Palm Product Knowledge¶
A knowledge management system that ensures every product, design, and engineering decision at Palm is grounded in real customer problems and desired outcomes — not internal assumptions.
Philosophy¶
The core belief behind this system: deep problem understanding beats solution suggestions. When you truly understand how treasury teams work, what they're trying to accomplish, and where they struggle — possible solutions become more accurate.
Every insight flows through the Jobs to Be Done framework, using the Desired Outcomes formula as a guide:
Example: - Job: "Reconcile bank transactions daily before 3 PM" - Outcome: "Minimize time required to match transactions across systems"
All extracted knowledge uses two-tier validation: - Confirmed Outcomes — 2+ independent sources corroborate - Emerging Signals — single source, needs corroboration before driving decisions
This keeps the knowledge base honest about what we actually know vs. what we've heard once.
Note: We use JTBD and the Desired Outcomes formula as a helpful framing tool, not a rigid standard. Not every job or outcome will adhere perfectly to the formula due to practical limitations — we do our best to capture the intent clearly.
Product Roadmap¶
Stack-ranked by priority. Each item is a job to be done — not a feature. The focus outcomes below each job are what we're actively investing in.
#1: Apply percentage-based assumptions to forecasts while preserving underlying patterns¶
Stage: Skateboard (Validate) | Domains: Scenario Modelling, Cash Forecasting
Why now: Every customer and prospect does scenario modelling in Excel. Simple percentage-based adjustments that preserve underlying patterns are more valuable than complex scenario trees.
Focusing on: - Minimize the risk of overestimating collections or underestimating outflows - Increase the ability to model scenarios quickly (e.g., -5% on collections) - Reduce manual effort in creating multiple forecast scenarios - Minimize the loss of weekly/daily patterns when adjusting forecast totals
#2: Configure forecasting models based on data quality and account characteristics¶
Stage: Skateboard (Validate) | Domains: Cash Forecasting
Why now: Customers with 100+ accounts need control over which models run where. Current one-size-fits-all approach doesn't scale.
Focusing on: - Minimize forecast errors caused by poor data sources (e.g., unreliable ARP data) - Reduce time spent correcting forecasts that could have been configured better
Already addressed: ML on/off toggle (✅)
Delivered
See total liquidity position including invested cash alongside operational cash — View
Upload forecast data from files without extensive reformatting — View
Voice of the Customer¶
Last updated: 2026-03-17
Emerging Signals¶
Not yet validated as jobs, but worth watching. Need more data points.
| Signal | Source | What We're Hearing | Questions to Explore |
|---|---|---|---|
| FX exposure and risk management | ON, Personio, Discogs (2026-03) | ON building FX risk management from scratch (Yulia: four exposure categories). Personio wants FX policy triggers and automated notifications. Discogs tracking FX losses on Bound trades. Three customers in one period — wasn't on the roadmap but demand is real. | Is this a standalone domain or a scenario building block? What's the minimum viable FX feature? How does it relate to the co-founder's FX Flow vision pillar? |
| Dynamic minimum cash threshold per entity | ON (2026-03-04) | Lucia wants a calculated minimum cash line (e.g., "2 weeks of upcoming payables") overlaid on forecast — moves with the forecast, not hardcoded. Used for CFO M&A guidance, regional buffer negotiations. "You keep telling me you need 200, but what we see is you'd be okay with 50." | Is this a scenario building block or a distinct feature? How do other customers think about minimum cash? |
