---
idea: Cost_to_Job_Margin_Guard
date: 2026-08-17
score: 63
verdict: proceed-to-test
report_version: 2
---

# Cost_to_Job_Margin_Guard Validation Report

> **Version 2, same day.** Version 1 scored this idea 78. The runbook then gained a required community sweep (Phase 1 Step 1b) with a hard 24-month freshness bar on community evidence and a resolution check on load-bearing stale complaints. Running those steps dropped the score to 63. Two things caused the fall: the community sweep found no datable buyer complaint anywhere in the last two years, and the stale complaint that carried version 1's headline argument failed its resolution check. The verdict is still proceed-to-test, but it is now three points above the band boundary rather than eighteen. Section 7 records what changed and why.

## 1. Verdict up front

Score 63 out of 100, which falls in the "above 60: recommend proceeding to test" band. **The total lands within 3 points of the 40-to-60 boundary, so this is flagged as required: any single further factor downgrade flips the verdict to refine.** Buyer reach falling to 1 gives 57, price clue falling to 2 gives 58, pain falling to 0 gives 55. The score is as marginal as it can be while still reading proceed.

The single strongest reason to proceed anyway: one dated, role-identified complaint from exactly the target buyer, and a manual pilot that costs nothing to run. A construction managing director at a 2 to 10 employee firm, on 25 March 2026, says an incumbent system "completely missed supplier invoices, or at best sent them across incorrectly". Against that, the £149 paid audit can be delivered by hand in five days, so the cost of finding out is a page and a fortnight rather than a build.

The recommendation does not depart from the banding, but three findings need to be read before anyone acts on it.

**First, the community sweep found nothing fresh.** Four platforms where this buyer talks to peers do actively discuss job costing and supplier costs (Screwfix Community, Electricians Forums, UK Business Forums, AccountingWEB Any Answers), and the failure mode is named there in the buyers' own words. But every one of those threads is fetch-blocked and undatable from search snippets, so under the runbook's hard 24-month bar none of it can feed a factor score. Reddit could not be reached at all: the search tool refuses the domain outright, so the runbook's "Reddit plus one niche forum" requirement was met on the forum side only. An absence of datable complaint is not proof of an absence of pain, but it is the opposite of the density the pain anchor asks for.

**Second, the stale complaint that carried version 1 failed its resolution check.** Version 1's headline quote was a February 2022 Simpro review: "receipts of goods purchased not going to jobs all the time ... the actual cost is not recouped in invoicing". That is load-bearing, so the runbook requires a check on whether the provider has since fixed it. Simpro shipped a Q1 2026 release themed "More Profit Per Job", described as "tighter control over the details that impact profit", with inventory "updating in real-time as materials are used on jobs" (signal 18). That does not prove the complaint is resolved, but it removes the right to score it. The quote now sits in Community voice as historical context.

**Third, the brief's 95.3 is not externally validated.** It comes from this repository's own UK trades market study under a different eight-factor AI-synthesised rubric. Under the runbook's evidence firewall it counts as one `INFERENCE` signal and fed no factor score. The gap between 95.3 and 63 is mostly a rubric difference, but it should not be read as corroboration.

## 2. The idea as understood

Match supplier documents (invoices, receipts, delivery notes) against the accepted quote, the job and the current invoice state for material-heavy trades of 3 to 20 staff, and surface only the exceptions: costs with no job, costs that moved since the quote, duplicates, and costs never recharged to the customer. Outputs are an exception queue and a margin-at-risk report.

- **Deliberately excluded:** no autonomous job allocation, markup, VAT or CIS treatment, accounting posting or customer billing. The owner approves every action.
- **Integrations:** email forwarding plus CSV first; read-only Xero or QuickBooks second; Tradify, Fergus or Simpro connectors and approved write-back later.
- **Switching strategy:** keep every current tool. Start with 30 documents and five jobs, importing references rather than migrating records.
- **Commercial model:** £149 paid audit, £350 to £750 setup, £149 to £249 per month, roughly £0.20 per document overage.
- **Paid validation as briefed:** five paid audits, requiring three conversions, value greater than three times the fee, and 70% or more repeated rules.

Gaps in the brief: no geography is stated (this report assumes the UK, consistent with the source study and this repository); no distribution asset or existing audience is stated; the brief does not say who inside the firm is the buyer, so this report names one from evidence; and the brief does not say whether "margin-at-risk" includes labour, which materially changes the build. This report scopes it to material and supplier cost only.

