-10: Any contractual commitments elsewhere (AWS EDP, Azure MACC, GCP commit)? Commit drawdown is a real blocker.

Economics (Q-ECON)

11. Q-ECON-01: What was last month’s storage bill, split into storage, requests, retrieval, egress? Egress share is the wedge.
12. Q-ECON-02: How much data leaves the cloud per month, and to where? Egress destination decides free vs paid.
13. Q-ECON-03: Which storage classes do you use, and do you ever pay early-deletion or retrieval fees? Tiering pain.
14. Q-ECON-04: Average object size and object count? Request fees and small-object penalties elsewhere.
15. Q-ECON-05: Does finance prefer opex monthly or a prepaid term? Pay-as-you-go vs 1 to 5 year commit.
16. Q-ECON-06: What did the last restore or data pull cost you? Makes egress pain concrete.
17. Q-ECON-07: What does a TB cost fully loaded on-prem: hardware, support, power, space, people? For on-prem displacement.
18. Q-ECON-08: When is your next hardware refresh or support renewal? Timing anchor.
19. Q-ECON-09: Do you have credits from a hyperscaler, and when do they run out? Startup trigger.
20. Q-ECON-10: Would a migration subsidy change your timeline? Universal Data Migration lever.

Backup and ransomware (Q-BKP)

21. Q-BKP-01: Which backup software and versions? Integration fit (Veeam, Rubrik, Commvault, etc.).
22. Q-BKP-02: Where do copies live today: primary, secondary, off-site? Map 3-2-1.
23. Q-BKP-03: Is any copy immutable? How is that enforced? Object Lock opening.
24. Q-BKP-04: What does your cyber insurer require? Insurance checklist is a forcing function.
25. Q-BKP-05: What are your RPO and RTO for tier-1 apps? Always-hot vs Glacier.
26. Q-BKP-06: When did you last test a full restore from off-site? How long did it take? Usually never. That is the pain.
27. Q-BKP-07: Retention policy: daily, weekly, monthly, yearly, how long? Sizing plus churn plus lock duration.
28. Q-BKP-08: Who can delete backups today? Could a compromised admin account delete them? Compliance mode value.
29. Q-BKP-09: Do you back up Microsoft 365 or Kubernetes? Veeam M365, Kasten.
30. Q-BKP-10: Is tape still in the picture? Who manages it? LTO replacement.
31. Q-BKP-11: Do you use a scale-out repository or direct-to-object? Veeam SOBR capacity tier vs direct.
32. Q-BKP-12: What dedupe ratio do you see? Translates logical to stored TB.
33. Q-BKP-13: Do you need copies in two geographies? Cloud Replication across regions.
34. Q-BKP-14: How many sites and endpoints? Computer Backup cross-sell.
35. Q-BKP-15: Have you been hit, and what did recovery look like? Listen. Do not pitch for a full minute.

Media (Q-MED)

36. Q-MED-01: What cameras and codecs? Average file size? RED/BRAW/MXF large-object fit.
37. Q-MED-02: How many hours or TB per week of new footage? Ingest sizing.
38. Q-MED-03: Which MAM/PAM/NLE do you use? iconik, Mimir, axle ai, Premiere, Resolve, Avid.
39. Q-MED-04: How does archive get pulled back for reuse? How often? Egress pattern.
40. Q-MED-05: LTO generation and number of tapes? Migration scope.
41. Q-MED-06: Do remote editors or freelancers need access? Presigned URLs, Suite Studios, LucidLink.
42. Q-MED-07: Do you distribute content to viewers? Through which CDN? Partner egress.
43. Q-MED-08: Any TPN or studio security requirements? TPN Blue Shield.
44. Q-MED-09: Do you need to start editing while uploads are in progress? Live Read.
45. Q-MED-10: Who owns the archive budget: production, IT, or library? Buyer.

