ESOS Phase 4 Evidence Pack from Supplier Energy Invoices

An ESOS Phase 4 evidence pack from energy invoices is a 12-month record of verifiable energy consumption — measured in kWh per fuel type, traceable back to supplier invoices and meter records — that a qualifying UK organisation must hold by the 5 December 2027 compliance deadline. The pack also has to carry intensity ratios for buildings, transport, and industrial uses, and a written progress narrative against the Action Plan commitments submitted in Phase 3.

GOV.UK's Energy Savings Opportunity Scheme guidance sets the 5 December 2027 Phase 4 compliance deadline, with qualifying UK organisations required to notify the Environment Agency of their compliance and to base their evidence on 12 months of verifiable data. Qualification is defined by Companies Act 2006 size thresholds applied at 31 December 2026: more than 250 employees, or more than £44m turnover combined with more than £38m balance sheet total. Roughly 10,000 organisations fall inside that line.

The fields the evidence pack needs map cleanly onto what a supplier energy invoice already carries. Total energy consumption in kWh per fuel type is the sum of billed kWh across every meter for the 12-month reference period. The source data references the auditor will ask for — invoice number, billing period, MPAN for electricity, MPRN for gas, supplier name — sit on the face of each bill. Intensity ratios are new in Phase 4: a numerator of building-, vehicle-, or process-level kWh against a denominator drawn from the organisation's own operational data (floor area, vehicle-kilometres, units produced). The Action Plan progress narrative, also new this phase, leans on baseline-versus-current invoice data to show whether the Phase 3 commitments were met.

Which regulator the organisation notifies depends on where it is registered. The Environment Agency handles England; the Scottish Environment Protection Agency (SEPA) handles Scotland; Natural Resources Wales (NRW) handles Wales; the Northern Ireland Environment Agency (NIEA) handles Northern Ireland. The compliance requirement and evidence shape are the same across all four; only the notification target differs.

ESOS is sometimes confused with the Streamlined Energy and Carbon Reporting (SECR) regime, which captures a different "large company" definition, runs annually, and lands its disclosure inside the directors' report. SECR matches the financial year; ESOS pins to a fixed reference window that has to include the 31 December 2026 qualification date. The two regimes share an invoice substrate but produce different deliverables — academy trusts and other SECR filers handling the annual disclosure can extract academy trust utility bills to Excel for SECR reporting using a similar field shape but a different reporting cadence.

One boundary to set before the operational walkthrough begins: this article covers the invoice-to-evidence-pack data assembly. The ESOS assessment itself — the appointment of a lead assessor, the site visits, the audit-quality verification, and the final sign-off — must be carried out by an accredited ESOS lead assessor registered with one of the approved professional bodies. The data work described in the sections that follow is the substrate the lead assessor signs off on, not a substitute for that sign-off.

---

What each line on a UK energy invoice yields

A UK B2B energy invoice carries a defined set of fields that map directly onto the evidence pack. The extraction shape per bill is:

• Billing period start and end dates. These bound the kWh figure to a specific window so it can be apportioned to the reference period.
• MPAN for electricity supplies, MPRN for gas. These are the supply identifiers — stable across supplier switches, unique per meter, and the join key from the invoice back to the building.
• Consumption in kWh for the billing period. This is the field that aggregates into the evidence pack's per-fuel total.
• Unit rate (pence per kWh) and standing charge (pence per day). Useful for cost analysis and for cross-checking the kWh figure against the billed amount.
• CCL line item. Climate Change Levy is charged separately on kWh, not on cost.
• VAT rate applied to the supply. Most UK B2B electricity and gas attracts 20% VAT, with a reduced 5% rate applying where supply qualifies as low-usage or non-business.
• Invoice number and supplier name. The source data reference an auditor will ask for.
• Estimated-versus-actual reading flag against each meter read on the bill.

The estimate flag is the field most often overlooked at extraction time and the most expensive to retrace later. ESOS expects verifiable data; an estimated-reading invoice needs either a subsequent true-up against an actual read or an explanatory note in the methodology. Capturing the flag while the invoice is in front of you costs nothing; reconstructing it across 200 bills three weeks before the assessor's deadline costs days. The flag appears as a one-letter code or short word alongside each meter read.