**Prior work (runbook Phase 1 Step 0).** `@docs/UK_Trades_Construction_AI_Software_Market_Study_2026.html` (13 August 2026) is the origin of this brief. It ranks Cost-to-Job Margin Guard first of eight wedges at 95.3/100, defines the £149 audit offer, and maps twenty incumbents with prices. `@docs/ideas/CIS_First_Invoicing/` is the only prior validation report and covers a different wedge (subcontractor CIS invoicing) for a different buyer, though the two share the Xero and QuickBooks incumbent layer. Per the runbook, the market study is treated as one `INFERENCE` signal (number 15 below); the incumbent prices it lists were re-checked directly against vendor pages before being used, and only the re-checked figures are scored.

## 3. Buyer and trigger

- **Buyer:** the owner or director of a material-heavy trade firm with roughly 3 to 20 staff (electrical, plumbing and heating, roofing, joinery, small building), or the office administrator or bookkeeper who reconciles the supplier account for them. Both roles appear by name in the sharpest complaints found (signals 1 and 2), and the bookkeeper side is where the failure mode is described most precisely (community signal 19).
- **Trigger:** the monthly supplier statement and payment run, when the merchant's statement has to be reconciled against invoices, delivery notes and what was actually charged out. A secondary trigger is the moment of invoicing a completed job, when the owner has to decide what materials to bill.
- **Current workaround:** for most of this market, not software. Peers on trade forums describe pen and paper, Excel for larger jobs, and merchant ordering discipline (a stage or job code in the order reference) rather than a system (community signals 20 to 22). Where software is used it is a job system (Fergus, Tradify, Simpro) that claims to match supplier invoices to jobs, plus Xero or QuickBooks with Projects or tracking categories.
- **Consequence:** the cost lands in the accounts as a merchant lump sum rather than against the job that consumed it, so job profitability is wrong; a cost that rose since the quote is discovered after the customer has been invoiced; and material that was fitted is never billed at all. This is a silent leak rather than a visible failure, which is precisely why it survives, and why the urgency score is low.

**Honest caveat on this section.** No evidence found in this run quantifies the leak in pounds for a named UK trade firm. The nearest quantified statements are vendor content marketing ("margin lives in the match", signal 5) and adjacent-sector hospitality material (signal 6). The size of the prize is currently an assumption, and the paid audit is the right instrument to measure it.

## 4. Evidence log

Confidence key: **High** = page fetched and content read directly. **Medium** = identified through search result summaries only, page not independently fetched; dates unverified.

Type key: `review` = review aggregator, `vendor` = vendor product or pricing page, `institutional` = trade body, regulator or statistical source, `community` = platform where the target buyer talks to peers, `inference` = prior AI synthesis.

Per the runbook, institutional and vendor sources are judged on whether the page is current, not on when the content was first written. Community signals face a hard 24-month freshness bar and are excluded from scoring if undatable. Review signals older than 24 months are flagged and, where load-bearing, resolution-checked.

