The Forecast That Shows Its Working
A risk score you cannot question is a risk score you cannot act on. Seven predictive packs that work from your own records with rules and statistics you can read, a Why this? panel on every row, actions that go through the normal approvals, and a check fourteen days later on whether the warning was right.
The buyer looked at the alert, high risk of stockout on cooking oil at the coast branch, and did nothing. Not because she disagreed, but because she could not tell whether she did. The screen gave a score and a colour. It did not say whether it had noticed the delivery due on Thursday, or the promotion that had doubled last week’s sales, or nothing at all. Two days later the shelf was empty, and the next alert she was shown, she trusted even less.
This is the problem with most predictive features in operations software, and it is not accuracy. A good-enough warning that people act on beats a perfect one they ignore, and people do not act on a number they cannot check. So the first decision behind the predictive packs was what not to build: no model trained on your data, no score that cannot be traced back to rows you could open yourself. Each pack is a stated calculation over your own records. That is less impressive in a demonstration and much more useful on a Tuesday.
Seven questions, each with its working
| Pack | What it reads | What it is asking |
|---|---|---|
| Stockout risk | On-hand, 30 and 90 days of sales, lead time, safety stock | Will this run out before the next delivery could arrive? |
| Supplier delay | The supplier’s late share over a year, the expected date, acknowledgement | Is this open order likely to be late? |
| Late payment | Days past due, the customer’s history, the balance | Will this invoice be paid late, or not at all? |
| Cash pressure | Cash, receivables and payables due in 30 days, your collection rate | Will next month’s bills outrun next month’s collections? |
| Till shrinkage | Shortages against cash sales over 60 days, sessions closed short | Is one till losing money in small amounts? |
| Approval SLA breach | A pending approval’s age against the usual time for its type | Is this approval about to hold something up? |
| Asset custody risk | Value, risk level, last verified, overdue returns, movement | Which assets are most at risk of going missing? |
Each pack shows its top fifty, banded high from 75 and medium from 50. The details are deliberately ordinary. Stockout risk is days of stock left against lead time and safety stock, with recent sales blended with longer-run sales and a seasonal uplift for peak months. Till shrinkage counts shortages and ignores overs, because an over on Tuesday does not repay a short on Monday; it usually means two counting errors rather than none. We wrote about how the underlying demand forecast works, and where it is blind, in demand forecasting, how it works and the peak our forecast cannot see.
Why this?
Every row has a panel that answers the buyer’s question from the opening paragraph. It lists the drivers and their values, the records and the time window it read, how many records the figure rests on, the confidence and why it is that confidence, the method in a sentence, and the recommended action. If the delivery due on Thursday is not among the drivers, you know the score did not see it, and you know what to do about that: overrule it, and say so.
No AI allowance spent, no data sent
The packs refresh every morning by running these calculations inside your workspace. No language model is called to produce a score, and none of your AI allowance is used. What leaves your server when you do use an AI feature is a separate question, answered in what leaves your server.
From a warning to an action, through the usual door
A warning nobody can act on from where they are reading it becomes a note on a pad. From a pack row you can raise a procurement request, request a stock transfer, create a follow-up task, alert a colleague, pin a report to the dashboard or open a support note. None of them is a shortcut. A request is raised as pending and goes through the same approval as one typed by hand, a transfer moves no stock until it is approved and dispatched, and each action needs the permission it would need anywhere else. Above a value you set, a request or transfer raised from an insight waits for someone allowed to approve AI-suggested actions, and the person who raised it cannot be that someone.
A prediction earns trust the same way a colleague does: by showing its working, and by being checked.
Fourteen days later
You can mark any insight helpful, not helpful, a false positive, accepted or dismissed, and the last two of those need a reason. Fourteen days on, AWRA records what actually happened: the row was resolved, improved, got worse or stayed where it was. Over a quarter, that tells you which packs your team is right to trust and which ones are crying wolf in your business, which no vendor benchmark can.
Three questions for any predictive feature
Why is this item high risk?
What you will probably hear
The model weighs many factors.
How to read it
Then nobody on your team can tell when it is wrong, which means nobody will act when it is right.
Was last month’s warning correct?
What you will probably hear
Our accuracy is very high in general.
How to read it
General accuracy is someone else’s data. Ask for the outcome of your own warnings.
What happens when I act on it?
What you will probably hear
It places the order for you.
