Retire the workbook everyone is afraid to touch — carefully.
GA Applications modernises business-critical spreadsheets into governed dashboard views. We map each workbook's purpose, owners and dependencies, surface the hidden knowledge in formulas and manual checks, document every calculation, then run old and new in parallel until the numbers reconcile — with a defined cutover and fallback.
Every established business has a workbook that matters more than anyone admits — maintained by one person, understood by fewer. Modernisation is not deleting it; it is giving its logic a safer, more visible home.
Demonstration uses a fictional workbook. Your engagement starts from the real file and its real owner.
Workbook anatomy — before and after
Interface demonstration — fictional dataToday: three layers blurred in one file
- Sheet 'PasteHere'
- 214 formulas
- Hidden sheet 'calc2'
- Chart tab
- Owner risk
After: layers separated, logic documented
- Entry
- Transformation
- 'calc2' knowledge
- Presentation
- Governance
Built for the workbook the business quietly depends on
This direction serves spreadsheet owners, finance and operations leads, and the managers who consume the numbers. Its decisions are risk decisions: Can we trust this figure enough to act on it? What happens when the person who maintains the file is away — or leaves? Which of our twelve workbooks actually needs modernising, and which are fine as they are? The engagement answers those honestly, including the last one: not every spreadsheet is a problem, and the review will say so.
The target operating outcome: the workbook's logic lives in documented, governed measures fed directly from source systems — with the old file running in parallel until the new view proves itself, and a written fallback until you are confident.
Discovery first: map, document, separate
Entry vs transformation vs presentation
Fragile workbooks blur three jobs in one sheet: entry (facts arriving — jobs done, invoices sent), transformation (facts becoming measures — totals, rates, flags) and presentation (measures becoming views). Modernisation separates them. Entry becomes direct feeds or disciplined capture. Transformation becomes documented measures with owners, each traceable to the formulas it replaces. Presentation becomes role views that read those measures — so the chart shown to management and the number in the ops review are the same number by construction.
The hidden knowledge is the real work
Every long-lived workbook contains logic nobody wrote down: the adjustment the owner applies "because February is weird", the manual check against last week, the hidden sheet that quietly fixes a known data problem. Discovery surfaces these deliberately — interviewing the owner, tracing dependencies, reading the formulas — and every surfaced rule gets one of three fates: documented as a written adjustment rule, fixed at the source, or retired as obsolete. Nothing is silently reimplemented, because that is how modernisation projects ship confident, wrong numbers.
Exception queues replace manual checks
The checks the owner runs in their head — "this total looks low", "that customer appears twice" — become explicit exception rules with thresholds and owners. The judgement is preserved; the dependence on one person's memory is not.
Dependency map — what feeds what
Interface demonstration — fictional dataFeeds in
- Job system export (Fri)
- Accounting CSV (Mon)
- Manual sheet from depot
- Owner's adjustments
Depends on
- Management weekly pack
- Payroll estimate tab
- 2 other workbooks link to it
- Owner present on Fridays
The map answers the question nobody asks until it is too late: what breaks if this file changes? Two downstream workbooks and a payroll estimate depend on the fictional file above — all discovered, all planned for.
Old and new run together until the numbers agree
Cutover is earned, not scheduled. The new view runs alongside the workbook for an agreed period — typically several reporting cycles — and every shared measure is compared within a stated tolerance. Differences are classified: a documented improvement (the workbook had a known error), an open question (investigate before proceeding), or a defect in the new view (fix it). Only when the agreed measures reconcile for the agreed period does cutover proceed — and the fallback path is written down before anyone relies on the new view.