### Scored signals

| # | Type | Source | Speaker | Behaviour observed | Cost or spend | Evidence date | Confidence | Link |
|---|---|---|---|---|---|---|---|---|
| 1 | `review` | Capterra, Simpro review | Braden B, Managing Director, construction, 2 to 10 employees | Implementation "was meant to reconcile with Xero however it completely missed supplier invoices, or at best sent them across incorrectly" | Failed implementation; amount not stated | 25 March 2026 | High | https://www.capterra.com/p/10529/Simpro-Enterprise/reviews/ |
| 5 | `vendor` | iTrade purchase order guide (UK) | Software vendor writing to UK trades | "Margin lives in the match. Every supplier invoice that never gets checked against an order is margin at risk, whether through price creep, short delivery, or someone else's parts on your account." Also: invoices arrive "priced differently from what was agreed", deliveries land "short and nobody checks", and "Materials get fitted and never billed" | No figures given | August 2026 | High; vendor marketing, not buyer testimony | https://www.itrade.net/post/purchase-order-software-for-trade-businesses-the-complete-uk-guide |
| 6 | `vendor` | Jelly blog, UK hospitality | Software vendor, adjacent sector | Nine named invoice-processing failures including duplicate payments and "no real-time price change alerts"; a product already sells exception detection into UK kitchens | "Over 60% of invoice errors come from manual data entry" | 10 June 2026 | High; adjacent sector, not trades | https://blog.getjelly.co.uk/common-invoice-processing-complaints-uk/ |
| 7 | `institutional` | FMB and CIOB State of Trade, H1 2025, 458 SME responses | Trade bodies surveying SME builders | "75% of builders have seen material prices rise", 67% report higher wages, 61% have increased the prices they charge | Material price rises across three quarters of the target market | Survey H1 2025, published 2 September 2025 | High (fmb.org.uk returned HTTP 403; read via Designing Buildings) | https://www.designingbuildings.co.uk/wiki/FMB%20and%20CIOB%20publish%20State%20of%20Trade%20survey%20findings%202025 |
| 8 | `institutional` | BCIS construction insolvency tracker | BCIS chief economist, citing Insolvency Service data | 3,805 construction firms insolvent in the year to June 2026; 309 in June alone, 16% of all insolvencies in England and Wales; specialised construction activities the largest single group at 173 | Business failure | Published 17 July 2026 | High | https://www.bcis.co.uk/news/construction-insolvencies-latest-news/ |
| 9 | `vendor` | Fergus UK supplier integrations page | Incumbent vendor | Supplier invoices "sent straight to Fergus, automatically importing line items and matching to the right jobs", with splitting across multiple jobs and phases; included in paid plans, auto email ingestion requires Pro | Included in £39 to £45 per month plan | Current, checked 17 August 2026 | High | https://fergus.com/uk/integrations/suppliers/ |
| 10 | `vendor` | Fergus UK pricing page | Incumbent vendor | Essentials from £39 per month per company with manual supplier invoices; Professional from £45 per month adds automatic supplier integrations and purchase orders; timesheet users from £12 | £39 to £45 per month plus £12 per timesheet user | Current, checked 17 August 2026 | High | https://fergus.com/uk/pricing/ |
| 11 | `review` | Capterra UK, Fergus profile | Aggregator plus a construction reviewer | 4.6 from 155 reviews, listed at £35 per user. Sharon, construction, 26 May 2025: "integration with xero and suppliers works [r]easonably well" | £35 per user shown | 26 May 2025 | High | https://www.capterra.co.uk/software/155571/fergus |
| 12 | `vendor` | Planyard pricing page | Direct competitor | Construction cost control with purchase invoice OCR, budget versus actual and Xero/QuickBooks sync; Essential $105 per project manager per month, Professional $158, Ultimate $210, support staff $43; supplier invoices absent from the $52 Estimating tier | $105 to $210 per project manager per month | Current, checked 17 August 2026 | High | https://planyard.com/pricing |
| 13 | `vendor` | Planyard job costing article | Direct competitor writing to contractors | Argues Xero tracking dimensions are limited for project plus job plus phase plus location, that option lists "balloon" and reporting becomes "messy", and that construction job costing "is typically driven by invoices and commitments (RFQs, POs, subcontracts, progress claims), not just time entries" | No figures | Published 10 September 2025, updated 18 February 2026 | High | https://planyard.com/blog/tracking-categories-in-xero-for-contractors-the-practical-setup-for-job-costing |
| 14 | `vendor` | Lentune knowledge base | Direct competitor | Sells three-way invoice matching that reconciles supplier invoices against purchase orders inside Simpro, also integrating Xero, Jobpac and Procore | Pricing not published | Current, checked 17 August 2026 | High | https://www.lentune.com/knowledge/invoice-matching-with-simpro-job-management-system |
| 15 | `inference` | Prior internal research, this repository | AI-assisted synthesis, `INFERENCE` | Ranks this wedge first of eight at 95.3/100 under a different eight-factor rubric; defines the £149 audit, £350 to £750 setup, £149 to £249 per month model; maps 20 incumbents | £149 audit; £149 to £249 per month proposed | 13 August 2026 | Recorded as prior synthesis, not primary evidence | `@docs/UK_Trades_Construction_AI_Software_Market_Study_2026.html` |
| 16 | `community` | Screwfix Community Forum | UK trades community | Forum reports approximately 106,868 members. Used here only as a channel-size figure for buyer reach, not as pain evidence | Not applicable | Current figure, unverified | Medium | https://community.screwfix.com/ |
| 17 | `community` | UK Business Forums | UK small business community | Reports 277,200+ business owner members with a dedicated tradespeople area. Channel-size figure only | Not applicable | Current figure, unverified | Medium | https://www.ukbusinessforums.co.uk/ |
| 18 | `vendor` | Simpro Q1 2026 release announcement | Incumbent vendor | Release themed "More Profit Per Job": "tighter control over the details that impact profit", AI-driven cash flow, inventory "updating in real-time as materials are used on jobs". Recorded as the required resolution check on stale signal 2 | Not stated | Q1 2026 | Medium (simprogroup.com returned HTTP 403; read via search summary) | https://www.simprogroup.com/company/press/simpro-q1-2026-more-profit-per-job |
| 19 | `vendor` | EzzyBills for Simpro | Direct competitor | Sells AI invoice scanning with purchase-order matching into Simpro, including handling suppliers that use a single item code | Pricing not established | Current, checked 17 August 2026 | Medium (page returned empty to the fetcher) | https://www.ezzybills.com/scan-invoices-into-simpro/ |

