Most close acceleration programmes are not wrong about what to do. They are wrong about when, and the cost of that is not a delay — it is a stage that delivers a fraction of its value because the thing it depended on was not in place.
This playbook sets out the sequence as a prerequisite chain: each stage moves a named capability to a specific maturity level, and the following stage assumes that level has been reached. Skipping is possible and common. It is also the single best predictor of a programme that lands at eight days and stops.
The target, and the right peer group
Before sequencing, set the target honestly. Against our benchmark peer set of 58 comparable organisations:
| Measure | Peer median | Top quartile | Interquartile range |
|---|---|---|---|
| Days to close | 6 | 4 | 5–8 |
| Close automation — routine steps automated | 55% | 75% | 40–70% |
Two observations that change how the target is set.
The median is the wrong target. A programme aimed at the median is aimed at being unremarkable, and by the time it lands the median will have moved. Aim at the top quartile or do not run the programme.
The interquartile range is wide, and that is the useful signal. A 5–8 day spread means context dominates — entity count, ERP quality, audit intensity, whether statutory and management close are separated. A single headline number imported from a conference slide is not a target; it is a way to lose credibility in month four.
The sequence
| # | Stage | Capability moved | Target | Prerequisite | Software? |
|---|---|---|---|---|---|
| 1 | Standardize the close | Close management | L1 → L2 | None — this is where you start | No |
| 2 | Wire the ERP data | ERP integration | L2 → L3 | Stage 1: a defined process to feed | Partly |
| 3 | Automate reconciliation | Account reconciliation | L2 → L3 | Stage 2: reliable, timely source data | Yes |
| 4 | Add transaction matching | Transaction matching | → L3 | Stage 3: certified balances to match into | Yes |
| 5 | Lock audit-ready controls | SOX controls & audit | L3 → L4 | Stages 1–4: a process that produces evidence | Yes |
The chain is the substance. Each prerequisite is not a nicety — it is the condition under which the next stage's value is available at all.
Stage 1 — Standardize the close
Moves: close management, L1 → L2. Software: no.
The close calendar becomes one calendar. Tasks have owners, dependencies, durations and a defined completion criterion. Exceptions are enumerated rather than absorbed.
This is organisational work and it is the stage most often skipped, because it produces nothing demonstrable and cannot be procured. It is also the stage that determines whether everything after it works, because stages 2–5 all encode whatever process they find.
Done when: two preparers produce the same output from the same inputs; the close completes correctly when your most experienced person is on leave.
Failure mode: documenting the current state and calling it standardization. A document records one version of a variable process; it does not remove the variation. Early warning: the documentation exercise finishes suspiciously quickly and nobody argued.
Typically worth: 1–2 days on its own, at no licence cost. It is the highest-return stage in the chain and the least visible.
Stage 2 — Wire the ERP data
Moves: ERP integration, L2 → L3. Software: partly.
Close-relevant data flows from source systems on a defined schedule, in a defined format, with a defined cut-off. The objective is not integration for its own sake — it is removing the manual extract-and-transform work that sits between the ledger and every schedule.
Done when: no close task begins with a person exporting something. Sub-ledger balances arrive rather than being fetched.
Failure mode: building integrations to a process that has not been standardized, which produces integrations to three variants and a maintenance burden that outlives the programme. This is precisely why stage 1 precedes it.
Early warning: integration scope keeps expanding because each team needs "their" file.
Stage 3 — Automate reconciliation
Moves: account reconciliation, L2 → L3. Software: yes.
Balances are matched against independent sources automatically, low-risk accounts are auto-certified against materiality thresholds, and preparer effort concentrates on exceptions.
This is the first stage where a dedicated platform is usually justified, and where the buy/extend decision genuinely arises. The selecting conditions are the standard ones: volume, entity complexity, or audit pressure exceeding what the ERP and spreadsheets carry.
Done when: the reconciliation population is risk-tiered, low-risk accounts certify without human touch, and the exception queue is worked rather than the whole population.
Failure mode: automating the reconciliation format rather than the reconciliation decision — a tool that produces the same schedules faster, with the same person still reviewing every line. Early warning: auto-certification rate stays near zero after go-live because no materiality thresholds were agreed.
Typically worth: the largest single reduction in close effort in the chain.
Stage 4 — Add transaction matching
Moves: transaction matching → L3. Software: yes.
Item-level matching underneath the balance-level reconciliation: bank lines, intercompany pairs, sub-ledger detail. This is where AI-assisted matching genuinely earns its place, because the work is high-volume, low-judgement, and tolerant of a review step.
It follows stage 3 rather than accompanying it because matching without certified balances solves the wrong layer: you can match items perfectly into a balance nobody has certified.
Done when: the unmatched queue is small enough to be worked daily rather than at period end — which is what moves the work out of the close window.
Failure mode: tuning match rules against last month rather than against the exception population, producing a high match rate and a residual queue of exactly the hard cases.
Stage 5 — Lock audit-ready controls
Moves: SOX controls & audit, L3 → L4. Software: yes.
Every close task carries evidence that a control operated: who prepared, who reviewed, against what, when, and what changed after review. The objective is that audit evidence is a by-product of doing the work rather than a project after it.
This stage is last because it has the most prerequisites — it can only certify a process that exists (stage 1), is fed reliably (stage 2), and produces certified balances (stages 3–4).
Done when: the audit request list is satisfied by export rather than by assembly.
Failure mode: treating this as a documentation stage. If evidence is produced by people rather than by the system, you have added work to the close in the name of controlling it.
What "done" looks like overall
A four-day close is not a faster twelve-day close. It is a different shape:
- Work has moved out of the window. Reconciliation, matching and much review happen
continuously; the close window is for judgement and consolidation.
- Exceptions are worked, not discovered. The queue exists year-round.
- Evidence is a by-product, not a project.
- The process survives absence. This remains the single best diagnostic.
Sequencing under constraint
If a budget freeze or headcount cap is active, stages 1 and much of 2 remain fully available — they are organisational work. Deferring stages 3–5 while completing 1–2 is a legitimate and frequently optimal position: it captures real days, costs no licence, and materially improves the eventual platform selection, because you will be buying for a process you understand.
- The order is a prerequisite chain, not a preference. Each stage assumes the previous
stage's maturity level; skipping produces a stage that delivers a fraction of its value.
- Stage 1 is the highest-return and least visible. It needs no software, produces
nothing demonstrable, and determines whether stages 2–5 encode a process or a variation.
- Aim at the top quartile, not the median — but set the target against your own entity
and audit complexity, because the peer spread of 5–8 days means context dominates.
Where this leaves you. Score your close against the five stages and identify the earliest one that is not genuinely complete. That stage — not the most appealing one — is where the next increment of value is.