The system only knows now

Three vendors, three architectures, one shared absence. Oracle Fusion, SAP S/4HANA, and Workday all overwrite state in place, so what was true on a past date is recomputed on demand, cached briefly, or gone. This page is the whole pattern.

RuleTreat every transactional system as a machine for now. History is a storage decision, and the store should be a warehouse you own.

Neverassume the vendor kept your past. On all three platforms, the default answer to what was true then is a reconstruction, a window, or nothing.

◆ The shared mechanismwhy three unrelated systems fail the same question.

A transactional system exists to process the next transaction fast, so it keeps one current version of every record and updates it in place. That design choice, made independently by every vendor, has one consequence: the past is not stored. When a report asks about a past date, the system either recomputes backwards from the present, or consults a short bounded cache of snapshots, or has nothing to consult at all.

Each failure this catalog documents is one of those three shapes wearing a platform costume. The reconstruction shape times out and shifts under concurrent activity. The cache shape has a grain and a window, and purges past it. The nothing shape sends analysts to folders of archived spreadsheets, which are a snapshot store built by hand.

Three delivered shapes of the same absence, and the one store that ends it.

◆ The evidence, platform by platformevery as-of failure this catalog documents, in one table.

Platform	What the question hits	The page
Oracle Fusion	Receivables aging recomputes from items since paid; on-hand stock is stored for today only; the reporting layer has no warehouse behind it.	Aging as of any date · Inventory as of any date · Reporting without memory
SAP S/4HANA	Key-date open items recompute with documented defects; historical stock reverses movements at query time; order status overwrites in place; embedded analytics loses archived data.	Open items as of a past date · Stock as of a past date · Order backlog · No history by design
Workday	Workforce snapshots are monthly, kept about 36 periods, purged beyond; history before go-live was never loaded as transactions at all.	Headcount as of any date · The trending store · History before go-live

No vendor can write this table, because each one sees only its own system. A practitioner who has extracted all three can, and the sameness is the finding: this is not an Oracle gap or an SAP gap or a Workday gap. It is what transactional architecture does to history, everywhere, and it will be true of the next platform too.

◆ The one fixsnapshot facts, owned, boring, and final.

The fix is identical on every platform because the problem is: dated snapshot tables in a warehouse you own. One row per thing per day, written by the extraction pipeline, retained for as long as you choose, rolled up through versioned dimensions so old rows stay comparable across reorganizations. Storage at this grain costs coffee money; the queries are filters.

The vendors will not build this for you on their side of the wall, and the reason is structural rather than malicious: their systems are optimized for the next transaction. Their analytics products that do keep history are separate purchases with their own windows, caps, and meters, documented on this catalog's cost pages. Owning the store is the only version of remembering that nobody can reprice, deprecate, or purge.

Use case

Problem

Every platform in the landscape fails the same question, what was true on a past date, and each failure is being solved separately with workarounds, add-ons, and spreadsheet archives.

What we build

One owned warehouse with snapshot facts as the standard pattern, fed by each platform's extraction path, applied identically to finance, inventory, orders, and people.

What you get

A single answer to the as-of question everywhere: stored, reproducible, joinable across platforms, and immune to vendor retention windows.

How many systems in your landscape can answer what was true then?

We build the store that remembers for all of them, in your cloud.

Talk to us

Terms on this page

ERP: Enterprise resource planning, the systems of record this page spans.
SAP: The vendor of S/4HANA. Systems, Applications, and Products in data processing.
state: The current version of a record, updated in place as transactions post.
reconstruction: Deriving a past value by reversing later activity out of the present one.
retention window: How long a delivered cache keeps snapshots before purging them.
snapshot fact: A table storing dated copies of a state, one row per thing per date.
grain: What one row means: one item, one worker, one order, on one date.
versioned dimension: A dimension keeping historical values with dates, so old rows roll up as they did then.
as-of date: The past date a question is asked about, as the data stood then.