kental
NL EN
Contact

For companies without a data department of their own

I have a question.
Nobody can answer it.

The numbers are there. There is just nobody who can work them out again every week, and nobody who can tell you whether the answer is right.

Kental works that question out for you, over and over. It checks for itself whether the outcome is reliable and warns you when it is not. The answer ends up in the system you already work in. You need no programmer and no data department of your own.

Measuring is knowing, and now doing something about it.

Put your question to us Or first see how it works.
The journey in five steps, click to read What convinces me at that point

What Kental is

A question that comes back every week becomes a calculation that repeats itself. You set how often that happens. We check that the outcome is right and put it where you need it.

You click it together

A calculation is a series of steps you put in order: where the numbers come from, what should happen to them, and what you want to know. Programming is allowed, but not required.

  • First see what is in your numbers
  • Compare, look for connections, look ahead
  • Write Python yourself if you want to go further

We check that the answer is right

A calculation can look fine and still give the wrong answer. Kental checks that for every result and says so straight away when something is off, with a better approach attached.

  • Every check with its measured value
  • A warning when you compare too many things at once
  • Plain language explanation under every result

The answer comes to you

You set when the calculation runs. The outcome is written to the system you already use, so your colleagues see it without logging in somewhere else.

  • To your own database, ERP or reporting
  • You decide when and how often
  • Your numbers stay inside your own environment

Who is Kental for?

For companies too big to answer important questions on gut feel, and too small for a data department of their own. Below is what size we mean, who works with it, and why they choose us.

Roughly 50 to 1,000 employees

Big enough that the numbers genuinely matter, small enough that nobody is working on them full time. Exactly the group that falls between two stools.

Without a data team of your own

You have a controller, an analyst, or someone who does it on the side. That is enough. You do not have to hire anyone to get started.

Without an expensive data platform

The well known data science platforms cost a multiple of this before you have calculated anything at all. Kental is built for companies where that does not add up.

Already running Databricks or something like it? Then Kental works on top of it. The heavy calculations stay in your own environment, and you use Kental for the question, the check and the delivery. You do not have to replace anything.

No persona selected, all five phases are visible.

Why companies come to us

You do not have to hire anyone

The question gets answered without building a data team or hiring a consultant for months.

Not a report, an answer that keeps coming

Set it up once and it is there again every time. Nobody has to rebuild it by hand each week.

We tell you when it does not add up

Other programs calculate on and stay silent. Kental checks the sum and warns you before you start steering on it.

What it does
in your line of work

The same way of working, a different question. Pick your sector and you see which question we answer there and what you do with the answer.

Which of our lines is running behind this week?

The same question in the meeting every week, and a different answer every time depending on who you ask.

+6,4 minlonger per order than the norm
31 uurof production time that costs every week
1 van 6lines that genuinely stands out

What you do with it: put line 3 on the agenda. Two other lines looked off as well, but that turned out to be chance. Without that check you would have started three improvement plans instead of one.

Which shipments are going to arrive late this week?

Today you hear about it when the customer calls. The data to see it earlier is already sitting in your own system.

142shipments at real risk of delay
3,2 dagenaverage delay within that group
€ 28kin penalty clauses you get ahead of

What you do with it: call the customers behind those 142 shipments today, instead of them calling you tomorrow. For the rest nothing changes.

Which invoices are going to be paid late this month?

Everyone knows the habitual late payers. The question is which of your ordinary customers slips this month.

€ 412kstays outstanding longer than agreed
11 dagenlate on average in this group
7 van 148customers that actually matter

What you do with it: call these seven before the due date instead of after. You do not need to chase the other 141.

Which systems are running up against their limits?

Today you buy capacity when something breaks, or far too much of it just in case. Both cost money you did not need to spend.

3 serversservers that will be full within six weeks
11 dagenbefore the first one grinds to a halt
9 van 12where you need do nothing for now

What you do with it: order extra capacity for those three. For the other nine you can safely postpone replacement by a year.

How much should we order per location?

One national average means slightly too much or slightly too little everywhere. Nobody calculates it per location by hand.

€ 96kless stock needed across twelve locations
4,1%fewer lost sales
12locations with their own forecast

What you do with it: adjust the purchase orders per location instead of working from the national average.

What will our consumption be tomorrow, hour by hour?

You buy on a daily total, but you pay for the deviation per hour. That gap adds up faster than you think.

2,4%gap between forecast and actual consumption
€ 18klower imbalance costs per month
24 uurahead, each hour calculated separately

What you do with it: buy against the hourly profile instead of a daily total, and correct only where the gap is large.