UK B2B invoices from British Gas Business, EDF Energy, SSE Business Energy, ScottishPower, Engie, Total Energies, npower Business, and the smaller specialist suppliers vary in layout but converge on those same fields. In practice the variation matters in three places:

• MPAN placement. Some suppliers print the full 21-digit MPAN (with the S-number prefix) in a header block; others list only the 13-digit core MPAN against each supply line on a multi-meter invoice. The 13-digit core is the bit that joins back to the building register — the prefix carries profile class, time pattern regime, and line loss factor information that is useful for energy management but not required for the ESOS source reference.
• Multi-meter invoice structure. SSE Business Energy in particular issues consolidated invoices with each MPAN appearing as its own sub-section, with its own kWh, reads, unit rate, standing charge, and CCL line. Extraction has to descend to the sub-section level — taking the invoice-level total alone collapses the per-meter detail the evidence pack needs.
• CCL position. CCL almost always sits as a distinct line item rather than being absorbed into the unit rate, but its placement on the bill (in the charges table, in a separate environmental-charges block, or as a footnote calculation) shifts between suppliers. A useful sanity check is to confirm the CCL kWh matches the billed kWh on the same line; a mismatch usually signals a partial-period CCL adjustment or a charity-exemption issue worth flagging.

For half-hourly-metered electricity supplies, the invoice still carries a billed kWh figure, but the verifiable substrate sits in the half-hourly data file held by the meter operator. The evidence pack can be built from the invoiced kWh for most reporting purposes; where an assessor wants to interrogate the underlying readings, the HH file is what gets produced. If a portfolio includes HH-metered supplies, the data request to the meter operator (MOP) goes alongside the invoice extraction rather than after it.

Aggregating across meters and sites

For a single-site organisation the per-meter and the portfolio total are the same figure. For everyone else, the evidence pack lives at the top of a three-level aggregation: each MPAN or MPRN is the atomic unit; meters group to a building or site; sites group to a portfolio total per fuel type. The line that goes into the evidence pack as "total grid electricity for the 12-month reference period" is the sum of every electricity-MPAN's reference-period kWh; the line for "total mains gas" is the same shape across MPRNs. The auditable detail, though, lives at the meter level — the assessor's spot-check question is "show me the invoices behind MPAN 2000023456789 for the reference period", not "show me the portfolio total".

MPAN and MPRN are the natural join keys back to the building. MPAN is the supply identifier for an electricity connection — 13 digits in its core form, 21 digits when the full S-number prefix is shown — and it stays the same when the organisation switches supplier, renegotiates the contract, or changes the customer account reference. MPRN performs the same role for mains gas connections, typically 6 to 10 digits, and is equally stable across supplier changes. The supplier's customer account number, by contrast, changes every time the contract changes, and is the wrong field to use as the join key. Building a meter register keyed on MPAN and MPRN — with supplier, account number, and tariff as attributes that can move underneath — is the spreadsheet shape that survives a four-year ESOS cycle.

The multi-supplier reality is the norm rather than the exception. A single building might have one electricity MPAN supplied by EDF and one gas MPRN supplied by British Gas Business. A portfolio of fifty sites might have invoices arriving from four electricity suppliers and three gas suppliers across overlapping but non-aligned billing cycles. Aggregation has to handle that variety without losing the supply-level audit trail, which is why the row shape from the extraction step — one row per MPAN per billing period, with supplier, billing dates, kWh, estimate flag, and source PDF reference all retained — is the right level of granularity to land on. The portfolio totals fall out as sums; the audit trail is preserved underneath.

For multi-site portfolios outside the UK the join key changes but the aggregation pattern carries — Australian portfolios use the National Metering Identifier in the same role MPAN plays here, and the workflow shape is recognisable across regulatory regimes. Teams running both UK and Australian operations can consolidate multi-site Australian utility bills with NMI as the join key on the same principles.

The fuel types that need separate totals for the evidence pack are: grid electricity (sum across MPANs), mains gas (sum across MPRNs), transport fuel (sum across vehicles and fuel cards, converted to kWh), and any other on-site fuel such as LPG, heating oil, or wood biomass (sum from delivery invoices or storage-tank records). Each fuel type has its own kWh column in the evidence pack and its own intensity ratio downstream. Transport fuel rarely arrives in the same shape as grid energy — the data substrate is fuel-card statements, mileage claim records, or driver expense logs rather than utility invoices — so it usually runs as a parallel sub-workflow rather than a column in the main extraction.

The numerator for a building intensity ratio (kWh per building per fuel type or fuel-combined) comes directly from the aggregation. The denominator — gross internal floor area for buildings, or vehicle-kilometres and tonne-kilometres for transport — lives outside the invoice data and comes from the property register, the lease portfolio, or the asset management system. The article focuses on the numerator because that is what the invoices produce; the denominator is a separate piece of work that the energy manager or facilities lead typically already holds.