AI and ML (Q-AI)

46. Q-AI-01: Where are your GPUs: which clouds, regions, on-prem? Partner egress and latency.
47. Q-AI-02: How big is the training corpus, and how fast is it growing? Capacity tier sizing.
48. Q-AI-03: How many times per month do you read the full dataset? Egress multiple; 3x threshold; Overdrive trigger.
49. Q-AI-04: Target throughput per node and aggregate? Overdrive tier (100 Gbps to 1 Tbps).
50. Q-AI-05: Do you have a hot tier (VAST, WEKA, Lustre, local NVMe)? Position B2 as capacity tier beneath it.
51. Q-AI-06: How do dataloaders access data: S3 API, FUSE mount, prefetch to local? Compatibility test.
52. Q-AI-07: Checkpoint size and frequency? Write bursts and retention.
53. Q-AI-08: Do you version datasets or models? With what? Object versioning, Pixeltable, artifact stores.
54. Q-AI-09: Do you move data between providers to chase GPU availability? The one-namespace story.
55. Q-AI-10: Any data residency limits on training data? US East/West, EU Central, CA East.
56. Q-AI-11: Are agents writing outputs to storage? MCP server.
57. Q-AI-12: What does an idle GPU-hour cost you? Turns throughput into dollars.
58. Q-AI-13: Inference: where do model weights load from, and how often? Cold start reads.
59. Q-AI-14: RAG: where do source documents and embeddings live? Object as source of truth.
60. Q-AI-15: Who benchmarked storage last time, and what was the test? Reuse their methodology.

CDN and application (Q-APP)

61. Q-APP-01: What CDN do you use, and what is the origin today? Bandwidth Alliance fit.
62. Q-APP-02: Cache hit ratio? Origin egress = misses.
63. Q-APP-03: Requests per second to origin, read and write? Transaction model.
64. Q-APP-04: Public or private content? Signed URLs? Cloudflare private bucket pattern.
65. Q-APP-05: Do you use S3 features like IAM roles, object tagging, website hosting, ACLs per object? Known gaps. Find them now, not in the POC.
66. Q-APP-06: SDK and language? boto3, aws-sdk-js, Go, etc.
67. Q-APP-07: Do you need event triggers on upload? Event Notifications (webhooks; request access).
68. Q-APP-08: Multi-tenant? One key per customer? App keys, up to 100M per account.
69. Q-APP-09: Latency targets for first byte? Region choice and CDN.
70. Q-APP-10: Do you serve users outside North America and EU? Region plus CDN design.

Migration (Q-MIG)

71. Q-MIG-01: Total TB and object count to move? Method choice.
72. Q-MIG-02: Available bandwidth and windows? Network vs Fireball (96 TB Synology).
73. Q-MIG-03: Source egress cost if moved over the wire? Universal Data Migration subsidy.
74. Q-MIG-04: Can apps tolerate dual-write for a period? Cutover strategy.
75. Q-MIG-05: Integrity requirements? MD5 verification per UDM FAQ.
76. Q-MIG-06: Who owns the cutover runbook? Resource reality.
77. Q-MIG-07: Any objects in Glacier/Archive tiers needing restore first? Hidden time and cost.
78. Q-MIG-08: Do bucket and object names follow S3 rules that B2 also accepts? Naming restrictions differ slightly.

Security and compliance (Q-SEC)

79. Q-SEC-01: Which frameworks apply: HIPAA, SOC 2, PCI, GDPR, CJIS, state RAMP? Map to the compliance page.
80. Q-SEC-02: Do you need a BAA or DPA? Available.
81. Q-SEC-03: Who manages keys? Need customer-managed keys? SSE-C.
82. Q-SEC-04: SSO for the console? Enterprise web console and OIDC for Computer Backup Enterprise Control [verify B2 console SSO options].
83. Q-SEC-05: Do you require private connectivity? Overdrive private networking.
84. Q-SEC-06: Log retention and audit needs? Bucket access logs.
85. Q-SEC-07: Data residency? Region selected at account creation, cannot change.
86. Q-SEC-08: Vendor questionnaire format: SIG, CAIQ, HECVAT? Whistic profiles.

Persona closers

• CIO: “If we prove the TCO and the restore time in 30 days, what stops you from signing?”
• Admin: “If I set this up with you in 20 minutes today, will you run your next restore test against it?”
• ML lead: “If we hit your GB/s target from your GPU cloud, is storage solved for this training run?”
• MSP: “If unit cost and restore egress are fixed, which client do you move first?”

05. Qualification: the STORE framework

A POC is a decision instrument, not a science project. Qualify before you build.