How to read it
An action that skips your approvals is a control you no longer have. It should raise a request, not settle one.
What AWRA OpsHub does today
- Seven predictive packs: stockout, supplier delay, late payment, cash pressure, till shrinkage, approval SLA breach and asset custody risk, each a stated calculation over your own records.
- A Why this? panel on every pack row, forecast and anomaly, on the web and in the mobile app, with drivers, window, record count, confidence and method.
- A morning refresh that uses no AI allowance, and a manual refresh.
- Six actions from an insight, each through the module’s normal permission and approval, with high-value requests and transfers waiting for a second person.
- Feedback with a reason, and the outcome recorded fourteen days later.
More we can add to your workspace
- Pack weights and bands tuned to your workspace, so a business with long lead times can treat thirty days of cover as tight.
- A pack built on your own question, such as donor-funded stock nearing its expiry or a project nearing its budget.
- Outcome accuracy shown per pack over time, so the trust question has a chart.
Where we point you to a specialist
- We will not train a model on your records and present its output as a fact. Every score here can be traced to the rows behind it, and that is the point.
A pack on your own question uses the same scoring, explanation and action engine as the seven already built.
Anything above that you need, we can build for you
Everything listed above as something we can add describes what ships in the standard product today — it is a starting point, not a limit on what AWRA OpsHub can do for your organisation. Kenya's eTIMS integration and its maintained payroll engine are both in the product because clients needed them and commissioned them; neither appeared by itself, and the same door is open for whatever you just read about. One qualification so this is worth what it claims: a small number of things on this blog we deliberately leave to a specialist rather than build — a statutory ledger we will not sign our name to, a rule that would decide a tax question for you, a clinical or member-funds record that belongs in a regulated system — and where that is true the post says so in those words. Everything else is a scope, a timeline and a price.
The operational work, which is what most commissions actually are
An extra approval stage in a chain that does not match the standard one, a custom field set on employees or assets that only your sector needs, an expiry that has to block an order rather than send an email, a report your board asks for in a shape nothing produces, or a scanner or weighbridge feeding the goods-in door. These are the commissions we are asked for most often and the smallest ones we quote — and unlike a revenue-authority pipeline, none of them waits on a regulator.
The module-shaped additions, which are the ones readers ask for most often
A price list with real discount authority, a customer-facing quotation that expires, a bill of materials or recipe costing, a staff advance that is issued, acquitted and chased, a member or unit ledger, a matching rule that holds a payment. Each of these is a build rather than a setting, and each has been quoted before — a bigger piece of work than a custom field, with a written spec and a date instead of a roadmap slide.
The report, document or pack nothing currently produces
The board pack in the shape your board actually asks for, a donor or funder layout, an invoice or receipt template carrying what your regulator or your customer expects, a dataset the report builder cannot reach yet. Usually the fastest thing on this list to deliver, because the data is already in the system.
Systems, rails and hardware you already run
The accounting package, CRM, online store, core banking or custom database you intend to keep — connected through our API so a fact is entered once and appears everywhere it is needed. Plus the physical edge: a scanner, a scale, a weighbridge or a till peripheral feeding the door it belongs to.
How it works: you describe the requirement, we return a written scope, timeline and cost, and once agreed it is built into your environment and maintained as part of the product. Nothing here waits on a regulator or a published specification, which is why operational builds are the ones we quote fastest. Tell us the requirement that would otherwise rule us out — that is a better first conversation than a demo.
Tell us what your operation needsSee predictive insights
Seven packs, each with its working shown, actions through your normal approvals, and a record of whether the warning was right.
Explore predictive insightsFrequently asked questions
Are the predictive packs machine learning?
No. Each pack is a stated calculation over your own records: averages, shares, ages and thresholds. No model is trained on your data and no language model produces the score.
Do the packs use our AI allowance?
No. They refresh every morning inside your workspace without calling a language model.
Can an insight place an order on its own?
No. A procurement request or stock transfer raised from an insight is created as pending and goes through the same approval as any other. Above a value you set, it also needs someone allowed to approve AI-suggested actions, who cannot be the person who raised it.
How do we know whether the packs are right for our business?
Every insight can be marked helpful, not helpful, false positive, accepted or dismissed, and fourteen days later AWRA records whether the situation was resolved, improved, got worse or stayed the same.
Which plans include the packs?
Premium and Enterprise include all seven. Pro includes every pack except late payment and cash pressure. Basic does not include AI insights.