The five steps

  1. 1 · The question
  2. 2 · Trying it out
  3. 3 · Going live
  4. 4 · Spreading
  5. 5 · Here to stay

Phase 1 of 5

1.

I have a question nobody picks up

I am not looking for software. I am looking for an answer.

This is where you actYou are involved here

What I want

I want to know every Monday which of our six lines is off that week, not worked out once and then forgotten, but something I can steer on while the week is still running.

What I do

I raise it in the meeting. No number comes back, only opinions. I build it myself in Excel, land on a six minute difference for line 3 and cannot tell whether that means anything, and next week I get to do it all again. In the end I just type my question into Google, in my own words.

What I run into

  • Articles that have my question as the title instead of a product name.
  • A talk at our trade association where someone describes my exact problem.
  • A consultant who can work it out. Three months, and a quote I am not taking to my director.
  • Ten dashboard tools that show me what already happened. I know that part.

What convinces me

"I read an article that describes my question, including the pitfall I already suspected but could not name. For the first time I think: someone understands this problem better than I do.

The questions it starts with

Manufacturing

Which of our lines is off this week?

Which orders will we deliver late?

Logistics

Which shipments will arrive late?

Which routes cost more than they bring in?

Retail & e-commerce

Which customers have stopped buying?

Did this promotion work, or would it have sold anyway?

Energy

What will our consumption be tomorrow, hour by hour?

Which connections are off?

Professional services

Which projects are running over budget?

Which quotes are we going to win?

What these questions share: they repeat, the answer belongs in a system that already exists, and an obvious approach quietly gives the wrong answer. That is the test of whether a question fits us, not the industry.

2.

I want to see for myself that it holds up

Fourteen days to try it, no salesperson on the line.

This is where you actYou are involved here

What I want

Try it myself, without a conversation with an account manager first. If I do not get it within an afternoon, it is not for me.

What I do

I create a trial account and use the sample data. Connecting my own data is not even allowed yet. I build my analysis in a few blocks and hit run, with the test I know from my training.

What I run into

  • Building blocks on a canvas instead of an empty code screen.
  • Under every result an explanation in plain language, which I can forward to my manager.
  • A twelve minute demo I can also watch without anyone getting my phone number.
  • A warning I had not asked for.

What convinces me

"The platform stops me. My test is not allowed on this data, it says, and here is why, with an alternative attached. I do not feel caught out. I feel protected.

The canvas, from question to scheduled export

This is what my analysis looked like: no empty code screen, but six blocks on a canvas. Each block is a step, from my source data to the answer that refreshes itself, as often as I want.

1
Connect

Connect the source through the guided wizard. Here: six production lines, week 29.

2
Profile

Kental reads the columns and sees that the lead times are right skewed.

3
Test

You pick the t test you know. A guardrail reads along, unasked.

Guardrail active
4
Assumption ledger

The assumption fails. Kental proposes Mann-Whitney and corrects using Holm.

Runs below
5
Forecast

Four blocks turn the corrected result into a forecast per line.

6
Schedule and export

Runs whenever you want, daily or hourly. The answer is waiting in your Delta table or ERP.

Block 4 is the block you can run yourself below. The rest of the chain is what surrounds it: from your own source to an answer that comes back by itself without you doing anything.

The moment itself, try it

This is exactly what happened on my screen. I test our six lines against each other, with the t test I know. The lead times turn out to be right skewed.

assumption ledger · weekly line comparison

Six production lines, week 29. Three lines appear to deviate from the average. The question is not how big the differences are, but which of them are real.

PassedIndependent samplesn = 2,410 measurements
PassedEqual variances (Levene)p = 0.214
RejectedNormality (Shapiro-Wilk)p = 0.003
RejectedCorrection for multiple testing6 tests, no correction
Two things do not hold here. The lead times are right skewed (skewness 1.84), which makes the t test unreliable, precisely for the outliers that matter most. And whoever tests six lines at once finds one "significant" on average purely by chance. Kental switches to Mann-Whitney and corrects the outcomes using Holm.
After correction one line remains: line 3 · p = 0.018
Its median lead time sits 6.4 minutes higher. The two other "deviations" fall away, they were chance. Without these two checks, an improvement plan would have been started for two lines where nothing is wrong.

This is what a finished result looks like

Every result carries its own evidence: a stamped checklist of the assumptions, with the measured value attached. One look and you know whether you can build on this number. The square closes it off when everything holds.