Use MEDDPICC if your team does. STORE is the storage-specific overlay that tells you whether Backblaze is the right answer.

S: Size and shape

• Stored TB today and in 36 months.
• Object count and average size.
• Read multiple per month (egress / stored).
• Green: 50 TB+, objects over 1 MB average, egress under 3x stored or flowing to partners.
• Yellow: tiny objects with very high request rates, or egress 3x to 5x stored to the internet.
• Red: under 5 TB with no growth, or a hard requirement for sub-millisecond latency.

T: Trigger and timeline

• Named event: renewal, incident, insurance, credits ending, refresh, migration.
• Date that forces a decision.
• Red: “just benchmarking prices.”

O: Owner and organization

• Economic buyer identified and met.
• Technical champion who will run the POC.
• Who blocks: security, procurement, an AWS commit owner.

R: Requirements fit

Check every hard requirement against documented capability:

E: Economics proven

• TCO built with their bill, not list estimates.
• Commit structure agreed in principle (PAYG, 1 to 5 year term, Overdrive multi-PB).
• Migration funding discussed (Universal Data Migration with qualifying commitment).

Red flags

• They need a feature listed as unsupported and will not change the app.
• Their egress goes to the open internet at 5x+ stored with no CDN (R2 may genuinely be cheaper; see 15).
• A large hyperscaler commit they must burn down this year.
• No champion willing to run a test.

Deal sizing quick math

Annual value ≈ stored_TB × $/TB/month × 12

• 500 TB at list = 500 × 6.95 × 12 = $41,700/yr.
• 2 PB at list = 2,000 × 6.95 × 12 = $166,800/yr.
• 5 PB Overdrive at $15 list = 5,000 × 15 × 12 = $900,000/yr. Note: deals over $500K sit outside public guidance, which is why they matter.

Exit criteria for each stage

20. Objection bank (60)

Format: They say / You say / Prove it. Acknowledge first, ask second, answer third. Never argue with the customer’s experience.

Price and value

1. “Wasabi is basically the same price.” You say: “On storage, close: $7.99 vs $6.95. The difference is what happens when you read. B2 includes 3x your stored data in free egress every month, unlimited through partners, and no minimum storage duration.” Prove: 14-tco table at their read multiple and deletion pattern.

2. “Glacier is way cheaper.” “It is, for data you never touch. What is your restore SLA, and what did the last retrieval cost?” Prove: restore scenario including retrieval fee, hours, egress.

3. “Your price just went up from $6 to $6.95.” “It did. Wasabi went up too. That is why term commits of 1 to 5 years lock the rate.” Prove: term option on quote.

4. “R2 has zero egress.” “True. R2 storage is $15/TB, more than double, and writes cost $4.50 per million. If you use Cloudflare, put B2 behind it and get $0 egress at $6.95.” Prove: 1 PB comparison.

5. “We get big discounts from AWS.” “Great, let’s use your effective rate. What is it on storage and egress?” Prove: rerun TCO with their numbers.

6. “$0.01/GB overage could surprise us.” “Overage only starts above 3x stored. At 1 PB you get 3 PB a month free. If you will exceed that, Overdrive gives unlimited egress.” Prove: break-even 3.8x.

7. “We cannot prepay.” “Pay-as-you-go monthly by card or invoice. No commitment.” Prove: pricing page.

8. “Savings are not worth the migration effort.” “Universal Data Migration offsets migration cost on a qualifying commitment and Backblaze runs it. What is your team’s cost per month of delay?” Prove: UDM page.

9. “Our storage bill is small.” “Then let’s find the line that hurts: restores, data movement, or the next refresh.” Disqualify politely if nothing hurts.

10. “Free egress is a gimmick; you will change it.” “It is published on the pricing page with a $0.01 overage. Term contracts lock pricing.” Prove: contract terms.

Technical fit

11. “Is it really S3 compatible?” “Core APIs yes, with documented gaps: no object ACLs, IAM roles, object tagging, website config, or browser POST. Which do you use?” Prove: run their app in POC.

12. “We use IAM roles.” “Not in the S3-compatible API. Use scoped app keys distributed through your secrets manager.” Prove: key scoping demo.

13. “We rely on object tags.” “Tags are not supported. Use prefixes or a metadata index. Many apps only need tags for lifecycle, which B2 handles by prefix.” Prove: lifecycle rule demo.