| Proactive data monitoring & alerts | Discogs, ON, Personio (2026-03) | Discogs missed a GBP account upload — caught manually through variance checking. ON has ~3,000 outliers from miscategorization. Personio wants counterparty risk updates, payment rail outage alerts. Customers catching data issues manually. | How to avoid alert fatigue? Build into existing workflows vs new notification system? What triggers are valuable? |
| Debt management is a nightmare | Avramar (expert) | Complex loan structures, multiple tranches, floating rates, securities tracking all in spreadsheets | How common is this? Enterprise only? |
| Payment execution | Highline | Explicit dealbreaker — won't buy without it. ~250 monthly payments. Also evaluating Kyriba, GTreasury, Nihilus | Build vs integrate? What's the market size for payment-first buyers? |
| Budget vs forecast variance | DAZN, ON | DAZN: compare treasury forecast against FP&A budget + EBITDA overlay. ON (Lucia): "Our controlling team plans P&L for the next five years... it's very difficult for me to tell if it's realistic. So I use Palm to do a sense check." | How does this relate to direct/indirect bridging? Now validated by 2 sources — consider promoting. |
| Push reporting for regional teams | ON | Regional finance teams lack basic visibility. Scheduled PDF reports to APAC/Americas/EMEA hubs — not requiring Palm login. $8M IC payment from Canada went unnoticed | Is this access control, or a distinct distribution model? |
| Forecast explainability at category level | Personio, ON | Tom wants to see WHY a forecast value is what it is. ON's FC Validation deck builds full WMAPE methodology in Excel because Palm doesn't have it — bias tracking, category-level accuracy, KPI framework. | Adjacent to accuracy KPIs but distinct: controllability, not just transparency |
| Invoice discounting integration | Nash Squared | Nadim to send details. Non-standard financial instrument | How common is this in the ICP? |
Sales Pulse¶
What prospects are asking about in sales conversations.
The pattern: Excel is still the universal competitor — every prospect maintains parallel spreadsheet processes regardless of existing tooling. But the conversation is shifting. Prospects aren't asking "can you forecast?" — they're asking "can I trust it, can I stress-test it, and can my team use it?" Enterprise names are entering the pipeline (Mercedes-Benz, SSE, Morrisons, Anglian Water), all bringing multi-entity, multi-currency complexity. The PoC/sandbox motion is the deal accelerator: every advancing deal (Swiss Steel, Bio-Rad, Euroports, DAZN) is progressing through hands-on evaluation with their own data.
Key themes explained:
-
Excel is still the real competitor: Even with Kyriba, SAP TRM, or Oracle, treasurers maintain parallel Excel processes. Bio-Rad runs 20 different Excel reports daily despite Kyriba. Swiss Steel has 16 entities sending individual Excel forecasts since 2019. DAZN has a 2+ year custom Excel model with deep organizational trust.
-
Role-based access control: Euroports (critical for April close), Digital Realty, Nash Squared, and ON all need RBAC. Entity-level permissions are table stakes for enterprise deals and the gateway to multi-user expansion.
-
Scenario planning / stress testing: DAZN (acquisition what-ifs, 10% EUR shock), Swiss Steel, Euroports all asking for what-if modelling. Validates our #1 roadmap bet. ON and Personio have both validated the prototype from the customer side.
-
Data agnostic connectivity: Palm's ability to ingest via API, SFTP, MT940, BAI, CSV, and email is consistently called out as a differentiator. Nash Squared (15 banks), Dr.Max (900 accounts), Swiss Steel (30+ entities) all see this as a key advantage vs competitors requiring specific integrations.
-
Multi-entity complexity at scale: Swiss Steel (30+ entities), Dr.Max (550 entities, 900 accounts), BRF (global treasury with complex SAP) — the bigger the company, the worse the consolidation problem. This is our sweet spot.
-
AI categorization resonates: Nash Squared responded strongly to ~90% accuracy vs rules. Foot Locker's daily manual review of 3-4 transactions out of 1,000-2,000 validates the pain point.
