LEARN / FILE 03

03. ICPs and personas

Sell to the workload, then to the person. The same customer can be three ICPs.

Ideal customer profiles

Who: mid-market to enterprise IT running Veeam, Rubrik, Commvault, Cohesity, MSP360, Synology/QNAP, or Kasten.
Trigger events: ransomware incident (theirs or a peer’s), cyber insurance renewal questionnaire asking for immutable off-site copies, tape library end of life, AWS bill review, Wasabi price increase (Wasabi pay-as-you-go list moved to $7.99/TB, per wasabi.com/pricing checked 2026-10-14).
Data shape: 50 TB to multi-PB, write-heavy, read rarely except restores and restore tests. Retention-based churn.
Why B2 wins: Object Lock at no extra cost, always-hot restores, no minimum duration so retention churn does not trigger early-delete fees, free egress for restore tests up to 3x.
Deal size signal: 200 TB+ or a multi-year commit.

Who: post houses, broadcasters, sports teams, agencies, houses of worship, public TV.
Triggers: LTO generation migration, shuttle drives, remote collaboration, MAM rollout (iconik, Mimir, axle ai, CuttingRoom, Backlight), TPN audit.
Data shape: 100 TB to 10 PB, large files (RED, BRAW, MXF), bursty reads, reuse of archive.
Why B2 wins: archive stays online, free egress up to 3x, TPN Blue Shield status, Live Read for growing files, integrations with MAMs, Fireball and Universal Data Migration for tape.
Proof: Philadelphia Eagles (800 TB, retired LTO-6), Austin City Limits (4,000+ hours of tape), Bell Bank studio, American Public Television, Jazz at Lincoln Center (1.3 PB).

Who: model labs, AI startups, enterprise ML platform teams, research institutions.
Triggers: egress bills moving data to GPU clouds, credits running out, multi-cloud GPU sourcing, dataset growth past 1 PB.
Data shape: PB-scale, read-heavy in bursts, throughput-bound, multi-site.
Why B2 wins: one S3 namespace that feeds any GPU cloud, partner free egress to CoreWeave and Vultr, Overdrive dedicated throughput to 1 Tbps with unlimited egress, no upload or delete fees.
Proof: Decart (0 to 16 PB in 90 days), Hume AI (~10 GB/s uploads, consolidated multiple hyperscale stores), Mirage/Captions (Overdrive, 95% estimated savings), Segmed (750 TB medical imaging).

Who: SaaS platforms storing user content, e-commerce, gaming, surveillance, IoT.
Triggers: COGS review, CDN bill, scaling past free credits.
Why B2 wins: drop-in S3 replacement (Tribute: “$15K saved per month”), CDN partner egress at $0, free transactions.
Watch out: very high read-op, small-object, high-egress-to-internet apps without a partner CDN can exceed 3x egress. Model it.

Who: MSPs with 20 to 2,000 clients, VARs, distributors.
Why B2 wins: predictable unit cost, Partner API and Groups for sub-account provisioning, Object Lock for insurance requirements, Computer Backup for endpoints.

Who: GPU-as-a-Service providers.
Why B2 Neo wins: white-label storage in weeks, Partner API, custom domains, SSO, no egress fees, no storage team to hire.

Owns: budget, risk, vendor count, board narrative on ransomware and AI.
Fears: surprise bills, a failed restore, vendor viability, migration risk.
Wins with: a 3-year TCO with egress modeled, insurance and compliance evidence, a named reference in their industry, a migration plan that does not consume the team.
Language: “predictable”, “resilience”, “exit cost”, “board”.
Do not: go deep on erasure coding unless asked.

Owns: the backup jobs, the restore SLA, the 2 a.m. page.
Fears: a new tool that breaks jobs, immutability that blocks legitimate deletes, throttled restores.
Wins with: a 20-minute Veeam or Synology setup demo, Object Lock behavior proved live, restore speed test, clear docs.
Language: “SOBR”, “capacity tier”, “immutable”, “3-2-1-1-0”, “RTO”.
Do not: oversell features they will find missing (no lifecycle transitions between classes, because there is one class).

Owns: GPU utilization, data pipelines, dataset versioning, cost per training run.
Fears: idle GPUs, egress tax, throughput ceilings, rate limits, an S3 API gap that breaks a dataloader.
Wins with: a throughput benchmark from their GPU cloud, partner egress confirmation, Overdrive sizing, dataloader compatibility (boto3, s3fs, PyTorch S3 connectors [test before claiming]).
Language: “GB/s per node”, “checkpoints”, “shards”, “time to first batch”, “hot tier vs capacity tier”.

Owns: margin, standardization across clients, insurance requirements.
Fears: egress surprises during a client restore, support quality, billing complexity.
Wins with: unit economics per client, Partner API demo, Object Lock in their backup stack, Computer Backup at $99/year per computer.

Fears: variable bills, commitment risk, prepayment.
Wins with: pay-as-you-go vs 1 to 5 year commit comparison, rate lock, the formula sheet in 14-tco-economics.md.

Fears: data exfiltration, insider deletion, compliance gaps.
Wins with: Object Lock compliance mode, scoped app keys, SSE-B2 default, SSE-C, MFA, SOC 2 Type 2 report via Whistic, BAA, TPN, access logs.