14. “Is it fast enough?” “For which workload? Hume AI hit about 10 GB/s uploads. Overdrive goes 100 Gbps to 1 Tbps dedicated. Let’s measure from your network.” Prove: throughput test.

15. “Latency for small files?” “B2 is optimized for throughput and capacity. For sub-millisecond small I/O, keep a hot tier.” Prove: honest architecture.

16. “Single region per account is a limitation.” “It is a design choice that keeps residency explicit. Cloud Replication covers multi-region with a second account.” Prove: replication setup.

17. “No storage classes? We want tiering.” “One hot class at a price below most cool tiers. No retrieval fees, no minimum durations. Tiering is complexity you no longer need.” Prove: price compare vs S3 IA ($12.50/TB) and Azure Cool.

18. “Does it work with our backup software?” “Veeam, Rubrik, Commvault, MSP360, Synology, TrueNAS and hundreds more have guides. Rubrik needs SigV4 enabled by their support.” Prove: integration guide.

19. “What about Kubernetes?” “Kasten K10 and Velero-style S3 targets work; Object Lock with Kasten is documented on the Veeam page.” Prove: test.

20. “Can we mount it like a drive?” “Via tools like Suite Studios, LucidLink, rclone mount, Mountain Duck. Object storage is not a POSIX filesystem; know the trade-offs.” Prove: Suite Studios integration.

21. “Event triggers?” “Event Notifications send signed webhooks on create, update, delete. Access is enabled by Support.” Prove: demo 6.

22. “We need versioning.” “B2 keeps versions by default. Lifecycle rules control how long.” Prove: console.

23. “Can we host a static site?” “Not via S3 website config. Put Cloudflare or Fastly in front, with $0 egress.” Prove: CDN guide.

24. “Our app uses SigV2.” “B2 requires SigV4. Most SDKs default to V4; update the client.” Prove: SDK config.

25. “Max object size?” “Large files via multipart. [verify current max object and part size in docs before quoting]”

Trust, risk, and company

26. “Backblaze is a consumer backup brand.” “B2 is 62 percent of revenue, 100,000+ customers, and CoreWeave signed a $335M multi-exabyte deal.” Prove: Q2 2026 results, 8-K.

27. “Will you be around?” “Public company, adj. EBITDA margin 30 percent in Q2 2026, ARR $177M growing 21 percent.” Prove: IR.

28. “Your drive failure rate went up.” “Yes, Q2 2026 AFR was 1.73 percent, and Backblaze published it. Durability comes from 17+3 erasure coding across 20 pods, not from drives never failing.” Prove: Drive Stats, durability doc.

29. “You are too small for our scale.” “Over an exabyte under management; Decart went 0 to 16 PB in 90 days.” Prove: case study.

30. “Support quality?” “Premium support add-on; Overdrive includes the highest tier; customers like Mirage cite ‘real human support’.” Prove: references.

31. “What if you get acquired?” “Contract terms and S3 compatibility protect you. Your data is portable with 3x free monthly egress.” Prove: egress math for exit.

32. “Hard drives are old tech.” “HDD is how the industry stores exabytes economically. Even CoreWeave uses Backblaze for its HDD tiers under a flash cache.” Prove: CoreWeave announcement.

33. “We had a bad experience with an alt-cloud.” “What broke? Let’s put that exact test in the POC.” Prove: POC criterion.

34. “Your SLA?” “[pull current SLA]. Durability designed at 11 nines.”

Security and compliance

35. “Are you HIPAA compliant?” “We sign BAAs for covered entities. HIPAA compliance is shared.” Prove: compliance page.

36. “FedRAMP?” “No. GovRAMP Progressing and TX-RAMP Provisional.” Prove: compliance page. Qualify out if hard requirement.

37. “Is data encrypted?” “SSE-B2 AES-256 by default on new uploads since Sep 14 2026; SSE-C if you hold keys.” Prove: blog.

38. “Can your staff delete our backups?” “Not in compliance mode, not before expiry.” Prove: demo 3.

39. “We need customer-managed KMS.” “SSE-C today means the client supplies the key per request. Native KMS integration is not documented.” Prove: SSE docs. Escalate roadmap ask.