| Theme | Frequency | Recent Example | Date |
|---|---|---|---|
| Manual Excel forecasting | Very High | 16 entities sending individual forecasts since 2019 — Swiss Steel | 2026-03 |
| Role-based access control | High | Entity-level permissions blocking purchase — Euroports, Digital Realty, Nash Squared | 2026-03 |
| Scenario planning / stress testing | High | Acquisition what-ifs, currency shocks — DAZN, Swiss Steel, Euroports | 2026-03 |
| Data agnostic connectivity | High | MT940, BAI, SFTP, API, CSV, email flexibility — multiple prospects | 2026-03 |
| Multi-entity complexity | High | 550 entities, 900 accounts, one person in Excel — Dr.Max | 2026-03 |
| AI categorization | Medium | ~90% accuracy vs manual rules — Nash Squared, Foot Locker pain | 2026-03 |
| Payment execution | Low | Explicit dealbreaker for Highline (~250 monthly payments) | 2026-03 |
Pipeline snapshot: 23 active deals, 13 new in 2 weeks, £500k weighted pipeline. Dunelm most advanced (S4, close target Apr 30). Bio-Rad PoC kicking off. Swiss Steel largest deal (£70k). Euroports down to Palm vs one other vendor.
CS Pulse¶
What current customers are struggling with or asking for.
The pattern: The customer base is shifting from "adopting Palm" to "building on Palm." All three active customers showed strong engagement this period — Discogs building dashboards live in sessions, ON expanding into BigQuery integration and FX risk management, Personio planning full daily process migration by end of 2026. Pulse AI landed well across all three (Discogs: $60k investment calculation demo, ON: wants newsletter, Personio: wants market-intelligence digest). The new cross-customer theme is FX: ON building FX risk management from scratch, Personio wants policy automation, Discogs tracking FX losses — none of this was on the roadmap. Meanwhile, proactive monitoring is a gap: Discogs missed a GBP account upload undetected until manual variance checking.
Key themes explained:
-
Pulse AI momentum across all customers: ON wants Pulse newsletter for seasonality insights. Discogs responded strongly to the demo ($60k investment calculation example, historical balance queries, FX trade support). Personio wants daily digest combining internal data + market events. All three exposed to Pulse this period with positive reception.
-
FX emerging as cross-customer theme: ON building FX risk management from scratch with Yulia (four exposure categories, CHF reporting currency needed). Personio wants FX policy triggers and automated notifications. Discogs tracking FX losses on Bound trades. This wasn't on the roadmap but demand is real from 3/3 customers.
-
Self-serve adoption inflection: Discogs building dashboards live in session (Jamie and Stephanie). ON getting self-serve dashboard simplification. Personio planning sprint-based module adoption. Cash analysis report described as "most popular report in Palm across all customers." Customers moving from "show me" to "let me do it."
-
AP/AR integration crystallizing: ON's BigQuery integration for AP/AR data is the first concrete implementation (~1 week Palm-side setup, Q2 target). New stakeholder Andrew Vendittelli (tech lead/PM) engaged. If successful, becomes the template for other customers.
-
Proactive monitoring gap: Discogs missed a GBP account upload — caught through manual variance checking. ON has ~3,000 outliers from miscategorization. Personio still cross-checking Palm against legacy files. Customers catching data issues manually. Automated monitoring would increase trust.
-
Scenario planning expanding in scope: ON wants scenarios for growing categories. Discogs wants currency appreciation/depreciation modeling and stress testing. Personio wants worst-case from historical extremes. The need goes beyond simple percentage adjustments into currency-specific and timing-based scenarios.
| Theme | Customer(s) | Severity | Recent Example | Date |
|---|---|---|---|---|
| FX exposure / risk management | ON, Personio, Discogs | High | Yulia building FX risk from scratch (4 exposure types), Personio wants policy automation, Discogs tracking Bound losses | 2026-03 |
| AP/AR data integration | ON | High | BQ integration planned, Andrew Vendittelli engaged, ~1 week setup, Q2 target | 2026-03 |
| Proactive data monitoring | Discogs, ON | High | GBP account upload gap undetected; ~3,000 outliers from miscategorization | 2026-03 |
| Scenario planning | ON, Personio, Discogs | High | ON: growing categories. Discogs: currency stress testing. Personio: worst-case from historicals | 2026-03 |
| Pulse AI adoption | ON, Personio, Discogs | Positive | All 3 customers demo'd, Discogs $60k investment calc example landed well | 2026-03 |
| Self-serve dashboard adoption | Discogs, ON | Positive | Jamie/Stephanie building dashboards live, cash analysis "most popular report" | 2026-03 |
| Personio full migration planned | Personio | Positive | Tom planning full daily process migration to Palm by end of 2026 | 2026-03 |
| CHF reporting currency | ON | Medium | Yulia needs CHF (defaults to USD functional), balance sheet vs transaction exposure | 2026-03 |
| Filter/chart issues | ON (Yulia) | Medium | Amount <0 filter not working, Y-axis showing "time", line chart not enabled | 2026-03 |
Product Observations¶
Patterns and insights from the product team.
