Questions · Aspen Vault
The questions we are asked before anyone asks about price.
Plain answers, in the language of the core shed rather than the language of software. If your question is not here, ask it and we will add it.
- 01What is a knowledge graph, and why does it matter in mining?
- 02Do we have to replace the systems we already run?
- 03Our oldest records are typewritten and hand-drafted. Are they usable?
- 04What happens when the archive contradicts itself, or the quality control failed?
- 05What will Aspen Vault not do?
- 06Who in the organisation is able to use the result?
What is a knowledge graph, and why does it matter in mining?
Every operation has one person who knows where everything is. A knowledge graph is the first system able to keep what she knows after she has gone.
The person who knew where everything was
Thirty years on the property. A memory shaped like a filing system. Ask her which holes went into the eastern ridge before the haul road was cut, and she tells you. Ask which of them went to the Vancouver laboratory and which went to Perth, and she tells you. Ask whether the Perth numbers from the 1998 programme can be trusted, and she says not entirely. The standards were drifting by the second batch, everyone knew, and the decision at the time was to live with it and move on.
Now ask the database the same three questions.
Every one of those facts is in there. The hole names are in there. The laboratory codes are in there. The certificates are scanned and filed, the standards results are in an appendix, the survey file is on a network drive. Nothing is missing.
And the database still cannot answer, because the answer was never in the facts. The answer lives in the connections between the facts. And the connections live in one person, who retires in March.
A knowledge graph is the attempt to write down what she knows, in a form a machine can travel.
One sample, and everything true about it
Take a single interval: sample 41-B, from a box in the core shed.
Sample 41-B
Sample 41-B is cut from the interval 212.4 to 213.9 metres. The interval sits in hole RC-98-14. The hole was collared at a surveyed point on the ridge, and bends away from the collar azimuth by eleven metres at depth, according to a downhole survey the contractor ran in November. The hole was drilled during the 1998 programme. The sample was assayed by a laboratory in Perth, using a four-acid digestion, with a detection limit of five parts per million for arsenic. The result was published on certificate 98-4471, page three. The sample was bracketed by a certified reference material, and the reference material failed by nine percent low. The interval lies twelve metres below the water table, and falls four hundred metres inside the permit boundary. Two resource statements, in 2004 and 2011, drew on this sample.
Every phrase in italics is a relationship. In a knowledge graph each one is written down and stored as a thing in its own right, as real and as durable as the grade itself.
The sample is a point. The relationships are the roads leading away from it. Put every sample, certificate, hole, survey station, laboratory, method, batch, campaign and boundary into the same country, with all the roads drawn, and you have a graph.
How a question travels
Here is a question a mine manager asks in an ordinary week.
Show me every interval above the arsenic penalty threshold, sitting below the water table, within fifty metres of surface. And tell me which of them a resource statement depends on.
A person answering by hand needs the assay certificates, the survey files, the topographic surface, the monitoring record and two historical reports open at once, and needs to hold the reconciliation in their head while reading. It takes a fortnight, if the right person is available, and the result is a number with no way back to its evidence.
In a graph, the question simply walks the roads. Arsenic values above a threshold → the intervals carrying them → the holes containing those intervals → the surveyed traces putting each interval in real space → the elevation of the surface above it → the water table below it → and out the other side, the certificates, the batches, the failed standard, and the two resource statements resting on all of it.
The answer arrives as a picture of the orebody. Every coloured interval in the picture can be followed back to a page of a certificate.
Why the systems you already own cannot do this
None of this is a criticism of your database, your laboratory information system or your modelling package. They were built to do something else, and they do it well.
A table has to know the question before the data arrives.
Columns are decided at design time. Somebody in 2003 decided which fields the assay table would hold, and no question falling outside those columns has been askable since. A graph takes a new kind of relationship the way a country takes a new road. The existing roads are undisturbed.
A join needs a key, and forty years of records do not share one.
Holes were renumbered when the property changed hands. The coordinate system changed in 1996. A 1974 laboratory code and a modern LIMS field describe the same digestion under different names. In a graph you state the equivalence once, as a relationship, and every question asked afterwards inherits it.
A report is a snapshot. A graph accumulates.
Each new report restates a position and quietly replaces the last one, until the current team is working from a summary of a summary. In a graph each new document joins the web rather than overwriting it, so a question spanning four decades and three operators returns all three operators' evidence, side by side, with the disagreements intact.