**Signals demoted out of scoring in version 2.** Signal 2 (Capterra, Simpro, office administrator, 23 February 2022: "receipts of goods purchased not going to jobs all the time ... the actual cost is not recouped in invoicing"), signal 3 (Software Advice, Simpro, undated: "we were told that the only solution was for the supplier to change their invoices to suit Simpro"), signal 4 (Capterra, Tradify, sole trader, 2 February 2021: "importing supplier statements against invoices and jobs could be improved") and the July 2018 Fergus fragment ("we never miss charging on materials"). All are older than 24 months or undatable. Signal 2 was load-bearing in version 1 and failed its resolution check against signal 18. They appear below as historical context only.

### Community voice

The runbook requires the buyer's own recurring words for the problem. **Every community source below is fetch-blocked and undatable from search snippets, so all of it is historical context that may be resolved, and none of it fed a factor score.** Reddit could not be reached at all: the search tool refuses the reddit.com domain, so the "Reddit plus one niche forum" requirement was met on the forum side only. This limitation is itself a finding, and it is the main reason pain evidence and urgency fell in version 2.

**Communities observed actually discussing the problem** (four platforms, roughly fifteen distinct threads surfaced):

| Platform | What is discussed there | Promotional posting |
|---|---|---|
| Screwfix Community Forum | Job costing software, pricing up jobs, markups, software for running a small construction business, clients asking for copies of materials invoices | Not established; rules unchecked |
| Electricians Forums | "Anybody using Job Management Software?" (three pages), "Does anyone use a good job tracking software?", quotation and costing spreadsheets | Not established; rules unchecked |
| UK Business Forums | "What apps are trades using for quoting jobs now?", construction project management software, accounting software alongside self-bookkeeping | Not established; rules unchecked |
| AccountingWEB Any Answers | Repeated "job costing software for a construction firm" questions from accountants serving trade clients | Professional community; pitching likely unwelcome |
| BuildHub | Self-builders and renovators, not trade firms | **Explicitly forbids advertising:** "no advertising of goods or services is allowed" |
| Reddit | Could not be assessed | Not reachable by this agent |

**The problem in buyers' and their accountants' own words** (historical, may be resolved):

- *"if you have a supplier invoice relating to more than one job, many systems can't handle it"* (AccountingWEB Any Answers). This is the failure mode stated more precisely than any vendor states it, and it is the single most useful phrase found for smoke-test copy.
- *"checking for postings unallocated to jobs in the 'No Name' column"* (AccountingWEB, describing QuickBooks). "Unallocated postings" and "No Name column" are the buyer-side vocabulary for what this product calls an exception queue. Note that this also means the report partly exists already, for free, inside QuickBooks.
- *"receipts of goods purchased not going to jobs all the time ... if you do not activate updates the actual cost is not recouped in invoicing"* (Capterra, Simpro, office administrator, February 2022). Version 1's headline quote, now historical and resolution-checked against signal 18.
- *"stage codes that are used as part of the order reference when ordering materials"* (Screwfix Community). The free process fix, described by tradespeople as normal practice.
- Software framed by its ability to ensure "better tracking of materials ensures all charges are captured" (Electricians Forums). The benefit language the market already uses.

**Notable objections heard** (historical, may be resolved, and all disconfirming):

- *"stay away from software options unless tendering for multi-trade works like extensions"* (Screwfix Community). Peers actively advising against the category.
- *"most tradespeople don't use mobile apps or software"* (Screwfix Community).
- Excel is described as the popular tool for breaking down costs on larger jobs (Screwfix Community).
- Price anchoring in buyers' own words: simPRO Office quoted at "£104 per month", the common products "around £45 per month", basic systems "around £50 a month" (Electricians Forums). Nobody in these threads discusses paying £149 to £249 per month for a monitoring layer.
- Adoption is slow: "it typically takes 12-24 months to fully realize the benefits" (Electricians Forums).
- A cancellation complaint: on cancelling, the account was deleted immediately, the paid month was lost and historical data became inaccessible (Electricians Forums). Relevant to how a data-holding product must behave.

