An end-to-end loop running today: detect variance with statistical, seasonality-aware baselines → diagnose drivers → narrate in plain language → stage recommendations → grade outcomes in the Outcome Ledger.
Watchtower Design Partner Program
Shape Watchtower on your retention data — and keep the receipt
Thrash Analytics is selecting one or two B2B SaaS companies as design partners for the Revenue-Retention Watchtower. You bring one retention workflow and read-only data access; you get an honest readiness report, a replay of your history reconciled to your own dashboards, a live operating loop with your owners — and founder pricing locked in as the product matures.
Founder pricingScoped at the fit call
A pilot fee set for a design partner, not a customer — meaningfully below future pilot pricing and locked in as Watchtower matures. The fixed-fee diagnostic is folded in.
Honest status
Proven on synthetic data. Your data is the next milestone.
Watchtower grades its own recommendations honestly — including the misses — so it owes the same honesty about itself. Everything below runs end to end today, deterministically, on seeded and adversarial synthetic data. No customer data has flowed yet, and a clean synthetic run proves mechanics, not calibration. The design partnership is the deliberate step from slice-proven to proven-on-your-data.
A confidence engine wired to data-quality scans, with knockout floors that cap how much the system may do when data quality slips.
An adversarial readiness rig: a Stripe-shaped data generator with labeled ground truth and a reconciliation harness that fails loudly when computed metrics drift from the truth.
A read-only semantic-layer adapter (LookML today) that reports which retention metrics your model actually supports — before anything runs.
An operator UI over the live pipeline output: ranked situations, quantified drivers, plain-language narratives, and the Outcome Ledger.
What you get
A working system fitted to your workflow — with proof at every step
The partnership is sequenced so each stage earns the next: readiness before replay, replay before monitoring, monitoring before recommendations.
Founder pricing, locked
A pilot fee set for a design partner, not a customer — scoped at the fit call, with the fixed-fee diagnostic folded in and your pricing locked as Watchtower matures.
The readiness report first
An honest coverage report against your semantic layer and billing data: which retention metrics are supported, which are degraded or blocked, and the confidence ceiling your data can earn — before you commit to the pilot.
Historical replay, reconciled
Watchtower replays your history and reconciles its numbers against the dashboards you already trust, within stated tolerances. Proof comes before monitoring.
A live loop with your owners
One retention workflow — detect, explain, recommend, approve, measure — with your customer-success or RevOps owners approving every action. Nothing executes without a human.
An Outcome Ledger receipt
Every recommendation recorded as a claim and graded against what actually happened — hits, misses, and correct holds. You keep the receipt either way.
Roadmap leverage
Your workflow, metric definitions, and governance constraints shape the first SKU — with direct access to the founder-engineer building it.
What we ask
One workflow, real data access, and candid feedback
The ask is deliberately small in surface area and serious in commitment — the partnership only works if the loop runs on a workflow that matters to you.
A B2B SaaS business where retention KPIs matter — NRR, GRR, churn, contraction — with enough accounts for variance to be meaningful.
Read-only access to billing and subscription data (Stripe-shaped fits best today) and to your semantic layer or warehouse. LookML coverage analysis is built now; your stack shapes what gets built next.
One accountable owner for the pilot workflow — someone in customer success or RevOps who can approve and act on recommendations.
A short weekly working cadence and candid feedback while the loop hardens on your data.
Permission to publish the outcome as an anonymized case study — the honest receipt, whatever the grades say.
Guardrails
Governed from the first byte
The same governance Watchtower applies to its own recommendations applies to the partnership itself.
How it runs
Five stages, each with an exit ramp
Either side can stop at any gate. The point of the sequence is that trust is earned with evidence, not asserted in a deck.
Fit call
Thirty minutes on the workflow, the KPIs at stake, and the data you have. Either side can pass with no friction.
Readiness report
Coverage of your semantic layer and billing data, data-quality readiness, and an honest confidence ceiling — the diagnostic, applied to the partnership.
Reconciled replay
A historical replay of your retention metrics, reconciled to your existing dashboards within stated tolerances. Go or no-go, decided together.
Live operating loop
Watchtower monitors, explains, and stages recommendations; your owner approves; actions are tracked to completion.
Outcome receipt
Graded results in the Outcome Ledger, an anonymized case study, and your founder pricing locked for what comes next.
The system today
The operator view a design partner starts from
Screens below show the working system on simulated data — labeled as such deliberately. Replacing simulated with reconciled-real is the work we do together.
One or two partners. One workflow each.
Honest grades, whatever they say.
If the replay does not reconcile or the loop does not earn trust, you will know — and so will we. That is the point of the program.
Agentic Analytics Readiness Scorecard
Score one KPI, queue, or exception pattern in 10-15 minutes. Email is requested only after the scorecard is complete.
Readiness band, knockout check, two gaps, and first fixes.
Start the ScorecardWe do not sell your information.