Pro-rating billing periods to the 12-month reference window

The reference period for Phase 4 has to be 12 consecutive months that include the qualification date (31 December 2026) and end before the 5 December 2027 compliance deadline. Most organisations align the window with whatever operational year is easiest to defend — a calendar year ending 31 December 2026, a financial year ending 31 March 2027, or a 1 July 2026 to 30 June 2027 window for organisations on a July-anchored cycle. The choice is the organisation's to make, provided the window contains the qualification date.

Once the window is set, the boundary problem appears immediately. Invoices arrive on the supplier's billing cycle, not on the reference window's. A quarterly gas invoice running 1 November 2026 to 31 January 2027 straddles a 31 December 2026 reference window end. The bill's total kWh covers 92 days; only the 1 November to 31 December slice (61 days) falls inside the reference period; the 1 January to 31 January slice (31 days) belongs to the next compliance cycle. The pro-rata kWh that goes into the evidence pack for that invoice is 61/92 of the billed kWh.

Three apportionment methods are in regular use, in descending order of accuracy and ascending order of availability:

• Meter-read true-up. A meter read taken at the boundary date itself produces an exact split — kWh from the previous read to the boundary read is in-period, kWh from the boundary read forward is out-of-period. This is the cleanest method and the rarest, because it requires a planned read on the boundary date and most suppliers will not schedule one on request.
• Half-hourly apportionment. For HH-metered electricity supplies, the half-hourly data file from the meter operator gives the exact kWh consumed in each half-hour, which sums to an exact in-period figure regardless of where the invoice's billing period falls. Use this where the HH data is held.
• Days-based pro rata. The default. Multiply the billed kWh by the ratio of in-period days to total days in the billing period. Simple, defensible, and the method an assessor expects to see unless better data is available.

Whichever method is used, each split needs to be recorded on the row of the spreadsheet rather than worked out and discarded. The columns that earn their place are: original invoice billed kWh, total days in the billing period, days inside the reference window, days outside, apportionment method (days / HH / read), and the resulting in-period kWh. An auditor checking the evidence pack should be able to reconstruct any split arithmetic from the row without going back to the original invoice. Storing the workings inline also protects against the spreadsheet being re-opened in 2032 by someone who was not in the room when the choices were made.

A handful of edge cases deserve their own row treatment:

• Monthly invoices with drifting read dates. A monthly bill may run 3 December to 2 January in one cycle and 3 January to 1 February in the next. The reference-window split still works on days; the drift just means the days-in-period calculation has to be done per invoice, not assumed from the month label.
• Annual reconciliation invoices. Some suppliers issue an annual true-up that adjusts prior periods after the fact (most commonly on HH supplies or where a meter read was estimated for several billings in a row). The corrected kWh should be attributed to the original billing periods it relates to, not the date the reconciliation invoice was issued. This may mean reopening earlier rows to apply the correction.
• Credit notes. A credit note reversing a previously-billed kWh figure is attributed to the same billing period as the original invoice it corrects, with the kWh shown as negative on a new row that references the credit note number. This preserves the audit trail back to the source documents.

The discipline that matters across all of these is keeping the apportionment workings visible. The evidence pack's value to the assessor — and to the auditor years later — is not the headline kWh figure but the ability to reconstruct how it was built.

Verifiable versus estimated readings, and the methodology note

ESOS expects verifiable data. In working practice that means data tied to a physical measurement: half-hourly settlement data from the metering system, an actual meter read (taken by the supplier, by the customer, or by a third-party reader), or audited fuel records for non-metered fuel types. Estimated data — the supplier's modelled consumption figure based on prior usage and seasonal profiles, used when no read has been taken in the billing period — is not verifiable in the ESOS sense and needs either a true-up against a later actual read or an explanatory note in the methodology.

The distinction shows up on the bill itself, against each meter read. Most UK B2B invoices print a one-letter code alongside the read: A for actual, E for estimated, C for customer-submitted, D for deemed. A handful of suppliers spell the type out in full ("Estimated read", "Actual read") and a few only surface the type in a read history table at the back of the invoice. Capture the flag against each read at extraction time. Where a billing period contains multiple reads with different flags, the conservative approach is to record both reads in the row detail and summarise the period as estimated unless every read in it is verifiable — it is cheaper to upgrade a period to verifiable later than to retrospectively downgrade it under audit pressure.

