Implementing reliable change data capture and propagation across microservices takes weeks or months of developer time, even for simple use cases like propagating a preference change.
A managed platform that sits between your source database and downstream consumers, handling CDC, outbox pattern, delivery guarantees, and automatic reconciliation out of the box. Deploy via config, not custom code.
subscription
The Reddit signal is strong: a developer spent 3 weeks of actual time spread over months on what should be a commodity problem. The pain is real, recurring (every new service needs it), and felt by the people who would champion buying a solution. However, it's infrastructure pain, not business-critical-emergency pain — teams tolerate it because they don't know alternatives exist. Docking 2 points because many teams have 'good enough' workarounds (polling, manual syncs, shared databases) and the pain is chronic rather than acute.
TAM estimation: ~500K companies running microservices on AWS globally. Perhaps 50K-100K have the specific cross-service data propagation problem and the sophistication to adopt a managed CDC platform. At $500-$2,000/mo average, that's $300M-$2.4B addressable market. However, this is infrastructure — it competes for the same budget as Confluent, AWS services, and internal platform teams. The realistic serviceable market for a startup is the mid-market segment (50-500 engineers) who can't justify building in-house but need more than AWS DMS. That's maybe $50-200M.
Mixed signals. Engineering teams clearly value their time (3 weeks of dev time = $15K-$30K in loaded cost), which makes a $500-$1,000/mo tool an easy ROI sell. However: (1) Infrastructure purchasing often requires VP-level approval, (2) Many teams default to 'build it ourselves' for infrastructure, (3) Open-source Debezium exists and is 'free,' (4) AWS-native teams may prefer to suffer with DMS rather than add another vendor. The strongest WTP signal is from teams who've already tried and failed with DIY — they've felt the pain and know the true cost. Cold outreach WTP is lower.
This is genuinely hard to build well. A solo dev MVP in 4-8 weeks is possible only if heavily scoped: support one source DB (PostgreSQL), one delivery mechanism (webhooks or SQS), basic outbox pattern, and simple reconciliation. The core CDC engine can leverage Debezium under the hood. But the 'managed' part requires: multi-tenant infrastructure, reliable delivery with retries/DLQ, a control plane UI, monitoring/alerting, and — critically — the consistency guarantees that are the entire value proposition. Getting consistency right at the infrastructure level is a 6-12 month engineering effort to do properly. An MVP that demonstrates the concept is feasible; an MVP that delivers on the promise of 'built-in consistency guarantees' in production is much harder.
The gap is remarkably consistent across all competitors: nobody owns the full pipeline from source DB change to confirmed downstream consumption. Confluent/Debezium handle capture but not consumption. Airbyte/Fivetran/Estuary target warehouses, not microservices. AWS DMS is unreliable and limited. Materialize is a database, not a delivery platform. The specific combination of managed outbox pattern + automatic reconciliation + config-not-code + end-to-end consistency is genuinely unserved. The risk is that Confluent or AWS could close this gap with product updates, but they've had years and haven't prioritized it because their core business is general-purpose infrastructure.
Extremely strong subscription fit. CDC pipelines are always-on infrastructure — once deployed, they must run continuously. Switching costs are very high (migrating CDC pipelines is risky and painful). Usage grows naturally as teams add more services and more sync patterns. The value compounds as more of the organization's data propagation flows through the platform. This has all the characteristics of sticky, expanding infrastructure revenue: high retention, natural expansion, and increasing switching costs over time.
- +Clear, consistent market gap — no competitor owns the full CDC-to-consumption pipeline with consistency guarantees, managed outbox, and automatic reconciliation
- +Strong pain signal with quantifiable ROI — 3 weeks of dev time per implementation vs. $500/mo subscription is an easy business case
- +Extremely high recurring/retention potential — always-on infrastructure with high switching costs and natural expansion
- +Growing market driven by microservices adoption — every monolith decomposition creates new customers
- +Config-not-code positioning differentiates from all existing tools which require significant engineering to operate
- !Technical complexity is high — delivering 'built-in consistency guarantees' in production is a 6-12 month engineering effort, not a weekend project
- !Platform risk from AWS — if AWS ships a polished managed CDC service (beyond DMS), it could commoditize this overnight given its distribution advantage
- !Confluent could close the gap — they have the Kafka ecosystem, Debezium connectors, and engineering resources to build an opinionated CDC-to-consumption product if they choose to prioritize it
- !Enterprise sales cycle for infrastructure — convincing teams to trust a startup with their data propagation layer requires significant credibility and SOC2/security compliance
- !Open-source competition from Debezium ecosystem — teams that have already invested in Debezium may prefer to build the consumption layer themselves rather than adopt a new vendor
Fully managed Apache Kafka with CDC connectors
Industry-standard open-source CDC platform that captures row-level changes from database transaction logs
AWS-native CDC replication from source databases to targets
Real-time CDC and ELT platform built on Gazette distributed streaming. Captures database changes and delivers to destinations with exactly-once semantics via declarative YAML/JSON pipeline definitions.
Streaming SQL database that maintains incrementally updated materialized views over CDC feeds from PostgreSQL and MySQL. Real-time derived views with strong consistency.
PostgreSQL-only CDC platform with: (1) Managed outbox pattern — provide a client library that writes to an outbox table transactionally, and a managed connector that reads and publishes changes. (2) SQS/webhook delivery to downstream consumers with at-least-once guarantees, retry policies, and dead-letter queues. (3) A reconciliation dashboard showing source-to-consumer state with one-click re-sync for drifted records. (4) Deploy via a single YAML config file. Skip: multi-database support, complex transformations, self-serve UI. Target: 3-5 design partners from the Reddit/HN community who are actively struggling with this problem.
Free tier (1 source, 2 consumers, 1K events/day) to prove value and reduce adoption friction → Team plan at $299-$499/mo (unlimited consumers, 100K events/day, basic reconciliation) for small engineering teams → Business plan at $999-$2,499/mo (multiple sources, advanced reconciliation, SLAs, SSO) for mid-market → Enterprise at $5K+/mo (dedicated infrastructure, custom compliance, premium support) → Usage-based pricing layer on top for high-volume customers to capture expansion revenue.
3-5 months to first paying customer. Month 1-2: Build PostgreSQL CDC + outbox MVP with webhook/SQS delivery. Month 2-3: Recruit 3-5 design partners from communities (Reddit ExperiencedDevs, HN, relevant Slack groups). Month 3-4: Iterate based on design partner feedback, add reconciliation. Month 4-5: Convert design partners to paid plans, begin content marketing (blog posts showing '3 weeks of CDC work replaced by 10 lines of YAML'). Revenue ramp will be slow initially ($1K-$5K MRR by month 6) given enterprise sales cycles for infrastructure.
- “one of our developers has been working on implementing the solution something like 3 weeks of actual developer time, spread out over several months”
- “This is insane to me”
- “suddenly we have many sources of truth instead of one”
- “There has to be a better way”