40. “Do you have SOC 2?” “SOC 2 Type 2; report via Whistic.” Prove: Whistic.

41. “GDPR residency?” “EU Central in Amsterdam; DPA available; data stays in the selected region.” Prove: regions doc.

42. “Is it air-gapped?” “Logically: off-site, separate credentials, immutable. Not a physical gap.” Prove: architecture.

43. “What about insider threats?” “Scoped keys, compliance-mode lock, MFA, access logs.” Prove: key demo.

Organization and process

44. “We have an AWS commit to burn.” “Then start with the copy that must sit outside AWS: the immutable backup. It does not touch your commit’s core workloads.” Prove: backup design.

45. “Procurement takes six months.” “Start PAYG on a card for the POC; paper the term deal in parallel.” Prove: PAYG.

46. “Security review will take forever.” “Whistic profiles have SOC 2, HECVAT, and questionnaires ready.” Prove: Whistic.

47. “We do not have staff for a migration.” “UDM is white-glove.” Prove: UDM.

48. “Our MSP decides.” “Let’s bring them in; Backblaze has an MSP program and Partner API.” Prove: partner page.

49. “We are standardizing on one cloud.” “For compute, sensible. For the backup copy, standardizing on the same provider puts the copy in the same blast radius.” Prove: 3-2-1-1-0.

50. “Not a priority this quarter.” “What changes next quarter? Insurance renewal, refresh, credits ending?” Prove: trigger.

AI specific

51. “We need flash for training.” “For the working set, yes. B2 holds the full corpus and checkpoints and refills the flash tier.” Prove: whiteboard 11.

52. “Our GPUs are in a hyperscaler.” “Then model egress from B2 at 3x free or Overdrive unlimited, versus hyperscaler storage price.” Prove: TCO scenario.

53. “Neoclouds can build storage with Ceph.” “They can. Then they run a storage company. B2 Neo launches in weeks.” Prove: white paper.

54. “Is CoreWeave a B2 Neo customer?” “CoreWeave signed a multi-exabyte storage agreement; Backblaze powers HDD tiers in CoreWeave AI Object Storage. The SEC filing does not label it B2 Neo.” Prove: 8-K wording.

55. “Can agents safely use storage?” “B2 MCP Server exposes only tools the key permits.” Prove: demo 7.

56. “Throughput claims are marketing.” “Agreed. Test from your GPU nodes with your shards.” Prove: demo 8.

Media specific

57. “Our MAM uses tags.” “Test it. Most MAMs keep metadata in their own DB.” Prove: POC.

58. “Tape is cheaper per TB.” “Media is. Add library maintenance, LTO migrations, off-site handling, restore delays. Eagles retired LTO-6.” Prove: case study.

59. “Uploads of 8K are too slow.” “Size the link; use Fireball for backlog; Live Read starts editing while uploading.” Prove: transfer table.

60. “Studios require TPN.” “Backblaze has TPN Blue Shield status.” Prove: compliance page.

24. Field lab: drills, flashcards, quiz

Fifteen minutes to get sharp. Do one scenario drill out loud every morning of your first month.

Scenario drills

Each drill: read the setup, set a 10-minute timer, draw the board, say the first three questions, name the risk, name the proof.

Drill 1: The insurer letter

Setup: 400-person manufacturer, Veeam v12, 180 TB of backups on a local linux repo, cyber insurer renewal in 60 days demands an immutable off-site copy. Board: whiteboard 09. First questions: retention schedule? RTO for ERP? who has admin on the Veeam server? Risk: Object Lock not enabled before connecting Veeam. Proof: 20-minute setup and failed delete demo. Numbers: 180 × 6.95 = $1,251/month list.

Drill 2: The egress shock

Setup: SaaS video platform, 600 TB in S3 Standard, 400 TB/month egress to Fastly. Ask for the bill, then check it with the math below. Math check: storage 50,000×0.023 + 450,000×0.022 + 100,000×0.021 = 1,150 + 9,900 + 2,100 = $13,150. Egress 399,900 GB: 900 + 3,400 + 7,000 + 249,900×0.05 = 12,495 → $23,795. Total ≈ $36,945. B2: 600 × 6.95 = $4,170; egress to Fastly $0. Risk: app uses object tagging for moderation flags. Proof: POC with their SDK; tag dependency replaced by DB column.