The true-up route resolves most estimated readings without any methodology note being needed at all. When a subsequent invoice reads the meter for real, the actual read implicitly corrects the cumulative kWh figure since the previous actual read, even if several intermediate billings were estimated. The cumulative consumption between two actual reads is verifiable; only the apportionment to individual billing periods inside that interval relies on the estimates. For most purposes the in-period kWh figure can be calculated from the bookending actual reads and the days-based apportionment described in the prior section, and the methodology note becomes redundant.

Two practical points follow:

• Request a meter read at the reference-window end where possible. A read on or close to the 31 December 2026 boundary (or whichever boundary the chosen window has) turns the entire reference period into a verifiable consumption figure without further reasoning. Suppliers will not always agree to a one-off planned read, but it is worth asking, particularly for higher-consumption supplies where the estimate risk is material.
• Attribute true-up corrections back to the periods they relate to. If an actual read in February 2027 corrects estimated reads stretching back to October 2026, the corrected kWh belongs to the October, November, December, and January billing periods proportionally — not to the February invoice that delivered the correction. This is the same apportionment discipline as the reference-window split, applied to a different boundary.

Where estimated readings remain in the reference period unresolved by a subsequent true-up — the meter could not be read, the read failed, the supply was inaccessible, the bill cycle ended at the reference window without a final read — the evidence pack needs a brief written methodology note explaining the situation. A defensible note is short, specific to the meter and period, and references the underlying invoice numbers. It identifies what was estimated (the MPAN or MPRN and the period), explains why a verifiable figure could not be obtained (no access to the meter, supplier estimate accepted in lieu of read), and states the basis used (the supplier's estimate, on the basis that it derives from prior actual reads and the seasonal profile). The aim is not to argue the estimate is verifiable — it is not — but to make clear what was done and why.

The audit risk to flag at extraction time, not at submission, is a portfolio where estimated readings make up a material share of the reference period. The Environment Agency and its devolved counterparts can request verifiable evidence for any consumption claimed; if a portfolio is more than marginally estimated, the assessor will raise it, and the time to get reads or to arrange site visits is long before the submission window closes. Surfacing the estimate share as a column total during extraction — a percentage of in-period kWh sitting on E-flagged reads — gives the compliance lead a number to act on rather than a problem to discover at sign-off.

Intensity ratios and the Action Plan progress narrative

Two of Phase 4's deliverables did not exist in Phase 3 and trip up most repeat compliers: energy intensity ratios across buildings, transport, and industrial uses, and a written progress narrative against the Action Plan commitments submitted at the end of Phase 3. Both are built on the per-meter kWh figures the prior sections produced.

The intensity ratio is a kWh-per-unit-of-activity figure expressed at a level the organisation's operations make meaningful. For buildings the standard ratio is kWh per square metre of gross internal floor area over the reference period. For transport the standard is kWh per vehicle-kilometre for fleet operations or kWh per tonne-kilometre for freight, depending on what the operation actually does. For industrial uses the ratio is kWh per unit of output — units produced, tonnes processed, hours of plant running, or whatever production measure the organisation already tracks. The numerator is always invoice-derived kWh; the denominator is always operational data the organisation already holds.

For the building ratio specifically, the assembly is straightforward once the meter-to-building mapping is in place. Take the building-level kWh sum from the aggregation step (decide upfront whether the ratio combines all fuels or splits electricity from gas, and apply that choice consistently across every building reported); pair it with the gross internal floor area from the property register or lease; divide. The result is a kWh per square metre figure that can be reported per building, per building type, or as a portfolio total. The choice of fuel combination should be documented in the methodology — an all-fuels ratio and a split-fuel pair of ratios both work, but mixing the convention across buildings is the kind of inconsistency that costs assessor time.

The transport ratio takes one extra step where the energy source is liquid fuel rather than electricity. Diesel and petrol consumption from fuel-card statements or mileage records arrives in litres and has to be converted to kWh using a published conversion factor. The standard UK source is the Department for Energy Security and Net Zero (DESNZ) greenhouse gas conversion factors, published annually and applied at the calorific value appropriate to the fuel grade. For electric vehicle charging the conversion does not arise — the kWh comes off the charging invoice or workplace charging records directly. Vehicle-kilometre denominators come from the fleet management system or, for owner-driver mileage claims, the expense system.

