Bridge direct and indirect cash flow forecasts to explain variance drivers¶

Investment Thesis¶

Treasury teams at mid-to-large companies run parallel forecasting worlds — direct (cash-based, treasury-owned) and indirect (budget/P&L-based, controlling-owned) — and spend significant manual effort in Excel reconciling them each cycle. Five sources (Euroports ×2, Personio, Levi's, ON) validate that the real pain isn't building the numbers, it's explaining working capital variance drivers without interviewing every local entity. A Palm bridge that inherits the category alignment and lets users drill from a consolidated variance down to driving transactions is a differentiator vs spreadsheet workflows and vs competitors that treat direct and indirect as separate modules.

Notes on Scope & Direction¶

YTD progress-bar view by category (Jen & Emma, 2026-05-11)

Direct cash and indirect budget rarely align month-to-month — some categories are evenly spaced in the budget (e.g., insurance budgeted at 100 EUR/month, 1,200 EUR annual total) but lumpy in cash (one annual payment). Comparing them point-by-point creates false variance and noise.

Working hypothesis (a strong one, not yet a confirmed fact): what actually matters for these categories is year-to-date progress against the annual commitment, not the month-by-month delta. Idea: a category-level progress-bar view where each bucket shows YTD spend as a % of the annual budget.

How a user reads it: - Insurance bucket at 0% → click → see no payments have gone out yet. Expected if the lump-sum hasn't fired. - COGS bucket at 70% when only 50% of the year has elapsed → click → investigate why it's running ahead.

Why this could land in v0: it absorbs the timing-mismatch problem (lumpy cash vs evenly-spaced budget) without forcing a synthetic point-in-time match, and gives a fast "is this category on track?" read at a glance. Complements — not replaces — the side-by-side bridge view, variance flags, and drill-down.

Bridging — canonical domain for direct/indirect reconciliation and variance explanation
Categorization — mapping bank-statement categories to FP&A budget line items is a prerequisite
Variance Analysis — overlap on drill-down and commentary patterns
Cash Forecasting — the direct forecast being bridged

Desired Outcomes¶

Current Focus¶

These are the outcomes we're actively investing in.

#	Desired Outcome	Evidence
1	Minimize the time required to identify deviations between direct cash flows and indirect budget	Euroports, Personio, Levi's, ON
2	Reduce the frequency of having to ask local entities for variance explanations	Euroports
3	Increase visibility into working capital movement drivers (AR vs AP breakdown)	Euroports
4	Increase the ability to drill from consolidated variance down to entity and transaction level	Euroports

Already Addressed¶

No existing features address these outcomes yet.

Not Yet Addressed¶

Known outcomes we're not focusing on yet.

Desired Outcome	Evidence
Increase the ability to capture variance commentary directly against bridge line items (replacing Excel free-text boxes)	Euroports
Minimize the manual effort to produce board-ready cash flow bridge reports	Euroports
Minimize the time required to produce T+1 monthly cash burn analysis	Personio
Reduce the manual effort to reconcile transactional forecasts with FP&A categories	Euroports, Personio
Increase visibility into which differences are timing vs permanent	Levi's
Increase the speed of comparing direct forecast vs FP&A forecast for the same period	Personio, ON
Persist confirmed variance explanations against past bridge periods, so the same investigation isn't repeated months later	ON
Improve the precision of future variance explanations by feeding confirmed causes back into the model	ON
Reduce the number of hand-maintained assumption tables required to translate plan into cash (AR aging, VAT timing, payment-term tables)	ON
Decrease the structural complexity of the bridge by replacing rule chains with statistical inference (e.g., infer warehouse cash from sales patterns instead of via the inventory→COGS detour)	ON
Surface entity- and currency-specific shortfalls revealed by the bridged balance-sheet projection	ON

Current Approach¶

Stage: Skateboard (Validate)

V0 prototype walks the user through a consolidated direct-vs-indirect view with three core moves:

Top-of-page variance flags — draw attention to where direct and indirect are drifting apart (especially working capital)
Category mapping UI — each indirect (budget) bucket shows its direct (Palm category) equivalents and how much flows from each
Drill-down on variances — from variance figure → driving transactions (eventually entity → transaction level)

V0 leans on categorized transactional data only; AR/AP granular feeds and factoring integration are follow-ups. Commentary capture (replace the Excel free-text box) and AI-assisted variance explanations are planned for later iterations.

What's Validated¶

Overall direction — "This is more than enough validation for us to keep going in this direction" (Emma, summarizing Matthias' feedback, 2026-04-15). Corroborated by Lucia (ON, 2026-04-29): "I'm feeling quite happy. I feel like we've high-level validated the direction this is going to take."
The bridge as the automated plausibility check — Lucia explicitly maps the v0 to her existing workflow: "you're reducing an activity that is very high touch, which is us in an Excel checking and seeing how the forecast compares to the long-term plan." (ON, 2026-04-29) — this is the core JTBD framing the bridge serves.
Top-of-page variance flags — "I think that makes sense, taking into account that we're actually focusing on the working capital" (Matthias, Euroports, 2026-04-15)
Category mapping UI — "And then with the drill-down capability, obviously these categories would need to be changed. Yeah, I think it would be [helpful]." (Matthias, 2026-04-15). Corroborated for ON: "we already follow the same categories across platforms. So this would be a very easy case for the tool to do the comparison." (Lucia, 2026-04-29)
Drill-down from consolidated → entity → transaction — "What would be helpful is if you go from consolidated to entity level and then ultimately transaction level" (Gurjit summarizing, Matthias confirming, 2026-04-15)
Need for commentary capture in Palm — "The question is then, how do you capture those comments? If that's not possible, it would be me joining a Word doc or something" (Matthias, 2026-04-15). Lucia adds the deeper need: confirmed explanations should persist against past bridge periods AND feed back to the model — "ideally it would be kind of fed back in, so it helps identify this directly the next time. A bit more like the categorization prompting." (Lucia, 2026-04-29)
Cross-functional persona (treasury + controlling) — Lucia: "this role who uses this is a bit blurred because it's a conversation that's probably half treasury, half controlling. But super happy to go into this in-between area." (ON, 2026-04-29)

Key Learnings¶

Strategic learnings beyond v0 validation, surfaced from feedback sessions. These shape future scope and framing rather than current build.

