5. Pre-Mortem: It's September 2028 and LedgerSync Is Dead
Three most likely causes of death, ranked by probability. They compound: #1 makes #2 fatal.
Cause #1: Episodic churn outran acquisition (probability ~45% of failure cases)
What happened. Practices signed up for a cleanup, ran two or three jobs, and cancelled once the file was clean. Monthly logo churn ran 9–11%, close to Data Fetcher's 10% gross. The App Store and search brought 8–12 new practices a month, so MRR flattened at about $1.5–2.5k by month 12 and never moved. The scheduled-import feature shipped late and few practices set it up.
Early warning signs
- By month 6, fewer than half of paying practices have run a job in the last 30 days.
- Most cancellations cite "don't need it right now".
- More than 30% of revenue comes from single businesses rather than practices.
- Annual plans are under 25% of new sales.
Fixable? Partly. Sell to practices only, push annual plans, and make the monthly job (scheduled import, month-end checks) part of onboarding, not an optional extra. You can't fully fix the fact that cleanups come in waves.
Cause #2: The platform or an AI agent absorbed the job (probability ~35%)
What happened. Two things. (a) Xero shipped bulk void and bulk recode natively, or made them a JAX agent action ("void these 300 invoices"). Its 2022 answer was "not planned", but Xerocon 2026 was about agents. (b) Practices started using an AI assistant connected to Xero's official MCP server, which already creates and updates invoices. For simple changes it was "good enough", and free. Trials fell by half in a quarter.
Early warning signs
- Xero announces bulk actions or JAX "bulk" workflows at Xerocon, or in the product-ideas thread's status.
- Xero's MCP server adds void or bulk-update tools.
- Trial volume from searches like "bulk void Xero" drops while rankings hold.
- Xero changes developer tiers or connection pricing again with short notice.
Fixable? Partly. You can't stop Xero. You can own what native tools and agents skip: preview, snapshot, reverse, change log, practice-wide view. Agents that write to ledgers make practices more nervous about proof, not less. You could even expose your safety layer to agents (a "preview and log" step), but not before $5k MRR.
Cause #3: One bad job broke trust (probability ~20%)
What happened. A bulk recode ran on a client with a lock date the pre-check missed, or a bulk void removed payments that were reconciled. The practice spent a weekend repairing the file, posted about it in a practice Facebook group, and left a 1-star review. With only 15 reviews at the time, the listing rating dropped below 4.3 and trials dried up.
Early warning signs
- Any job that writes and then fails halfway without a clean reverse.
- Support tickets that start with "it changed things I didn't select".
- Pre-flight checks skipped "to make it faster".
- Record types added before their reverse path was tested.
Fixable? Yes, if built in from day one. No write without a preview. Snapshot before every write. Stop the job on the first unexpected error. Test every record type's reverse path on the demo company before release. Keep professional-indemnity-style wording out of the product: you're a tool, the practice approves every change.
Honourable mentions (less likely, still real)
- Xero pricing shock. Connections grow faster than revenue and push you onto the Advanced tier (AUD 1,445/month) early. Fix: auto-disconnect idle orgs.
- Clone with a free tier. Someone uses AI to clone the tool and gives it away. The free bulk-void tools already show this. Your defence is reviews, reliability and the practice workflow, not code.
- Founder-market fit gap. Without bookkeeping experience you'll build safe-looking features that don't match how practices actually clean files. (Team background wasn't provided. See doc 09.)
- Certification delay. Xero certification takes longer than planned and you can't list until month 5–6. Sell direct in the meantime.