-
The trust chain is sequential: Data quality → categorization → forecast accuracy → scenario analysis → decision-making. Each link broken undermines everything above it. Connectivity failures erode trust even when long-term forecasting performs well. As Gurjit put it: "Until someone trusts the outcomes hand over fist, there's no execution happening."
-
Short-term vs long-term forecast are two different products: ML is valuable for weeks 5-13+ (strategic: investments, FX hedging), but treasurers' magnifying glass is on the next 2 weeks (operational funding decisions). Short-term needs deterministic data (AP/AR, payroll), not probabilistic ML. We may not be aggressive enough on the deterministic layer. ON's validation deck confirms: Palm at 4.5% WMAP at 2 weeks vs Kyriba at 12.5% — short-term is where we win hardest.
-
Forecast validation as a sales weapon: ON's head-to-head vs Kyriba (8% vs 16% WMAP) is now a formal deck presented to leadership. Federico/Giannis built the methodology (snapshot vs horizon WMAPE, bias tracking, category-level KPIs) — this is the kind of proof point that closes deals. We should make it easy for every customer to run this comparison and productize the validation workflow.
-
FX is the next cross-cutting domain: Three customers independently raised FX in one period. ON is building risk management from scratch, Personio wants policy automation, Discogs is tracking losses. This aligns with the co-founder's FX Flow vision pillar but emerged from customer pull, not top-down strategy. The question is whether to tackle FX as a standalone domain or weave it into scenarios (currency stress testing) and reporting (exposure dashboards).
-
Pulse AI landed — now it needs direction: All three customers responded positively to Pulse, but each wants something different: Discogs wants ad-hoc Q&A ($60k investment calculations), ON wants scheduled newsletters (seasonality insights), Personio wants a market-intelligence-enhanced daily digest. The product question is which mode to double down on first.
-
The spreadsheet gravity well: Even companies with enterprise TMS revert to Excel. Bio-Rad runs 20 Excel processes despite Kyriba. DAZN has a 2+ year Excel model with deep organizational trust. Personio is in "quasi-process" state using Palm + legacy files for cross-checking. The transition from parallel running to full adoption is the key CS challenge.
-
PoC/sandbox is the deal accelerator: Swiss Steel, Bio-Rad, BRF, DAZN, Euroports — every advancing deal is progressing through hands-on evaluation. The faster we can spin up a sandbox with their data, the faster we close.
-
Confidence > Accuracy: Treasurers will accept a less accurate forecast if they understand and trust how it was generated. "Can I trust it?" is the universal question. Federico (ON): "More about building trust... methodology is a draft that we can improve over time."
-
Proactive > Reactive for AI: The highest-value AI use case isn't chat — it's scheduled analysis that surfaces things users didn't know to ask about. Giannis ran CapEx categorization analysis for Amanda (precision improved 90→96%); she didn't know to request it. Meanwhile, Discogs's missed GBP upload and ON's 3,000 outliers show that proactive data monitoring — upload verification, outlier alerting, freshness checks — would compound trust. Lower-effort than new features, high-impact for adoption.
-
IC is a multi-layer problem: The Feb IC sync mapped three layers: (1) Is it intercompany? (2) What type — cost plus, dividend, loan? (3) What's the operational implication? ZBAs are solvable, but cost-plus settlements — large, AP-driven, treasury lacks visibility — are the high-value target. ON's scenario session added FX dimension: failed IC payments create open currency exposure.