Drill 3: The GPU nomad

Setup: AI lab, 3 PB corpus in GCS, renting GPUs at CoreWeave this quarter and maybe Lambda next. Reads full corpus ~2x/month. Board: whiteboard 11. Answer: B2 standard: 3,000 × 6.95 = $20,850/month; 2x read to CoreWeave = $0 partner egress. To non-partner at 2x: still inside 3x free. Risk: dataloader uses GCS-specific client. Proof: throughput test from CoreWeave nodes.

Drill 4: The tape library

Setup: regional broadcaster, 1,200 LTO-5 tapes (~1.5 TB each native ≈ 1.8 PB), library end of support, archive pulled weekly for retrospectives. Answer: UDM tape migration with commit; B2 library + MAM; delivery through Cloudflare. Risk: tape read failures; budget for unreadable media. Proof: Eagles and ACL references.

Drill 5: The Wasabi renewal

Setup: MSP with 900 TB on Wasabi, renewal at $7.99, clients restore ~1.2x stored per month in aggregate during incidents. Answer: 3x free egress vs fair-use; no minimum duration for short-retention clients; Partner API. Math: $7,191 vs $6,255/month at list. Risk: migration of backup chains; plan re-seed vs copy.

Drill 6: The neocloud founder

Setup: 8,000-GPU neocloud, customers keep data in AWS, founder considering Ceph. Board: whiteboard 11, B2 Neo variant. First questions: customers’ data location? storage headcount? billing integration? Proof: Partner API demo, custom domain, CoreWeave reference pattern (stated precisely).

Drill 7: The Azure shop

Setup: hospital system, all Microsoft, M365 backups to Azure Blob in the same tenant. Answer: separate-blast-radius copy on B2 via Veeam M365, BAA, compliance mode. Risk: MACC commitment politics. Proof: BAA, SOC 2, failed delete demo.

Drill 8: The R2 developer

Setup: startup storing 40 TB user images on R2, 300M Class A ops/month (thumbnails), Cloudflare CDN. Math: R2: 40×15 = $600 + 300×4.50 = $1,350 → $1,950+. B2 behind Cloudflare: 40×6.95 = $278, egress $0, ops free. Risk: Workers bindings to R2 need rework to S3 calls.

Flashcards (front / back)

1. B2 list price? / $6.95 per TB per month, billed per byte-hour.
2. Free egress allowance? / 3x average monthly stored; then $0.01/GB.
3. Partner free-egress list? / Fastly, Cloudflare, bunny.net, CacheFly, CoreWeave, Equinix Metal, Vultr, phoenixNAP.
4. Overdrive starting price? / $15/TB/month, multi-PB commit.
5. Overdrive throughput? / 100 Gbps to 1 Tbps.
6. Overdrive egress? / Unlimited free, any destination.
7. Erasure coding scheme? / 17 data + 3 parity across 20 pods.
8. Designed durability? / 11 nines annually.
9. Regions? / US West, US East, EU Central, CA East.
10. Can you change an account’s region? / No.
11. Object Lock retention range? / 1 to 3,000 days.
12. Can Object Lock be disabled on a bucket? / No.
13. Compliance vs governance? / Compliance cannot be shortened or removed; governance can with privileged key.
14. Legal hold expiry? / None until removed.
15. Lifecycle rules per bucket? / Up to 100.
16. Lifecycle rule fields? / fileNamePrefix, daysFromUploadingToHiding, daysFromHidingToDeleting, daysFromStartingToCancelingUnfinishedLargeFiles.
17. App keys per account? / Up to 100 million.
18. Default encryption? / SSE-B2 AES-256 on new uploads since Sep 14 2026.
19. SSE-C? / Customer supplies AES-256 key per request.
20. S3 API gaps? / Object ACLs, IAM roles, tagging, website config, browser POST.
21. Signature version? / SigV4 only.
22. Rubrik gotcha? / Defaults to SigV2; support must enable V4.
23. Veeam lifecycle setting? / Keep all versions.
24. Class D cost? / 2,500/day free then $0.004 per 10,000.
25. Cross-region replication requirement? / Second account in target region.
26. Can replicated data be replicated again? / No.
27. Existing-file replication timing? / Daily at midnight UTC.
28. Fireball capacity? / 96 TB Synology NAS, 10GbE RJ-45.
29. UDM integrity method? / MD5 byte-for-byte verification.
30. MCP server tool count? / 40 (17 native, 19 S3, 4 analytics); read-only profile 20.
31. Computer Backup price? / $99/year per computer.
32. HIPAA answer? / BAA available for covered entities.
33. Gov programs? / GovRAMP Progressing Snapshot, TX-RAMP Provisional.
34. CoreWeave deal size? / ~$335M, multi-exabyte, five-year (5 and 7 year order forms).
35. Q2 2026 B2 growth? / +34% to $26.6M.
36. ARR Q2 2026? / $177.3M, +21%.
37. Overdrive break-even read multiple? / ~3.8x stored per month.
38. Q2 2026 Drive Stats AFR? / 1.73%.
39. Free storage? / First 10 GB.
40. Minimum storage duration? / None.