**What the sweep did not find, despite looking:** no dated thread in the last 24 months in which a UK trade owner describes losing money to an unmatched or unbilled supplier cost; no thread quantifying that loss; no thread asking for a product like this.

**Budget note.** Roughly 70 minutes of searching across both versions, and 19 numbered signals plus the community sweep. Version 2 spent its budget on Step 1b and the resolution check. Sources that refused automated fetching: community.screwfix.com (TLS certificate failure), electriciansforums.net (403), ukbusinessforums.co.uk (403), accountingweb.co.uk (403), simprogroup.com (403), uk.trustpilot.com (403), fmb.org.uk (403), community.xero.com (DNS failure), apps.xero.com (timeout). reddit.com is refused by the search tool itself.

## 5. Kill list

| # | Disconfirming finding | Seriousness | Source |
|---|---|---|---|
| 1 | **The incumbent already claims to do exactly this, for free within a plan the buyer is already paying for.** Fergus sends supplier invoices "straight to Fergus, automatically importing line items and matching to the right jobs", splits them across jobs and phases, and includes it in every paid plan at £39 to £45 per month. That is a tenth of the proposed subscription. | **High.** Forces the wedge to be about exceptions the incumbent does not surface, and about firms with no job system at all, not about matching per se. | https://fergus.com/uk/integrations/suppliers/ and https://fergus.com/uk/pricing/ |
| 2 | **Three live direct competitors already occupy the space.** Planyard sells purchase invoice OCR plus budget versus actual from $105 per project manager per month; Lentune sells three-way matching into Simpro, Xero, Jobpac and Procore; EzzyBills sells AI invoice scanning with purchase-order matching into Simpro. | **High.** The proposed £149 to £249 per month sits in a band already contested by products with more build behind them. | https://planyard.com/pricing ; https://www.lentune.com/knowledge/invoice-matching-with-simpro-job-management-system ; https://www.ezzybills.com/scan-invoices-into-simpro/ |
| 3 | **The community sweep found peers advising against the whole category.** Tradespeople tell each other to "stay away from software options unless tendering for multi-trade works", note that "most tradespeople don't use mobile apps or software", and recommend Excel. The price anchors they quote to each other are £45 to £104 per month, well below the proposed subscription. | **High.** This is new in version 2 and it is the most direct evidence available about how the offer will land. It argues that the buyer's default answer to any new subscription is no. | Electricians Forums and Screwfix Community, via search snippets (Medium, undatable) |
| 4 | **Part of the exception report already exists free inside QuickBooks.** Accountants describe running Profit and Loss by Job and reading unallocated postings from the "No Name" column. The unmatched-cost half of the product is a report a bookkeeper can already run. | **Medium to high.** Narrows the defensible wedge to changed-since-quote and never-recharged, which are the harder halves to compute. | AccountingWEB Any Answers, via search snippets (Medium, undatable) |
| 5 | **The stale complaint that carried version 1 failed its resolution check.** The February 2022 "receipts not going to jobs ... cost not recouped in invoicing" review is the clearest statement of the product thesis found anywhere, and Simpro has since shipped a Q1 2026 release themed "More Profit Per Job" with real-time inventory updates as materials are used. The complaint may simply be fixed. | **Medium to high.** Does not prove resolution, but removes the right to score the complaint and is why pain evidence fell from 2 to 1. | https://www.simprogroup.com/company/press/simpro-q1-2026-more-profit-per-job (Medium) |
| 6 | **The pain is silent, so it may not be bought.** No dated evidence found anywhere, including four community platforms, of a UK trade firm quantifying its own leakage or searching for a fix. Contrast the CIS wedge, where a deadline and a penalty force action. | **Medium.** A leak nobody has measured is hard to sell against, which is exactly why the £149 paid audit is the correct first offer: it converts an invisible problem into a number before asking for a subscription. | Absence of evidence across both versions |
| 7 | **A free process fix exists and is already normal practice.** Tradespeople describe using stage or job codes in the merchant order reference so materials tally to jobs and anything added in error is visible. Discipline at the point of order removes much of the need for detection afterwards. | **Medium.** The counter is that this depends on every operative every time, which is what fails in a 3 to 20 staff firm. | Screwfix Community, via search snippets (Medium, undatable) |
| 8 | **Financial data sensitivity raises the cost of the first yes.** The offer asks a firm to hand over 30 supplier documents, five quotes and its job list, exposing supplier pricing and margins to a stranger. A community cancellation complaint (account deleted immediately, historical data lost) shows this market has been burned on data custody before. | **Medium.** A conversion risk on the smoke test: a weak result may reflect reluctance to share rather than absence of pain. | Electricians Forums, via search snippets (Medium, undatable) |
| 9 | **One relevant community explicitly forbids pitching**, and the others' rules are unchecked. BuildHub states "no advertising of goods or services is allowed". AccountingWEB is a professional community where a pitch would likely be unwelcome. | **Medium.** Directly constrains the Phase 3 channel choice and is why buyer reach stays at 2 rather than rising. | https://forum.buildhub.org.uk/ |
| 10 | **Searches for failed competitors returned nothing.** Searched for shut-down or failed UK supplier-invoice-matching and job-costing products for trades across 2024 to 2026. No corpses found; Fergus, Tradify, Simpro, Planyard, Lentune, EzzyBills and Buildxact are all still trading. Nobody has burned publicly, and nobody has broken out. | **Low**, and ambiguous in both directions. | Searches run 17 August 2026 |