From Lucia (ON, 2026-04-29)¶

Two-tier validation framework is how ON treasury thinks about forecast confidence. "Two ways of telling if a forecast is good": (1) variance analysis vs past actuals, (2) plausibility check vs controlling's long-term plan. The bridge is fundamentally the automated plausibility check. Without it, treasury manually does the gut-feel comparison in Excel/Anaplan before acting on the forecast. Framing implication: position the bridge as "the second tick in your head" — not as a separate analytical surface.
Time-horizon split — ≤13wk Palm wins, >13wk FP&A wins. Lucia is explicit that the long-term plan should NOT bias the short-term operational forecast ("you come up with a better forecast by ignoring the long term in the short term"), but FOR >13 weeks (especially >6–7 months) the FP&A long-term plan must dominate to maintain organizational alignment ("it just creates friction in the conversation with controlling when they say, yeah, but you're not taking our numbers"). Product implication: the bridge view shouldn't blend the two indiscriminately — the weighting between Palm bottom-up and FP&A top-down should depend on where you are in the horizon.
Strategic ambition: replace Anaplan's rule-based translation logic entirely. Beyond v0 (visualize + explain variance), Lucia wants Palm to ingest P&L + balance sheet plans and produce direct cash flow itself, replacing the hand-maintained Anaplan tables (AR aging splits, VAT timing, inventory→COGS chain). She'd "pitch this back to FP&A and say, can we treat this as official?" This is a much bigger arc than the current v0 scope but matters for product positioning — Palm as the production source of the bridge, not just a viewer of it.
"No touch" vs "low touch" distinction. Daily operation should be no-touch (model autonomous, treasury just confirms). But the override path must remain — "I always want to have the option to go in and say, nah, let me type in five instead of two" for special-knowledge cases. Product implication: don't confuse no-touch with no-override. Both are needed simultaneously.
Explainability is the gating constraint for FP&A adoption. "You're never going to trust a black box." Lucia uses Gemini-in-Sheets as the reference: a tool that shows what it's doing step-by-step. Without that, controlling/FP&A won't accept Palm's logic as the official cash translation source — and the strategic ambition above (replacing Anaplan) becomes politically blocked.
Boundary-testing pilot offer. ON volunteered to test the tool's limits with three input granularities: (1) ON's already-translated direct cash flow with same categories (easy case), (2) indirect with same categories, (3) raw P&L. Lucia: "We can test where the tool finds its limits and what kind of additional mapping or prompting it needs." Use ON as the pilot to find where model-driven inference works and where it breaks.
Persona is treasury + controlling + (potentially) CFO. Lucia explicitly frames the bridge as a cross-functional surface: "this role who uses this is a bit blurred because it's a conversation that's probably half treasury, half controlling." Treasury uses it to plausibility-check; controlling uses it to challenge regional teams (e.g., APEC consistently overshoots); CFO uses it to "understand how my P&L transforms into [cash]." Don't over-optimize the UI for treasury alone.

From Matthias (Euroports, 2026-04-15)¶

Working capital deviation is the analytical core. "What I need to answer is always: what's the working capital deviation? What's driving that?" The bridge isn't about reconciling the whole P&L → cash translation — it's specifically about explaining the working capital movements that show up between forecast and actuals. EBITDA + adjustments come from budget directly; the variance lives in working capital lines.
Board reporting is the output artifact. Matthias has to produce a quarterly board deck with specific columns (Actual / Last Estimate / Previous Estimate / Prior Year / Budget) and free-text commentary per row. Today this is rebuilt manually from Excel each quarter. The bridge should produce this directly — not just be an analytical view.
Local entity forecasts have systematic bias that the bridge surfaces but can't fix alone. AP-staffed forecasts overestimate payments; conservative entities like Finland match budget; Spain deflects. Group treasury manually overlays targets to compensate. Implication: forecast accuracy KPIs per entity (a "scoreboard") are a complementary surface to the bridge, not in scope for v0.

Next Milestone¶

Follow-up session with Matthias in ~1-1.5 weeks to share v0 progress for testing
Incorporate Friday IT/accounting meeting feedback from Matthias on AR/AP data availability
Categorization workshop with Giannis to define bank category → budget bucket mapping
Scope BigQuery integration with ON's data lake (long-term FP&A plan as next AP/AR-style table)
Plan ON pilot for boundary testing (direct → indirect → raw P&L)

Feedback Log¶

Date	Company	Validated	Summary	File
2026-04-15	Euroports	Yes	V0 direction validated. Top-of-page variance flags, category mapping UI, and drill-down all resonate. Key asks: entity drill-down with groups, commentary capture in Palm, forecast accuracy scoreboard, factoring customer-level drill-down.	View
2026-04-29	ON	Yes	Direction validated. ON volunteered as pilot to test tool boundaries (direct → indirect → raw P&L). Key asks: variance feedback loop with persistence, historical commentary view, explainability surface, BigQuery ingestion for long-term plan. Strategic ambition: replace Anaplan's rule-based bridge logic.	View