Potrebujem nadzor nad prodajo.
Napoved, ki preživi sejo uprave.

Forecast Reality Check — demo run

Input: sample-pipeline.csv (15 open deals, Amberline Software's live pipeline as of 2026-08-28) against historical-conversion-rates.csv (trailing 8 completed quarters ending 2026-06-30, n=54–140 deals per stage).

Baseline used:

Stage order for the remaining-days waterfall: Discovery → Qualification → Proposal → Negotiation → Closed Won.

Step 1 — Fantasy test, every deal

remaining_needed = max(median[stage] − days_in_stage, 0) + sum(median[later stages])

9 of 15 deals flag Fantasy.

Step 2 — Sandbagging test, the 6 deals that passed Step 1

gap = win_rate_from_stage − stated_confidence_pct; flag at gap ≥ 20 points.

3 of 15 deals flag Sandbagged. 3 of 15 clear as Accurately forecast.

Full result

Pipeline value by flag

Total pipeline: €1,340,000

Read this straight: barely 1 in 7 euros of this pipeline is both correctly dated and correctly weighted. Just over half is forecast to close on a date the team's own history says it can't hit. Just under a third is being under-called by reps sitting on deals that, by the team's own numbers, are healthier than they're admitting to.

Honesty check

• The historical table's sample sizes range from n=54 (Negotiation) to n=140 (Discovery) over 8 quarters — reasonably solid, though Negotiation's n=54 is thinner than the others and its 84% win rate should be treated as a good-not-bulletproof baseline.
• All 15 deals had a stated confidence field, so all 15 could be run through both tests — no deal was scored on date-math alone for lack of data.
• Two deals (D-04, D-05, D-11) are flagged Fantasy and show confidence meaningfully above their stage's win rate — the worst combination in the file: a date that can't be hit, held with more conviction than the underlying stage typically earns.
• What to start capturing for a sharper run next time: a per-deal stage-entry-date history (not just the current stage's entry date), so a chronic-stall flag can distinguish "this deal has been slow everywhere" from "this deal was fast until it hit this one stage" — right now both look identical from a single stage-entry timestamp.

One thing worth flagging back to the founder

The chronic-stall override (Step 1, check 2) did real work in this run: three of the nine Fantasy flags — Ashford, Juniper Hall, and half of Castlemere's flag — would have been invisible to the date-subtraction formula alone, because a deal already far past its stage's median produces max(median − days_in_stage, 0) = 0, which reads as "nothing owed" rather than "badly overdue." Any competitor building this without the override would silently pass Juniper Hall Insurance — 88 days in a 10-day-median stage, four prior close-date pushes, still marked 95% — as an achievable near-term close. That gap is the actual product.

The CRM Confessional — demo run

Input: sample-pipeline-export.csv, 35 open deals, Fernbridge Industrial Supply's live pipeline as of 2026-08-28. No separate stage-duration benchmark was supplied, so per the rubric's fallback the median time-in-stage was inferred from the spread of deals currently sitting in each stage. No staleness cadence was stated, so the 21-day default applies.

Inferred stage benchmarks (median of days_in_current_stage across the deals presently in that stage — a soft proxy, not true historical stage duration; see honesty check):

Overrun gate: days_in_stage ÷ median > 1.3.

Summary

Total known pipeline: €2,406,500 across 34 deals with a stated amount. A 35th deal (Wraithmoor Systems, D-09) has no amount on file at all — its size isn't small, it's unknown, and that's a finding in itself, not a zero.

Categories overlap — a deal can trip more than one test. 23 of 35 deals (66%) trip at least one of the four hygiene tests, worth €1,714,000 (71.2% of known pipeline). Add the 3 unforecastable records and 26 of 35 deals — 74% of the file — have a problem serious enough to flag. Only 9 deals, worth €552,000, are clean across every test.

1. Stage overrun — 9 deals, €881,000 (36.6% of pipeline)

€881,000 (36.6% of pipeline) is sitting in a stage the record itself says it left behind weeks ago, meaning any stage-by-stage pipeline total your team reports upward is currently overstating how much of that money is actually still in motion.

Worth watching, not flagging — three deals sit right at the 1.3 boundary and would flip with one more slow week: D-33 Amberfield Logistics (Discovery, 13 days ÷ 10 = exactly 1.3, the gate requires strictly greater), D-34 Palewell Materials (Qualification, 18÷14 = 1.29), and D-32 Quillfeather Logistics (Proposal, 25÷20 = 1.25).

Engagement without progress — a different failure mode, called out separately

Two of the nine overrun deals have been touched recently even though their stage hasn't moved in months. This is not the same problem as neglect — it means someone is actively working the deal without it actually advancing, which is a coaching problem, not a visibility one:

• D-20 Farrow Point Insurance — 55 days in Qualification (3.9× median), last touched 8 days ago. Someone is emailing this account regularly; the deal just isn't moving stage.
• D-14 Draycott Manufacturing — 80 days in Proposal (4.0× median), last touched 12 days ago, next-step note reads "still exchanging comments internally on pricing tiers." €134,000 of live back-and-forth that the stage field says has been stalled for two and a half months.