**Searches that returned nothing usable, recorded as required:** no public churn or cancellation data for job-costing supplier-invoice modules; no dated thread of a UK trade owner explaining why they abandoned a job-costing tool and returned to spreadsheets; no Reddit evidence of any kind, the domain being refused by the search tool; no published figure for the value of material cost leakage in UK trade firms of any size; no dated community thread in the last 24 months describing an unmatched or unbilled supplier cost.

## 6. Competitor and incumbent map

| Option | Price (checked 17 August 2026 unless noted) | What it does about this problem | What buyers complain about |
|---|---|---|---|
| **Fergus** | £39 Essentials, £45 Professional per month per company; £12 per timesheet user; Capterra UK lists £35 per user | Auto supplier invoice import with line items, matching to jobs, splitting across jobs and phases | Job file management gaps; margins on materials "fiddly" to change at invoicing (2018, historical) |
| **Simpro** | Quote only; buyers quote each other roughly £104 per month for the Office package | Supplier invoices, purchase orders, job costing, Xero sync; Q1 2026 release themed "More Profit Per Job" | The sharpest complaints found, though only one is in-window: supplier invoices missed or sent across incorrectly to Xero (March 2026); receipts not going to jobs and cost not recouped in invoicing (February 2022, historical) |
| **Tradify** | £34 to £44 per user per month (prior study; not re-verified) | Job costing, quoted versus actual comparison, Xero and QuickBooks sync | Importing supplier statements against invoices and jobs "could be improved" (2021, historical) |
| **Planyard** | $105 to $210 per project manager per month | Purchase invoice OCR, approval workflow, budget versus actual, forecast, Xero and QBO sync | Not established in this run |
| **Lentune** | Not published | Three-way match of supplier invoices against purchase orders inside Simpro | Not established in this run |
| **EzzyBills** | Not established | AI invoice scanning with purchase-order matching into Simpro, including single-item-code suppliers | Not established in this run |
| **QuickBooks Projects** | Inside a QuickBooks Online Plus or Advanced subscription | Job costing, and Profit and Loss by Job with unallocated postings visible in the "No Name" column | The unmatched-cost report largely exists here already, free to anyone who knows to run it |
| **Xero Projects or tracking categories** | Inside an existing Xero subscription | Bills and spend assigned to a project, or every line tagged with job and cost code | Tracking dimensions limited across project, job, phase and location; option lists balloon; reporting becomes messy |
| **Dext or similar capture tools** | Approximately £25 to £30 per month for around 250 documents (Medium confidence) | Captures the document and codes it | Does not answer which cost changed or was never recharged: the gap this product targets |
| **Do nothing:** merchant statement, Excel, job codes in the order reference, memory | Owner and administrator time | Monthly reconciliation; job codes at the merchant make materials tally and flag anything added in error | Cost lands as a merchant lump sum, not against the job; skipping reconciliation for months leaves discrepancies that take days to untangle. This is what most of the market actually does, and peers recommend it |

The competitive shape is a barbell with a crowded middle. Below the proposed price, Fergus and Tradify bundle matching into a job system the buyer may already own, and QuickBooks already reports unallocated postings. Above it, Planyard, Lentune and EzzyBills sell the deeper cost-control layer. The proposed £149 to £249 per month sits between them without a job system underneath, at three to five times the price the buyer's own peers quote to each other. The switching strategy ("keep every current tool") is what makes that position defensible rather than merely squeezed, and it is the central commercial question the test must probe.