Quiz (answers at bottom)

1. A customer stores 200 TB and egresses 700 TB/month to the internet. B2 monthly list?
2. Which mode would you use for an insurer-mandated backup copy?
3. A customer needs survival of a full region loss. What do you design?
4. Which S3 call will be rejected if it includes x-amz-tagging?
5. Name two things Fireball pricing includes.
6. Why should you never say “unlimited free egress” for standard B2?
7. What is the first thing you do in a Rubrik POC?
8. At what read multiple does Overdrive beat standard B2 on list price?
9. What happens to existing files when SSE-B2 default is enabled?
10. What does the MCP server do with a read-only key?

Answers: 1) 200×6.95 = $1,390 + (700−600) TB × $10 = $1,000 → $2,390. 2) Compliance mode. 3) Cloud Replication to a bucket in a second account in another region. 4) PutObject or CopyObject. 5) Security deposit, non-refundable shipping fee, per-30-day rental. 6) It is 3x free then $0.01/GB; unlimited applies to partners and Overdrive. 7) Open a Rubrik support case to enable SigV4. 8) ~3.8x. 9) Not encrypted retroactively; only new uploads. 10) Exposes only read tools (read-only profile, 20 tools max).

31. Call plans

Every meeting has one job. Name it at the start, check it at the end.

A. First discovery call (30 minutes)

Job: decide whether there is a qualified storage problem worth a whiteboard.

B. Technical whiteboard (60 minutes)

Job: agree on a target architecture and POC criteria.

C. Executive briefing (30 minutes, CIO/CFO)

Job: get sponsorship for the POC and commercial path.

1. Their situation in three bullets (2 min).
2. The economics: one table, their numbers, sources dated (8 min).
3. Risk controls: immutability, compliance, migration support, company stability (8 min).
4. Proof: one reference in their industry (4 min).
5. Ask: approve POC and identify the commercial path (PAYG vs term) (8 min).

D. POC kickoff (45 minutes)

Job: zero ambiguity on scope.

• Confirm environment, data set, keys, endpoint, region.
• Walk the scorecard (17).
• Schedule check-ins (day 3, day 7, mid-point, final).
• Name the decision meeting date.

E. POC readout (30 minutes)

Job: convert results into a decision.

• Scorecard criterion by criterion.
• TCO refreshed with measured numbers.
• Proposed commercial structure.
• Ask for the decision or the specific blocker.

F. Partner QBR (45 minutes)

Job: pipeline and enablement.

• Joint wins and losses; what broke in integration.
• Top 10 overlapping accounts.
• One whiteboard enablement (09 or 12).
• Next quarter targets.

G. Neocloud architecture session (90 minutes)

Job: decide build vs buy.

1. Their GPU footprint and customer data locations.
2. Draw the split-stack problem: customer data in hyperscaler, compute on neocloud, egress in between.
3. Build-it-yourself cost model: hardware, Ceph/MinIO ops, on-call, drive replacement, upgrades.
4. B2 Neo: Partner API flow, custom domain, SSO, billing feed.
5. Time to launch and commercial model.

Call hygiene

• Notes in the CRM within one hour, with Q-codes.
• Every number tagged “customer said” or “we estimated”.
• Next step has a date, or it is not a next step.