Time matters. Your systems don’t agree.

Sequence is where the value is. Assembling a true timeline is a data engineering problem long before it is a modeling problem.

Two kinds of sequence

Every organization has both. They almost never live in the same system.

Measured continuously
Time series

Steady samples. Heart rate, vibration, price, load.

Recorded when something happens
Point process

Irregular events. A visit, a repair, a trade, a plan change.

Together
measurementseventsevent 1event 2nothing moveseverything moves

Apart, each one is just a record. Together, they explain “why.”

The why

Event 2 changed things. Event 1 didn’t.

Something to act on

Do more of what worked. Stop what didn’t.

an agentor your team

A timestamp is not a time

An entry can land in the record long after the event it describes.

What happenedissue startsfirst contactcause foundfix appliedWhat the record saysSame events. Different order. Variable, unrecorded lag.

The cause was found before the fix was applied. In the record, it appears after. Nothing in the schema marks the difference between when something happened and when someone entered it.

Three things to settle first

None of them are modeling questions.

Which time is this?

created_atwhen it happenedwhen it was enteredwhen the job ran?

Nothing in a schema says whether a date is when the thing happened or when someone typed it.

Is this the same event?

System ASystem B?same?mostly. not always.

Two systems overlap on most records and disagree at the edges. Whether that matters depends on the question.

What does one row mean?

1 row = a day1 row = an event1 row = a reading

Join records at different grains and you get numbers that look right and cannot be reproduced.

So we build MTN FieldMap first

What each field means, which concept it carries, and where the evidence runs out.

?Records systempatient_idsvc_dtSchedulingclient_noappt_tsDevice feedsubjectreading_atPersonEventWhen it happenedYOUR SYSTEMSWHAT THEY MEAN
mapped, with the evidence behind itflagged: asked, not guessed

MTN Guide builds this FieldMap from your schemas, documentation, and API definitions, never from your records, and keeps it current as those systems change. A timeline is an artifact of this work, not something waiting in the data to be found.

What a true timeline makes possible

  • Models that learn from a sequence of events rather than scoring one moment in isolation.
  • Agents that can act on what happened, in the order it actually happened.
  • Honest answers to what usually comes next, and where acting changes the outcome.

Build it with your own models, your own team, or a vendor you already trust.

Bring us one schema

Send a representative schema from a system you care about, and we will show you what the FieldMap finds, and what it cannot tell you without asking.