## 7. Score breakdown

| Factor | Weight | v1 | v2 | Anchor met and justification |
|---|---|---|---|---|
| Pain evidence | 25 | 2 | **1** | After the freshness discipline, exactly one dated in-window buyer complaint survives (signal 1, March 2026), and no speaker anywhere quantifies time or money. The community sweep across four platforms produced no datable complaint at all. Version 1's supporting complaints are older than 24 months, and the load-bearing one failed its resolution check against Simpro's Q1 2026 release. That is the weak anchor's description: occasional grumbles, no cost mentioned. Genuinely torn against 2; scored down per the tie-break. |
| Urgency | 20 | 2 | **1** | The strong anchor needs a cost, penalty or deadline on the day the problem occurs, and this leak is silent by construction. The community sweep showed peers handling it with Excel, pen and paper, and job codes at the merchant, which is literally the weak anchor: an annoyance handled whenever convenient. The monthly statement run is a real cycle, which is why this was 2 in version 1, but no evidence shows anyone treating it as pressing. Scored down per the tie-break. |
| Buyer reach | 20 | 2 | **2** | Unchanged, and better evidenced than before. Named channels with published figures exist (Screwfix Community approximately 106,868; UK Business Forums 277,200+) and the sweep confirmed all four platforms actively discuss this problem, which is what the anchor is really testing. It stays at 2 rather than 3 because the figures are search-mediated, the target slice is a minority within communities dominated by sole traders, one relevant community explicitly forbids advertising, and the others' rules are unchecked. |
| Manual pilot | 20 | 3 | **3** | Strong anchor met without qualification. Thirty documents, five quotes and a job list returned as a checked leakage report in five working days is a spreadsheet, a PDF reader and attention. No software is required to deliver the first paid outcome. This is the factor that keeps the idea worth testing. |
| Price clue | 15 | 3 | **3** | Strong anchor met: verified current spend at Fergus (£39 to £45 per month), Planyard ($105 to $210 per project manager), Dext (approximately £25 to £30) and quote-only Simpro. These are vendor pricing pages, judged on currency rather than age, so the freshness bar does not touch them. The community sweep corroborates the band from the buyer side (£45 to £104 per month). Held at 3 because the £149 audit undercuts remedial accountancy time and the subscription reframes a per-seat licence as managed monitoring, but note that no evidence anywhere shows this buyer paying £149 to £249 per month for a monitoring layer. |

Total: `round((1×25 + 1×20 + 2×20 + 3×20 + 3×15) / 3)` = `round(190 / 3)` = **63**.

Version 1 total was `round(235 / 3)` = 78. The 15-point fall comes entirely from pain evidence and urgency, both dropped one level after the community sweep and the resolution check. No factor was adjusted to move the total across a boundary; the boundary proximity is a consequence, not a choice.

**Boundary flag (required).** 63 is within 3 points of the 60 boundary. Any single further downgrade flips the verdict to refine: buyer reach to 1 gives 57, price clue to 2 gives 58, pain to 0 gives 55. The factor most likely to tip it is **price clue**, because it is the one strong-anchor score resting on an "or reframes" reading rather than a straightforward undercut, and because the community price anchors (£45 to £104 per month) sit well below the proposed subscription.

## 8. Proposed test design

- **Offer (smoke-test page copy direction):** lead with the buyers' own words rather than the product's. "Some of your supplier invoices cover more than one job. Which ones never reached a customer?" For material-heavy UK trade firms of 3 to 20 staff: send 30 supplier documents, five accepted quotes and your open job list. Within five working days you get back a checked leakage report naming every cost with no job, every cost that moved since the quote, every duplicate, and every material that was fitted and never billed, with the pound value of each. £149, credited against setup if you continue. Positioned honestly as a done-by-hand audit, not software.
- **Copy note from the community sweep:** use "unallocated", "not on a job" and "never charged out" rather than "exception queue" or "margin at risk". Expect and pre-empt the two objections peers actually voice: that this is what Excel and a job code at the merchant are for, and that job software already claims to do it.
- **Single commitment action:** pay the £149 audit fee and send the 30 documents. This is deliberately the strongest commitment layer available (behavioural, paid), and it is the one the brief specifies. An application without payment and documents does not count as qualified. A softer secondary action (upload three supplier invoices for a free sample check) may be offered as a step towards it, but only paid audits count against the gate.
- **First distribution channel:** trade-focused accountants and bookkeepers who already see these firms' merchant statements, asked to refer two or three clients each. The community sweep supports this choice and argues against forum-first distribution: AccountingWEB shows accountants repeatedly asking how to solve exactly this for construction clients, while the trade forums show peers advising each other away from software. Forum posting, if used at all, must follow a rules check per community: BuildHub forbids advertising outright, and the Screwfix, Electricians Forums and UK Business Forums rules are unchecked. Distribution remains a human action per the runbook.
- **Pre-committed decision gate:** the runbook default is adjusted upward in commitment strength and downward in volume, with reasoning. Because the commitment action is a £149 payment rather than a form fill, and because the target is a narrow slice, the gate is **5 paid audits sold within roughly 50 qualified visits over 10 working days**. From those five audits, the briefed thresholds must then hold: **3 conversions to setup or subscription, an identified leakage value greater than three times the fee (over £447 per firm), and 70% or more of the exception rules recurring across firms**. If fewer than 5 audits sell but 2 or more do, that is a refine signal on price or channel, not a stop.

