We would rather be measured than believed.
Regnum Personia is being run inside a working clinic: its real records, its real staff, real consequences when something runs out, and enough time for the results to mean something. Six builds, in order. Everything measured before we change anything, so that six months from now the improvement is a number instead of an opinion.
A demo can be made to say anything
Four things a live site gives you that sample data never will.
Real operational data
Records kept by people under pressure: inconsistent units, late entries, the odd blank column. The conditions the software has to survive.
Real users
Staff who did not choose the software and have a queue waiting. If a screen is awkward, it gets abandoned rather than reported.
Real consequences
A missed reorder is a patient sent away. That is the only stake that properly disciplines a forecast.
Time
Seasonality, supplier drift, an outbreak week. None of it shows up in a fortnight, all of it shows up in a year.
Six builds, in this order
The order matters more than the list. Prediction before measurement is decoration.
Ingestion
Take the clinic's information as it already exists, without asking anyone to change how they work.
Build 02Truth before prediction
Count the shelf. Compare it to the record. Write down the gap. That is the before picture.
Build 03Forecasting
Predict, wait, then score the prediction against what actually happened. Repeatedly.
Build 04Expiry intelligence
Not the expiry date, but the quantity that will still be sitting there when it arrives, and what that costs.
Build 05Financial intelligence
Connect physical stock to operational risk to money, in the language the owner actually thinks in.
Build 06Anomaly detection
Learn what normal looks like, then point at the days that are not, and say nothing about why.
Sit on top of what is already running
The first milestone is deliberately unglamorous. Before any intelligence, Regnum has to accept the clinic's existing information, in the shape it already exists, without forcing anyone to change how they operate.
That is the strategic point being proven, and it is the one that matters later: Regnum is a layer over an operational environment, not a replacement for it. A facility that cannot afford a migration can still be measured. A national system that already holds the records can still be read.
- Today: CSV and Excel, mapped once and remembered
- Next: scheduled pulls straight from the source database
- Then: an API contract, so the ingestion is repeatable rather than heroic
| Record | Why it is needed |
|---|---|
| Inventory records | Current position |
| Purchase records | What was bought, when |
| Supplier information | Who, and how reliably |
| Batch numbers | Risk lives at batch level |
| Expiry dates | The wastage clock |
| Consumption history | The basis of every forecast |
| Stock received | Confirms what arrived |
| Stock adjustments | Where the truth was corrected |
| Prices and costs | Turns units into money |
| Supplier lead times | Turns a forecast into a deadline |
Lead times are frequently missing at the start. Regnum reconstructs them from order and receipt dates until the real figures exist.
Measure first. Impress later.
For the first weeks the software is not trying to be clever. It is establishing whether the records can be trusted at all, by physically counting a sample of meaningful medicines and comparing the shelf to the system.
Repeated across the medicines that matter, this produces one number nobody currently has: baseline inventory accuracy.
What gets written down before anything is improved
- Stock-outs, and near-stock-outs
- Expired stock, and stock near expiry
- Overstock, and total inventory value
- Emergency orders placed, and what they cost
- Actual supplier lead times, per supplier
- Inventory discrepancies, line by line
A prediction that is never checked is a marketing claim
For each meaningful medicine, Regnum states a position, a rate, a deadline and a risk. Then it records what actually happened, and scores itself.
| Current stock | 420 units |
| Average consumption | 18 / day |
| Estimated days remaining | 23 |
| Supplier lead time | 14 days |
| Recommended action window | 9 days |
| Stock-out risk | Moderate |
| Predicted | Actual | Error |
|---|---|---|
| 80 units in 14 days | 84 | +4 |
| 210 units in 14 days | 198 | −12 |
| 45 units in 14 days | 61 | +16 |
Illustrative structure, not pilot results. Every forecast Regnum issues is stored with a due date, and settled against the count on that date, the misses included.
"Our AI predicts shortages."
Unfalsifiable, unremarkable, and indistinguishable from every other pitch in the room.
"Across the clinic pilot, Regnum achieved X% forecasting accuracy across Y medicines over Z days."
A different category of statement. It can be interrogated, reproduced, and held against us, which is precisely what makes it worth hearing.
Stop tracking expiry. Start predicting wastage.
"Amoxicillin expires December 2026" is a date. Any inventory package can print it, and it tells a manager nothing they can act on.
The useful question is how much of that batch will still be on the shelf when the date arrives, which requires the expiry date and the consumption rate in the same sentence. Once those meet, expiry stops being a record-keeping field and becomes a forecast with a currency value attached.
| Units remaining | 320 |
| Expires in | 83 days |
| Expected consumption before expiry | 190 |
| Estimated excess at expiry | 130 units |
| Financial exposure | Priced from your own cost records |
Excess is what the projection says will still be on the shelf on the expiry date, the figure a manager can still act on today.
And then, across facilities
The same arithmetic run at more than one site turns a write-off into a transfer:
In a country where stock sits unevenly across facilities, this is the difference between medicine expiring in one district and being unavailable in the next.
Physical inventory → operational risk → money
This is the deliberate separation. A logistics system tells you where the stock is. Regnum is built so the person who owns the practice can ask what it is doing to the balance sheet, and get the answer from the same records, on the same screen, the same day.
"How much money is sitting in medication inventory right now?"
"How much of it is at risk of expiring?"
"Which medicines are tying up the most capital?"
"What did emergency procurement cost us this quarter?"
"Where are we overstocked?"
"Which supplier price changes affected our purchasing?"
Point at the day. Say nothing about the reason.
Start simple and stay simple: establish what normal movement looks like for each item over its own history, then flag the deviations. No exotic modelling is required to be useful here, and over-engineering it early is how these systems lose their audience.
Each flag carries the window it compared against and the records behind it, so the review takes a minute rather than an afternoon.
The pilot scorecard
One page, kept from the first week. Baseline on the left, current on the right, and the difference between them is the entire argument.
| Measure | Baseline | Current | Change |
|---|---|---|---|
| Inventory accuracy (counted vs. recorded) | Capturing | Not yet | Not yet |
| Stock-outs per month | Capturing | Not yet | Not yet |
| Near-stock-out events | Capturing | Not yet | Not yet |
| Units expired | Capturing | Not yet | Not yet |
| Value of expired stock | Capturing | Not yet | Not yet |
| Stock at risk within 90 days | Capturing | Not yet | Not yet |
| Total inventory value held | Capturing | Not yet | Not yet |
| Overstocked lines | Capturing | Not yet | Not yet |
| Emergency orders placed | Capturing | Not yet | Not yet |
| Cost of emergency procurement | Capturing | Not yet | Not yet |
| Average supplier lead time | Capturing | Not yet | Not yet |
| Forecast accuracy (medicines × days) | n/a | Not yet | Not yet |
| Anomaly flags raised / confirmed useful | n/a | Not yet | Not yet |
The next site benefits from the first one
If you run a clinic, a pharmacy or a group and would rather see a method than a slide deck, we will walk you through the baseline process on your own records, and tell you plainly what Regnum cannot do yet.