-
Scenario analysis has four building blocks, not two: Percentages (what-ifs), composable named events, timing shifts ("delay IC payments by 30 days"), and dynamic calculated thresholds. Plus Discogs wants currency-specific scenarios (appreciation/depreciation modeling). The building blocks are: amount, percentage, timing, dates, calculated thresholds, and currency.
-
Regional enablement unlocks enterprise scale: ON wants 3 regional hubs, Euroports needs entity-level access for local CFOs, Nash Squared has global locations. RBAC + push reporting is the path to $200-400k ACVs. ON's scenario session validated entity-level scenarios as key to regional buffer negotiations.
-
Connectivity is a trust investment, not an execution investment: Strategic reframe from the product bets discussion. We don't build connectivity because it enables features — we build it because it enables trust.
12-Month Product Vision¶
Where Palm needs to be to win.
The shift: From "forecasting tool" → "treasury decision platform"
Must-Solve Problems¶
| Problem | Why It's Hard | Customer Evidence | Status |
|---|---|---|---|
| Scenarios & Assumptions | Layer user assumptions on ML without breaking it | ON (8/10 on prototype; dedicated session validated timing shifts, dynamic min cash thresholds, entity-level scenarios), Personio (worst-case from historicals), Discogs (currency stress testing), DAZN, Swiss Steel, Euroports | Assumptions backend shipped, timing shifts + min cash planned |
| Investments & Liquidity | Data from many sources, maturity modeling | ON (dashboard near-production, MMCs missing), Personio (migrating, Treasury Spring parser shipped, BlackRock issues), Discogs ($60k investment calculation via Pulse) | In progress — Treasury Spring ingestion shipped, investment data in all datasets |
| Debt Management | Complex structures, floating rates, covenants | Avramar (syndicated loans, 6+ tranches), Personio (debt covenant reporting consuming Tom's focus until April) | Not started |
| Forecast Explainability | Make ML transparent without overwhelming users | ON (formal validation deck: Palm 8% vs Kyriba 16% WMAP — "definitively the superior model", presented to leadership), Personio (wants category-level model explanations) | Limited — WMAPE methodology built by customer, needs productizing |
| FX Exposure & Risk | Multiple exposure types, policy rules, hedging workflows | ON (Yulia building from scratch, 4 exposure types, CHF reporting currency needed), Personio (FX policy triggers, interest rate sensitivity), Discogs (tracking Bound trade losses) | Not started — strong emerging signal from 3/3 customers |
| Intercompany Forecasting | Paired transactions, zero-sum constraint | ON (ZBA live, $8M missed payment, capex IC distinct from operational, failed IC payments create FX exposure), Personio (monthly sweep logic) | ZBA in progress, full IC not started |
| Proactive Intelligence | "Late" vs "not happening", surface unknown unknowns, avoid alert fatigue | Pulse demo well-received by all 3 customers. Discogs: missed upload caught manually. ON: wants Pulse newsletter. Personio: wants market intelligence digest | Pulse shipped (chat + digest), proactive monitoring not started |
| Direct/Indirect Bridge | Category mapping, working capital effects | DAZN (budget vs forecast + EBITDA overlay), ON (Lucia: "very difficult to tell if controlling's 5-year plan is realistic... I use Palm to do a sense check"), Personio, Euroports | Not started |
| AP/AR Integration | Every ERP is different, data quality varies | ON (BQ integration planned Q2, Andrew Vendittelli engaged, ~1 week setup), Bio-Rad (Kyriba API issues) | Not started — ON BQ integration is first concrete implementation |
| Access & Entitlements | Granular permissions without complexity, clean feature gating | ON (3 regional hubs, entity-level scenarios for regional buffer negotiations), Euroports (April deadline), Digital Realty, Nash Squared | Roles & permissions design complete, entity management page shipped |
Quick Links¶
- Domain Knowledge - What we know about treasury problems
- Transcripts - Raw customer conversations (see navigation)
- Features Registry - What we've built and why
- Roadmap - Jobs we're investing in
- Palm Product MCP - Connect your AI agent to this knowledge base
This page is the product team's homepage. Keep it current.