Industrial intensity ratios are the most variable in shape because the relevant denominator is whatever production measure makes operational sense — a manufacturer might use units produced, a hospitality group might use occupied room-nights, a food processor might use tonnes of finished product, a data centre might use compute-hours or IT-load kW. The denominator comes from the operational system that already tracks it; the invoice extraction supplies the numerator. The reporting choice that matters is consistency with prior reporting — if the organisation reported kWh per tonne in Phase 3, kWh per tonne is the ratio that earns comparability in Phase 4, even if a different denominator would feel cleaner in isolation.

The Action Plan progress narrative is the second new deliverable, and it is where the per-meter detail in the extraction earns its keep. Phase 4 requires reporting progress against the specific commitments made in the Phase 3 Action Plan — typically named sites, named measures (LED retrofits, BMS upgrades, boiler replacements, fleet electrification), and named kWh or percentage reduction targets against a baseline year. The substrate of the progress narrative is baseline-versus-current kWh on the same metering basis. A Phase 3 commitment to reduce HQ electricity by 15% from a 2019 baseline becomes, in Phase 4, a comparison between the 2019 reference-period kWh for the HQ MPANs and the 2026 reference-period kWh for the same MPANs.

If invoice extraction has been done with per-meter detail and source PDF references preserved, the comparison is a spreadsheet operation rather than an archaeological one — pull the baseline year's MPAN-level totals from the prior evidence pack (which still exists, because ESOS retention requires keeping it), pull the current year from the new pack, and the delta speaks for itself. Where measures were completed but the kWh did not move as expected, the narrative explains why; where commitments lapsed, the narrative says so. The invoice data does not write the narrative; it provides the numbers the narrative has to either justify or account for.

The data work supports the narrative. The narrative itself — what was done, what worked, what was abandoned, what replaced it — is the energy manager's account of the four years between phases, and is signed off by the accredited lead assessor who also signs off the assessment.

CCL line items, charity exemption, and the 12-year retention horizon

Two remaining invoice-level concerns sit alongside the main extraction work: how the Climate Change Levy line interacts with the kWh totals, and how the spreadsheet has to be shaped to remain auditable for the 12-year retention horizon ESOS imposes.

CCL appears as a separate line on most UK B2B electricity and gas invoices, charged on kWh rather than on cost. The main electricity rate is 0.775p per kWh during the 2025/26 financial year, rising to 0.801p per kWh from 1 April 2026; the gas rate is published separately by HMRC and applies on the same per-kWh basis. Because CCL is calculated against kWh, the CCL line gives a useful independent check on the consumption figure — the kWh implied by the CCL charge and rate should match the kWh on the supply line. A mismatch usually signals a partial-period CCL adjustment (most commonly because the rate changed mid-billing-period), a charity-exemption issue, or an invoicing error worth raising with the supplier.

The charity-exemption route matters for the subset of qualifying organisations whose activities meet the non-business definition for CCL purposes. Charitable-status organisations engaged in non-business activities — registered charities running non-commercial operations, academy schools, charitable trusts — can be CCL-exempt via the PP10 supplier declaration and PP11 customer certificate mechanism. Where the exemption is properly registered with the supplier, CCL should not appear on the invoice at all (or should appear at zero). Where the exemption applies in principle but CCL is still being charged in practice — a common situation where the PP11 was never lodged, or where it lapsed during a supplier switch — the invoices need correcting through the supplier's exemption team. That correction is a separate workflow from the ESOS evidence assembly, but the two overlap in practice: any organisation working through 12 months of supplier invoices for ESOS will see whether CCL has been charged correctly. Teams discovering a backlog can audit four years of academy trust energy invoices for wrongly charged CCL using the same per-invoice field extraction described earlier, with a different downstream purpose.

The retention horizon is the structural constraint most easily underestimated when the spreadsheet is first built. ESOS records must be retained for the compliance period plus two further compliance periods — roughly 12 years from the assessment date. For Phase 4 that means the evidence pack and its supporting invoices must remain accessible until somewhere around 2039. The compliance lead who built the spreadsheet will not be the one who answers questions about it in 2034 or 2038, which is what the structure has to be designed for.

