Get the category right before the features.
Most teams first compare Déjà to an incident-management tool or a wiki, which is the wrong shelf. Your GRC platform isn't a rival either — it's where the receipts land. This page compares categories rather than products, because what a given vendor ships changes quarterly and what a category is built for doesn't.
- category map
- five shelves
- also
- three mistaken for us · one we feed
- compared by
- primary job, not feature list
Five shelves, and only one of them is this one.
- Mistaken for
Postmortem & incident management
primary job · document and coordinate
Built to run the incident and write it up afterwards: timelines, follow-ups, action items, a readable narrative for humans. Good at that, and many teams should keep it.
Outputprose documents, authored after the fact
- Mistaken for
Paging & alert routing
primary job · get the right person awake
Built to route signals to on-call engineers and coordinate response. Déjà sits behind paging rather than in front of it — the page still comes from your paging tool.
Outputalerts, escalations, response coordination
- Mistaken for
Wikis & documents
primary job · hold whatever you type
The status quo, and the real incumbent. Postmortems live in pages written from scrollback and memory. Nothing is structured, signed or queryable, so audit time means reassembly.
Outputwhatever someone remembered to write
- Works with
GRC & compliance platforms
primary job · manage the programme
Control mapping, audit workflow and evidence collection across a whole compliance programme. Not a rival — the destination. Déjà makes the incident-evidence portion cryptographic and hands it over through its GRC connectors. Each is in beta: built to the vendor's documented API, and not yet validated against a live instance by us.
Outputprogramme management and audit workflow
- Is
Evidence custody
primary job · keep the record as it happens
Built to seal what connected systems reported and what changed, when a signal is scored, in a form a third party can check without asking anyone. Evidence is the primary output; everything else is a by-product.
Outputsealed receipts, checkable offline
Three of these get mistaken for Déjà, one works with it, and none of them was built for the same job. Keep the ones you already have — most of them are Déjà's inputs.
The same question asked of each shelf.
Compared by what each category is designed to produce, not by what any particular vendor shipped last quarter. Individual products vary, several do more than their category suggests, and any of them may have changed since this page was written.
| Question | Incident management | Paging | Wikis | GRC platforms | Evidence custody |
|---|---|---|---|---|---|
| Primary artefactwhat the tool exists to produce | A written postmortem | An alert and an escalation path | A page someone typed | A managed control programme | A signed cross-system receipt |
| When it is createdrelative to the event | Afterwards, by people | During, automatically | Afterwards, by people | Afterwards, by people | During, when each signal from a scored source is scored |
| Who vouches for itif a regulator asks how you know | You do | You do | You do | You do | A published key, checked on a machine we don't control |
| What happens when nothing is foundthe majority case | Usually an inconclusive write-up | Not what it's for | Often nothing at all | A control marked untested | A receipt stating what was examined, or which eligibility condition failed |
| Structured or prosecan a machine read it | Prose, some structure | Structured alerts | Prose | Structured control data | Schema-validated receipts, open format |
| Independent verificationwithout the vendor in the loop | Generally not its model | Generally not its model | No | Self-asserted by the customer | Offline signature check for signed receipt types |
| If the vendor disappearswhat survives | An export, unattested | An export, unattested | An export, unattested | An export, still self-asserted | Signed receipts still verify against the published key |
| Relationship to Déjà | Complementary | An input — keep it | Replaced for evidence, kept for narrative | Where the receipts land | — |
The row that defines the category is the third one. Every other shelf produces a record you have to vouch for. That is not a criticism of any of them — it is what they were built for, and for their primary job it is entirely sufficient. It stops being sufficient the moment the reader is someone with no reason to take your word.
Probabilistic root cause is the closest thing, and the furthest.
A growing set of tools infer likely causes from telemetry and rank them. They are impressive, they are often right, and they are solving a genuinely harder problem than Déjà solves — which is exactly why they cannot produce the artefact this product exists to produce.
A ranked inference has three properties that disqualify it as evidence. It can change: the same inputs, a model revision later, produce a different order. It has no floor: something is always ranked first, so "nothing was found" is indistinguishable from "the top candidate is weak." And it cannot be independently checked: verifying it means re-running the vendor's model, which means trusting the vendor.
You cannot sign a guess.
Signing it doesn't make it true — it just makes it a signed guess. Déjà takes the opposite trade. It attributes one narrow failure mode — a producer removed, renamed or retyped a field and a consumer broke on it — using published weights and a published threshold. Below the threshold it says so. Where it cannot score at all it says that instead. The scope is small enough that determinism is achievable, and determinism is the only thing that makes a signature meaningful.
If your problem is understanding the other failure modes, use an inference tool — it will help and Déjà won't. If your problem is producing a record someone else can check, ranking is the wrong shape of answer no matter how good the ranking is.
The two side by side: a guess is not evidence.
Three cases where we'd tell you to use something else.
Déjà does one job. If that isn't your job, one of the categories above is the better purchase and we would rather say so now than discover it in month four of a pilot.
- Case one
No compliance pressure yet
If nobody outside your company has ever asked you to evidence an incident — no examiner, no auditor, no client security review — you don't need this yet. Run postmortems in a wiki and revisit when your audit calendar is real, which in regulated markets arrives earlier than most teams expect.
Better fita wiki and your existing paging tool
- Case two
Your problem is alert routing
If the thing that hurts is getting the right engineer awake and coordinating the response, that is a paging problem and Déjà is not a paging tool. It sits behind paging, reading what your alerting already produces.
Better fityour paging tool — plus Déjà later
- Case three
You want help writing postmortems
If the bottleneck is prose — synthesising a timeline, drafting the narrative, tracking action items — incident-management platforms do that natively and well. Déjà produces structured receipts, not documents. Plenty of teams run both.
Better fitan incident-management platform — plus Déjà later
A fourth case worth stating: if your stack isn't on the status page, say so before a trial rather than after. One qualifying signal source is demonstrated today, and we would rather lose the deal than run a pilot that was never going to produce anything.
Underneath, not instead.
Your GRC platform keeps its job
Control mapping, workflow and programme management stay where they are. What changes is the quality of one input: the incident-evidence portion arrives as signed receipts rather than as documents someone assembled. Comparing Déjà to a GRC platform is comparing the pipe to the tank.Your observability and paging are inputs
Déjà reads what they emit. Without them there is nothing to seal, which is why it makes no sense as a replacement and every sense as a layer. Keep the contract, keep the tooling, keep the on-call rota.Your postmortems become narrative rather than evidence
This is the one real substitution. The write-up stops being the record of what happened and becomes the explanation of it — which is what a postmortem is good at, and what it was never reliable for.Your audit firm gets something to check
Rather than reading a narrative you wrote and asking you to stand behind it, they verify an artefact on their own machine. That is a different conversation, and it is the one this whole category exists to make possible.
Every category on this page produces a record. Four of them produce one you have to vouch for. If nobody has ever made you vouch for yours, buy one of those instead — and if someone has, that is the conversation worth having.