Parallel-run reconciliation — week 4 of 6
Interface demonstration — fictional data| Measure | Workbook | New view | Variance | Classification |
|---|---|---|---|---|
| Jobs completed (wk 29) | 34 | 34 | 0 | ✓Reconciled |
| Hours worked (wk 29) | 312 | 318 | +6 | iDocumented improvement — workbook excluded Saturday |
| Materials cost (wk 29) | $18,240 | $18,240 | 0 | ✓Reconciled |
| Utilisation % | 74.1 | — | — | ▲Open question — workbook formula under investigation |
| Backlog value | $96,500 | $97,100 | +$600 | ◐Defect in new view — fix before cutover |
Every surviving calculation, in writing
Measure: utilisation %
Replaces workbook formula =SUM(F2:F40)/SUM(D2:D40)*0.925. The 0.925 factor traced to a 2023 leave adjustment — now a written, dated assumption with an owner, reviewed annually.
Adjustment rule: depot manual sheet
Depot submissions arrive Mondays and are included in the following week's figures — previously an invisible lag. Now an explicit rule: submissions after Monday 12:00 roll to next period, labelled.
Exception rule: duplicate customer
Replaces the owner's manual scan for double-counted customers. Same customer name appearing in two source systems within one period raises a quarantine exception for human review.
Retired: February 'weirdness' fix
A manual February adjustment traced to a system that was replaced in 2024 — the cause no longer exists. Retired with the reason recorded, not copied into the new view by reflex.
What we need from you, and how delivery runs
Inputs
- The workbook itself (a copy) and time with the person who maintains it
- Read access to the systems it manually copies from, subject to interface review
- Knowledge of what depends on it — packs, payroll estimates, other workbooks
- An agreed tolerance and parallel-run period for reconciliation
Delivery & acceptance
- Modernisation review: purpose, owners, dependencies, honest verdict on whether to proceed
- Discovery: formulas, manual checks and hidden knowledge surfaced and triaged
- Build: feeds, documented measures, exception queues, role views
- Parallel run: reconciliation against the workbook for the agreed period
- Acceptance & cutover: gate criteria met (reconciled or documented, consecutively); fallback written and tested; workbook archived read-only; ownership handed to named people in your organisation
Right fit
- A recurring, decision-critical report depends on one person and one file
- Figures are re-keyed from systems every week
- You want the logic documented before the knowledge walks out the door
Wrong fit
- Ad-hoc analysis and one-off models (spreadsheets are the right tool for those)
- You want the workbook deleted on day one — parallel running is non-negotiable
- You want a guarantee the new view will match the old one (we reconcile; we do not assume)
Related directions and services
Frequently asked questions
Will you take our spreadsheet away?
Not until you are ready. The workbook keeps running while the replacement is built and reconciled in parallel. Cutover happens only when the new view matches the old within an agreed tolerance for an agreed period, and the fallback path is written down before anyone relies on the new one.
Our workbook has years of hidden logic. How do you handle that?
By treating discovery as a first-class step. Formulas, manual adjustments, lookup tables and the checks the owner runs in their head are all documented as written calculations with owners. Anything that cannot be explained is flagged as an open question rather than silently reimplemented.
What is the difference between data entry, transformation and presentation?
Entry is where facts arrive — jobs completed, invoices sent. Transformation is where facts become measures — totals, rates, flags. Presentation is where measures become views. Fragile workbooks blur all three in one sheet; modernisation separates them so each layer can be checked and changed safely.
Who owns the new view after cutover?
Named people in your organisation: a measure owner for definitions, a source owner for each feed and a view owner for the dashboard itself. GA Applications can support improvement after release, but the governance is designed so you are not dependent on any one person — including us.
Can some things stay in spreadsheets?
Yes, and some things should. Ad-hoc analysis and genuine one-off modelling belong in workbooks. Modernisation targets recurring, decision-critical reporting where fragility has a real cost — and the review will tell you honestly which of your workbooks qualify.
Which workbook would hurt most if its owner left tomorrow?
Send us a description of the workbook — what it decides, who maintains it, what feeds it. We will scope a modernisation that documents its logic, proves the replacement in parallel and keeps a fallback until you are confident.