Novus Metrics

Read the whole operation. Defend every number.

Demand and supply, where the work waits, how the machines run, what comes out right and how the routes run — read from the exports the business already keeps, week by week or year by year. Ask for a page in plain words, build it yourself, or open the method underneath; try a change beside how things run now; and trace every figure to the rows and the rule it came from.

Free for 30 days during the beta. Nothing is charged.

The whole operation

Every part of the operation, read from what it already records.

Five areas — Demand & Supply, Bottleneck, Equipment, Quality, Route & Location — each a question and the page that answers it, made from the sales exports, delivery sheets, job logs and machine readings a business already keeps.

Cebu bread · Step 2 · Review · Bottleneck

Where does the time go, and what holds the flow?

1

The data it reads

Work events — a step, a start, an end, and the case it belongs to

✓ already in the Library

2

The page

Every step, the time each takes, and where cases wait

Proof holds the line

2.1 h average wait, 61% of lead time

3

And the what-ifs that read it

Process — add a prover, or split the step

Starting the reviews

Want weekly or monthly reviews, but don't know where to start?

Say what you want to see each week, month, quarter or year. Novus proposes the pages, a question for each, and drafts the review on synthetic sales or a simulated floor from what you tell it, so everyone can agree what to track. As the real files come in, the same pages fill with real numbers, each set beside its preview. Or start from a layout for your kind of work, and build it yourself.

Cebu bread · Guided, then Review › Presentations — a sample
Novus

Start with three questions — a page for each, then the assumptions:

  1. What will customers want next month? · Demand & Supply
  2. Where does the time go? · Bottleneck
  3. How much comes out right the first time? · Quality

Their files aren’t in yet, so I’ll draft it from what you tell me. What’s the business like?

Novus

✓ Made three months of sales and deliveries from that — Synthetic

✓ The floor’s runs are in Log — Simulation

✓ Built “Cebu — monthly review”. Every page says what it stands on.

Demand & SupplyBottleneckQualityPresentations

Cebu — monthly review

for the operations director · 3 sections, then the assumptions

1What will customers want next month?From synthetic data

Needs its data

2Where does the time go?From Studio's simulation

Needs its data

3How much comes out right the first time?From Studio's simulation

Needs its data

Assumptions

4 markets · about 1,200 loaves a day · paydays +20% · delivered Mon, Wed, Fri

Synthetic sales and deliveries, made by Novus from this; Studio’s simulated runs. When the real files arrive, each page sets its preview beside them.

Every number says where it came from

Plan on a preview. Decide on what happened.

A made-up number is only safe to show if it can never pass for a real one. Novus keeps three kinds of data apart everywhere it shows them, and every result says which it was projected from.

Actual
What really happened, from your files. The only data that can raise a signal.
Synthetic
Made-up or modelled figures — yours or Novus's — for trying things out. Labelled on every page that shows them.
Simulation
What Studio's floor produced when it ran. Held against what the real floor did, never in its place.

A page built partly on real data and partly on a preview says that too — “partly synthetic or simulated” — so the moment the real numbers take over is visible, not assumed.

Your files, as they are

Bring the exports you already have.

Sales exports, delivery sheets, a log of when each job started and finished — as they come, including the report-style sheets people fill in by hand, with the date above the table and headers in two rows. Novus works out what each file holds and shows you exactly how it read it, column by column.

The preparation is yours, without a data team or a script. As it reads, Novus finds the table under the header block, joins the sheets that belong together, sorts out dates, returns and currencies, and matches each column to what it holds — and shows every step, so you can point a column somewhere else.

Correct a reading once and it’s kept for that layout: next week’s file reads the same way, with nothing asked. The file stays in the Library as you gave it.

Cebu bread · a board's chat with Novus — sample files
Novus

Three, from what this board already records:

Returns at Cebu · 9.1%

against a usual 4.8% — ₱18,400 at cost last month.

See it in Log › Analyze › Monarch

Pandesal selling out · 1 day in 4

Fri and Sat clear everything delivered, so next week's forecast is understated until that's read in.

See it in Review › Demand & Supply › Forecast

Mixer 2 downtime · 6.5 h

most of it waiting on a part, not repair.

See it in Review › Equipment

Monitor will raise a signal if any of them move — but only on readings you actually measure, never on a simulated one.

The findings are computed, not written

Every figure in a check-up comes out of code reading your rows — the same engines the rest of Metrics runs on. Novus words them and cites each one’s id, so you can check a finding instead of trusting it.

It says what it can’t read

A file it can’t place is kept and named as not read, never quietly forced into shape. When a copy would drop lines, it says how many and why.

CSV or Excel, one file or a folder

Drop a whole folder if that’s how it comes. Each file is listed with whether it was read in, and removing one never takes back what it already read.

What it reads

Movements
sales, deliveries and returns→ Demand, money and waste
Items and locations
prices, costs, shelf life, routes→ Puts money on everything else
Work events
a step, a start, an end, a case→ Where work waits and what gets sent back
Measurements
a reading against its usual range→ Quality, and the signals Monitor raises
Asset events
breakdowns, repairs, downtime→ Equipment losses
Limits and Orders
what can be carried, what's committed→ Load against capacity

Where each kind goes, in full: the User Guide.

Studio

Simulate the floor you're planning.