Three design choices follow from that horizon:

• Source PDF references inline, in a stable archive location. Every row in the spreadsheet should carry a reference to the source invoice PDF — file name, supplier, invoice number — and the PDFs themselves should sit in a long-term archive rather than the original shared inbox. A shared mailbox from 2026 may not survive a tenancy change, a domain migration, or a platform retirement. A dated archive folder on the organisation's primary storage, or a document management system with a defined retention policy, is the right place. The spreadsheet reference should be readable on its own; the PDF should be findable from the reference.
• Methodology notes inline, not in a separate document. Apportionment workings, true-up explanations, and any methodology note for estimated readings should sit on the relevant rows of the same spreadsheet, not in a companion document that may be lost. The evidence pack should be a self-contained unit: open the spreadsheet, see the kWh figure, see how it was built, see the source PDFs.
• Archive-grade spreadsheet structure. Avoid macro-dependent formulas, exotic cell formats, and links to external systems that may have been retired before the retention period ends. The spreadsheet that opens cleanly in 2039 is the one that uses standard functions, native Excel data types, and self-contained references. Power Query and other live-connection mechanics are useful during assembly but should be flattened to values before the pack is archived.

The auditable unit at the end of the retention period is the spreadsheet plus the linked PDFs. Everything else is convenience for the team building it today.

---

From a stack of PDFs to an ESOS-ready spreadsheet

The decision the in-house compliance lead is actually making at this point is about time. A Phase 4-qualifying organisation typically holds 12 months of PDF invoices across two to five electricity suppliers, one to three gas suppliers, and a fuel-card statement file, covering tens to hundreds of meters. Transcribing those invoices by hand against the field shape described earlier — billing period, MPAN or MPRN, kWh, unit rate, standing charge, CCL line, VAT, estimate flag, invoice number, supplier — is the working-week-of-data-cleanup problem every consultancy explainer assumes away. Before the aggregation step, before the pro-rating, before the intensity ratios, the data assembly itself is a multi-week piece of work for a sizeable portfolio.

Invoice-data automation compresses that work to hours. The same 12 months of PDFs upload as a single batch; a natural-language prompt names the evidence-pack fields and the row shape (one row per MPAN or MPRN per billing period, with the estimate flag and the source-PDF reference retained); the output is a single Excel file where each row can be filtered by MPAN, summed by site, pro-rated against the reference window, and audit-traced back to the originating PDF. The data assembly stops being the bottleneck. The compliance lead's time goes into the work that needs judgement — choosing the reference window, deciding fuel-combination conventions for intensity ratios, drafting the methodology note, writing the Action Plan progress narrative — instead of into transcription.

Invoice Data Extraction is the tool we build for that specific job. Twelve months of UK supplier energy invoices arrive as a mixed batch — British Gas Business PDFs, multi-MPAN SSE consolidated invoices, EDF B2B bills, ScottishPower statements, Engie commercial gas — and the prompt that produces an ESOS-shaped output looks something like "Extract one row per MPAN or MPRN per billing period for ESOS evidence pack assembly. Capture supplier name, invoice number, billing period start, billing period end, MPAN or MPRN, kWh, unit rate, standing charge, CCL kWh, CCL rate, VAT rate, total amount, and the estimated-versus-actual reading flag. For multi-meter invoices, produce one row per meter. Format dates as YYYY-MM-DD. Include the source PDF file name and page number on every row." The Excel output lands with each row referenced back to its source PDF and page, which is exactly the traceability the 12-year retention requirement asks for. Batch processing handles up to 6,000 files per job and multi-page consolidated invoices up to 5,000 pages, so even large multi-site portfolios run as a single batch rather than as a sequence of smaller passes. The same prompt produces the same row shape across every supplier in the batch, which is what makes the aggregation, pro-rating, and intensity-ratio steps downstream a spreadsheet operation rather than another transcription exercise. Teams running the assembly this way extract a stack of UK supplier energy invoices into one ESOS-ready spreadsheet in a single afternoon and spend the rest of the project on the parts that need a human judgement call.

The scope to be honest about: invoice-data automation handles the data assembly. It does not run the ESOS assessment, does not certify the evidence pack, and does not replace the accredited lead assessor whose sign-off is required for Phase 4 compliance. It removes the slowest part of the in-house path; it does not remove the lead assessor from it.

That sharpens the decision the reader arrived with. The three paths into the 5 December 2027 deadline are: outsource the whole exercise to a lead-assessor consultancy and pay for their data team to assemble the evidence; run the entire exercise in-house with manual transcription and accept the multi-week assembly cost; or assemble the data in-house with invoice-data automation and engage the lead assessor only for the assessment and sign-off. The third path becomes practical at the point the data assembly stops being the constraint. Whichever path is chosen, the operational shape of what the evidence pack needs from supplier invoices — the fields, the aggregation, the pro-rating, the verifiable-data discipline, the intensity ratios, the retention horizon — is the same. The only variable is who does the data work.