These thresholds are copied verbatim into the Phase 3 gate record before any distribution begins. This report is their canonical source.

## 9. Risks and unknowns

- **The size of the leak is unmeasured, and now known to be undiscussed.** No source found across two versions and four community platforms quantifies material cost leakage for a UK trade firm, or shows one complaining about it in the last 24 months. If the typical leak is £200 a month, a £149 to £249 subscription cannot be justified. The audit measures this, and the ">3x fee" threshold is the right test.
- **Willingness to pay £149 to £249 per month is contradicted by the only buyer-side price anchors found.** Peers quote each other £45 to £104 per month for full job systems. This is the primary commercial risk.
- **The evidence base may be an artefact of what this agent can reach.** Reddit is refused outright and every relevant forum 403s. The pain could be loudly and recently discussed in places that were simply invisible here. Anyone with forum accounts should spend an hour reading the named threads directly before accepting the pain and urgency scores; that single hour could move the score by 15 points in either direction.
- **The complaint may already be fixed.** Simpro shipped a 2026 release aimed at profit per job. If 30 documents from a current Fergus Professional or Simpro customer yield no exceptions, the wedge is dead for that segment and the audit will show it in week one, cheaply.
- **Part of the product is a free QuickBooks report.** The unmatched-cost half can be approximated by Profit and Loss by Job and the "No Name" column, so the paid version must earn its keep on changed-since-quote and never-recharged.
- **Trust and data sensitivity may suppress the commitment action,** and this market has a recorded grievance about a vendor deleting an account and its history on cancellation. A weak result may reflect reluctance to share rather than absence of pain, so the application must record why anyone declines.
- **False matches are commercially dangerous.** Confidence thresholds, audit logs and owner approval are build requirements, not refinements, which raises the floor on build cost above what "medium build" implies.
- **The £0.20 per document overage is untested** and no evidence was found about document volumes at this firm size, so plan allowances cannot yet be set.
- **The test will not settle:** long-term retention, whether the exception rules generalise beyond the first five firms' suppliers, whether merchant or job-system connectors are obtainable on reasonable terms, or whether the wedge works outside the UK.

## 10. Recommendation and next step

Score 63: recommend proceeding to test, but this is a marginal proceed, three points above the refine band, and it should be read as "the cost of finding out is low" rather than "the evidence is strong". The evidence is not strong. One dated complaint, no quantified loss, no community demand, three live competitors and an incumbent that bundles the core function at a tenth of the price. What carries it is that the manual pilot is genuinely free to run and the price clue is real.

The human is asked to approve **Phase 3 artefact creation only**: a smoke-test page carrying the £149 audit offer in section 8, the application and eligibility form questions, the decision gate record with the thresholds above written down before anything is published, and a ten-day plan against real dates. Distribution, including reading and complying with the posting rules of any community, will require separate explicit approval. Nothing has been built in this run.

Two alternatives are worth naming, because at 63 they are close to equally defensible:

- **Refine, evidence-first.** Spend one hour reading the named forum threads with a human account (Screwfix Community, Electricians Forums, UK Business Forums, AccountingWEB) before building anything. The pain and urgency scores rest on an absence this agent could not distinguish from a fetch failure. That hour could move the score by 15 points either way and is cheaper than a page.
- **Refine, offer-first.** Run three unpaid sample checks with firms already known to you, to measure the leak before asking anyone to pay £149 for a number that might be small. This follows from kill list findings 1, 5 and 6.

Proceed, refine, park or stop.

## 11. Decision log

| Date | Decision | Decided by | Reason |
|---|---|---|---|
| | | | |