A number in a cell has lost its parentage.
Copy a grade into a spreadsheet and it becomes an orphan: true-looking, untraceable. In a graph the number keeps the roads leading back to the laboratory, the method, the batch and the page. Presentation without provenance is a picture. Provenance is what survives review.
The short version
A traditional system stores facts and leaves the connections to people. A knowledge graph stores the connections themselves.
The one difference explains why a question crossing four types of record (assay, survey, terrain, water) is a fortnight of work in one and seconds in the other, and why the answer in the second case arrives with its evidence still attached.
No new drilling. No new sampling. No new laboratory work. Every metre you have ever drilled is already holding the information. What has been missing is the map of how it all connects.
And the person retiring in March
She is the reason any of this matters.
Her successor will be capable, and will start with a fraction of what she carried, not through any failing, but because nobody can read a forty-year archive and hold it accurately in mind. The record outlives everyone who understood it, and each handover loses a little more of the reasoning behind decisions still standing.
A knowledge graph is not a replacement for her. It is the first system able to keep what she knows after she has gone.
↑ All questionsDo we have to replace the systems we already run?
No. An engagement needs one dataset and permission to read it.
There is no platform commitment, no migration, and no change to how your teams work. Your database stays where it is, your modelling package stays where it is, and your people carry on using both. Aspen Vault reads the record and builds the graph alongside it.
One archive, one defined scope, one fixed timeline. Every stage is scoped and priced before it begins, and you keep the output whatever you decide afterwards.
A knowledge graph is an addition to your evidence base, not a replacement for your operating systems. If an engagement ends after the first stage, you still hold everything produced in it.
↑ All questionsOur oldest records are typewritten and hand-drafted. Are they usable?
Those are usually the records with the most unread information in them.
Format spread is the ordinary condition of a long-lived property, and the ingestion engine is built for it: typewritten certificates scanned to PDF, hand-drafted sections on mylar and paper, legacy databases and loose spreadsheets, modern laboratory exports, implicit surfaces from Leapfrog, mine designs from Deswik. Any format, any decade, any vendor.
Reconciling them is the work. Detection limits move across decades, a 1974 laboratory code and a modern field describe the same digestion under different names, and a single site may pass through four owners over six decades with a different vocabulary each time. The semantic layer records each equivalence once, so the whole archive becomes comparable rather than merely co-located.
Nothing is discarded on the way in.
↑ All questionsWhat happens when the archive contradicts itself, or the quality control failed?
You see the contradiction, and you see the failure.
We show the data as it is, including the gaps, the failed quality controls and the questions the archive cannot answer. A hole with no downhole survey, a certificate with no method recorded, a campaign whose standards failed and whose results were used anyway: naming the holes in the record is part of the product, not an embarrassment to be smoothed over.
Batch failures are isolated during refinement rather than averaged away, and quality control is shown throughout the delivered view. Where two operators disagree, both positions stay in the graph with their sources attached, so a reviewer can see the disagreement instead of inheriting someone's resolution of it.
A finding higher than the number you were hoping for is reported exactly as readily as a lower one. A record showing only its strengths is a brochure.
We build the record.
We do not shape it.
What will Aspen Vault not do?
Four things, and we say so plainly before an engagement rather than after it.
Out of scope: resource estimation, competent person sign-off, geochemical or tailings modelling, and consenting advice. Those remain the work of qualified people carrying professional liability for their judgements.
Aspen Vault tells you what is in the rock and where the evidence for it sits. A competent person signing under PERC, JORC or NI 43-101 can move from a coloured interval on a section to the document underwriting it in a few steps, which is the difference between an assurance and a document, and it makes their work defensible rather than doing it for them.
↑ All questionsWho in the organisation is able to use the result?
Anyone with a question. Knowledge belongs to everyone in the decision, not only to the few specialists able to read the files.
Questions are asked in plain language and answered as sections, plans and three-dimensional views, with the certificates and survey traces behind each answer attached. A record nobody outside the geology team can interrogate is not a record. It is a bottleneck.
The same holds outside the company. The record going to the community, the board and the regulator is one record, with no second version prepared for any of them.
Our full policy set is published in full: professional conduct, business ethics, competition, information security, data protection, modern slavery, trade sanctions, and health, safety and environment.
↑ All questionsAsk us the one we have missed.
We would rather answer a hard question in public than a soft one in private. Send it, and if it is a question other people are also carrying, it goes on this page.