Lay out the line, counter, clinic or office the work runs through and run it as a flow. The runs become Simulation data, so Review has pages to show from the first afternoon — and once the real floor is recorded, the simulation is held against what it actually did.

Here it is on a small firm’s tax returns: the steps land, the run starts, and the returns pile up in front of the one reviewer.

Tax returns · a small firm · Studio — a sample
▶ Run
Returns done 0Intake to filed 1.9 days

Intake

15m

Prepare

80m · 3

Review

41m · 1

Sign-off

1 day

File

10m

Review holds the line. One reviewer for three preparers, busy 85% of the day — returns queue in front of it.

Try a second reviewer in What-ifs before hiring one.

Monitor · Studio vs actual

  • Returns done a week60→57match
  • Waiting before Review66 min→1 h 35off
  • Review · minutes a return41→43match

Within 10% counts as a match. Waiting is off — returns sit longer before Review than the model thinks.

What if

See what it could look like, beside what it is.

Eight levers — demand, capacity, process time, quality, reliability, supply, policy and money — each moved by a percentage, a shift or a set value, on everything or only the places, items or steps you choose. The staff, process, stock and warehouse engines run the change and set it beside how things run now: as a table, as a chart of what moved, measure by measure.

Tax returns · a small firm · What-ifs · Process — a sample

Lever

DemandCapacityProcess timeQualityReliabilitySupplyPolicyMoney

Move

By %ShiftSet to

Amount

 

On

EverythingIntakePrepareReviewSign-offFile
Run20 replications · same random numbers as now

A second reviewer on Review

Simulated
NowWith it
Returns done a week6064+7%
Intake to filed2.3 days1.8 days−22%
Waiting before Review66 min3 min−95%
Reviewers · busy85%47%−38 pts
Prepare · busy55%61%+6 pts

What moved

Returns done a week
Intake to filed
Waiting before Review
Reviewers · busy

Sign-off is the longest wait now — a day a return. The second reviewer sits idle half the day; try shortening sign-off before hiring.

And also, the editor itself.

For when you want the controls: draw the floor Schematic, Standard or Illustrative, place the teams, set the layout, the calendar and how each step behaves. This is the real Studio, on the same tax-returns floor, running in your browser — open Review, give it a second reviewer, and press Run.

A sample floor, kept in this browser. Item extensions need a Project Board.

Underneath

Open any page, and the method is there.

Every page can be asked for in plain words — and every page can be opened to the rule that made it, changed, and checked. The engines are tested against the published worked answers of university operations-research courses, so a figure you defend upward is one you can work out by hand.

Forecast · Configure

Method

NaiveHolt-WintersCrostonAdvanced

Lifts

Paydays +20%Rainy season −8%Fiesta, 3rd Sun

Buffer from the spread

One σ for all daysOne for each group

WAPE

8.4%

Bias

+1.2%

vs naive

−31%

Forecasting

Naive baselines to beat, Holt-Winters for trend and season, Croston for the slow movers. Paydays, seasons and special days as their own lifts. Buffers from the spread — one σ for all days, or one for each group of days — and accuracy kept as WAPE and bias against the naive.

Stock plan · Rules

On the shelf

Order-up-to(s, S)(r, Q)
s420
S1,100
Lead time1 day

In the warehouse

EOQBase stockWagner–Whitin

Materials

Frozen fence3 days
BufferDDMRP

Stock and materials

Order-up-to, (s, S) and (r, Q) on the shelf; EOQ, base stock and Wagner–Whitin in the warehouse. MRP with time fences and calendars, DDMRP buffers, and where to hold stock across the network (MEIO).

Studio · Options

Warm-up60 min
Replications20
Seed7
✓Common random numbers between options✓Pick the best · KN, 95% sure

Units a day

1,180 ± 24

95% interval, 20 runs

Against M/M/c

within 2%

the queue formulas

Simulation

A discrete-event floor with warm-up and replications, confidence intervals on every output, and common random numbers between options. Rinott's and KN's procedures pick the best option with a stated chance of being right — and the floor is checked against M/M/1 and M/M/c.

Quality · control chart

X̄–RI–MRpnpcu

Cp

1.42

Cpk

1.18

OEE

71%

six losses

Quality and equipment

X̄–R, I–MR, p, np, c and u charts, their baselines kept with who set them and why. Cp, Cpk and Ppk against spec. OEE on total time with its six losses, time between failures and to repair, and when to service by renewal.

Route & Location

Stops11
Vehicles2
Each carries600

Exact loop

64 km

up to 13 stops

Depot to Mandaue

−14 km

a day

Routes and places

Each route's loop, exact up to 13 stops, with vehicles and their capacity; where a depot or a store should go, and what moving it does.

Layers · what this page stands on

Cebu movements 2026-11.xlsxActual
Sales, Aug – OctSynthetic
The floor's runsSimulation

Rule · Advanced, made Oct 29 · 549 rows read

Export CSVExport XLSXHow it's worked out

Every figure, traceable

Which rows each number came from, which rule made it, and whether it stands on actual, synthetic or simulated data — exportable as plain rows, to check in your own tools.

Read the whole operation. Defend every number.

Start a Metrics project, describe the business or lay out its floor, and present the first review this afternoon. Bring the real files when you have them — the review will check its own preview.