2. Missing next-step — 14 deals, €1,072,000 (44.6% of pipeline)

D-02 Hallowfen Energy (€58,000) · D-30 Griswold & Pike (€27,000) · D-25 Petrel Bay Robotics (€64,000) · D-13 Ossuary Bio Labs (€38,000) · D-16 Bramwell Group Ltd (€47,000) · D-31 Ferrow & Dunn (€26,000) · D-20 Farrow Point Insurance (€89,000) · D-26 Ainsdale Robotics (€92,000) · D-29 Halloway Pharma (€81,000) · D-03 Bellhaven Retailers (€77,000) · D-35 Ombersley Textiles (€112,000) · D-15 Carrow Dean Foods (€96,000) · D-23 Torvane Industrial (€87,000) · D-11 Wickfast Logistics (€178,000)

€1,072,000 — 44.6% of this pipeline — has no logged next action, meaning almost half of next month's forecast is a guess about deals nobody has committed to a specific move on. Two of these are also your two largest stage-overrun deals (D-11, D-35): not only stalled, but stalled with no stated plan to un-stall them.

3. Stale record — 9 deals, €773,000 (32.1% of pipeline)

€773,000 (32.1% of pipeline) hasn't had a human touch in over three weeks, meaning almost a third of this pipeline could have already died, been lost to a competitor, or been quietly closed by the buyer walking away — and this CRM would show no evidence of it either way. Two of these (D-18 Kettlemoor and D-27 Corvane) are not stage-overrun — they're proof that "recent stage entry" and "actively managed" aren't the same fact.

4. Duplicate-looking records — 2 clusters, 4 deals, €187,000 combined

Cluster A — "Bramwell Group"

Same normalized name ("Ltd" stripped) and identical domain (bramwellgroup.com), logged separately by two reps, both still open.

Cluster B — "Thackeray & Finch"

These matched on identical contact domain (thackerayfinch.com), not on name-string similarity alone — punctuation-stripping turns "Thackeray & Finch" into "Thackeray Finch" and "Thackeray and Finch Co" into "Thackeray and Finch," which are close but not an exact match. Domain is what confirms this is one account being worked twice, one rep still in Discovery while another already sent a Proposal — worth flagging as a process gap on its own, independent of the hygiene cost.

Two reps, two deal records, two different valuations of the same relationship. This isn't just clutter — €187,000 is currently counted twice toward pipeline total on the optimistic reading. On the conservative reading (one real opportunity per account, valued at whichever record is smaller), €78,000 of stated pipeline doesn't exist — it's the same deal, counted once too often.

5. Unforecastable — 3 deals, missing a field this audit needs to work at all

These three cannot be scored on any of the four tests above — not "assumed clean," genuinely unscoreable. D-09 is the more serious of the three: a deal that's apparently far enough along to have a next-step note ("Discovery call scheduled Sept 2") but no dollar figure means whatever total pipeline number gets reported this week is missing at least one deal's worth of upside entirely, not just misjudging it.

Honesty check

• Staleness cadence: not supplied, defaulted to 21 days.
• Stage benchmarks: not supplied, inferred from the median days_in_current_stage of the deals currently sitting in each stage — this is a proxy for "how long deals typically stay in this stage," not a measurement of it, because a snapshot of open deals only shows time spent so far, not full stage-to-stage duration from closed history. Negotiation's median (11 days) is built on only 5 deals after excluding D-19 for its missing stage-entry date — thinner evidence than the other three stages and worth treating as approximate.
• 23 of 35 deals (66%) trip at least one of the four scored hygiene tests; adding the 3 unforecastable records, 26 of 35 (74%) have some problem. 9 deals (26%), worth €552,000, are clean.
• Three deals sit within a hair of the stage-overrun gate (ratios 1.25–1.30) and would flip on the next weekly run without anyone doing anything differently — worth a standing watch, not a flag today.
• What to start capturing for a sharper run next time: actual closed-deal stage-duration history (not just current open-deal residency), so the overrun benchmark stops being a proxy — and a required account-domain field at deal creation, which would have caught both duplicate clusters automatically instead of needing a manual audit to surface them.

One thing worth flagging back to the founder

The most useful number in this report isn't any single category — it's that 44.6% of pipeline value has no logged next step. Every other test here (overrun, stale, duplicate) measures a symptom of neglect that's already visible elsewhere in the record. A missing next-step is different: it's the one field that directly answers "what happens to this deal next month," and on this file, the honest answer for €1,072,000 of it is "nobody knows." A forecast built on this pipeline isn't wrong by a little on nearly half its value — it's undefined.