Weekly line comparison · week 29
Line 3 runs 6.4 minutes behind
Reliable
+6.4minutes longer per order than the norm
31 hrsof production time that costs every week
1 of 6lines that genuinely stands out

What you do with it: put line 3 on the agenda. Lines 2 and 5 stood out in the raw numbers too, but stay within the margin. Without the checks below, improvement plans would have been started for those as well.

What this answer was checked against
Enough measurements to draw a conclusion2,410 orderschecks out
The lines are comparable to each otherverifiedchecks out
Outliers do not skew the averagedifferent test chosenchecks out
Corrected for six lines at once6 comparisonschecks out
3.

It has to work on my own numbers

Between my enthusiasm and the result stands IT.

This is where you actYou are involved here

What I want

My own data in it, and one analysis that repeats itself, so I do not have to run and format it again every Monday.

What I do

I request access from IT. I wait. I send a reminder. Then I connect the source, look at what is in the data, rebuild the analysis on real numbers and put it on a schedule.

What I run into

  • A wizard that walks me through the connection step by step.
  • Three questions from IT I cannot answer myself: where does our data sit, who can reach it, and what if we stop.
  • Eleven days where nothing happens and I feel like I bought the wrong thing.
  • My own numbers, which turn out messier than I thought.

What convinces me

"The first morning the answer is already there, in the system I open anyway. I did nothing for it that day. That is the moment it stops being a trial.

4.

My colleagues want this too

I sold nothing. I just showed my answer.

This is where you actYou are involved here

What I want

That the people who do something with my numbers can simply see them, without me making an export and mailing it round every week.

What I do

I share the link with a few colleagues. I add two analyses. And because I get curious, I try a Python block for the first time, for something the building blocks just cannot do.

What I run into

  • Colleagues who have already seen my numbers before I send them round.
  • My chart in a presentation I did not make myself.
  • A second analysis I set up in an afternoon, because the source is already there.
  • A colleague asking whether I can build this for their department too.

What convinces me

"The department next door asks whether they can have it too. Not because I recommended it, I did not even know they were watching. They had already seen it.

5.

I pass it on

The one who hit the brakes now defends it.

This is where you actYou are involved here

What I want

That this stays. No more annual discussion about whether we carry on, and no dependence on my own enthusiasm.

What I do

I help IT with their requirements instead of working around them. I stand next to our IT manager at an industry afternoon. And I pass it on to someone at another company with the same question I had eighteen months ago.

What I run into

  • Logging in with our own accounts, an audit log, everything in our own environment.
  • An IT manager who suddenly talks about "our platform".
  • Questions from peers I know from the association: does it do what it promises?
  • Our communications department, careful about our name in a customer story.

What convinces me

"IT presents it internally as a win: fewer loose spreadsheets, everything traceable. The person who held me back at the start now defends it in the management meeting.

In practice:
your receivables

Take the finance question. Everyone knows the habitual late payers; the question is which of your ordinary customers slips this month. Kental works that out and puts the answer straight into the overview your controller already uses.

What your colleagues see

No new program to open. The list appears in the receivables overview you already go through, with the expected payment date next to it.

Receivables overview 148 customers · updated 04:00
customerdue dateamountexpected indays lateaction
Van Dijk Logistiek7 Aug€ 128.40026 Aug19call
Meijer & Zn11 Aug€ 96.75025 Aug14call
Brouwer Techniek14 Aug€ 41.20015 Aug1none
De Wit Groothandel18 Aug€ 33.90017 Aug0none
Hoekstra Bouw21 Aug€ 27.60020 Aug0none
Filled in by Kental · 5 of 148 rows shown · everyone with access to this overview sees it

What to
expect

An honest picture of how a project like this runs, including the things that slow it down. So you know in advance what you are saying yes to, and what we need from your side.

1

Start with one question, not five

Pick one that repeats every week and that someone is waiting on. That first question decides whether the rest follows.

2

You do not need a data scientist

You do need someone who knows where the numbers sit and what they mean. Usually the analyst doing it in Excel today.

3

The connection goes through your IT

Access to the source is almost always the longest part. Involve IT in week one, not at the first error message.

4

Your own data is messier than you think

Empty fields, three spellings of the same name. Count on a cleaning round before the first result holds.

5

Scheduling is the real starting point

A one off analysis is an experiment. Only when an answer is waiting at fixed moments does anything change about how you work.

6

Do not let it rest with one person

If only one colleague knows how it fits together, it stops the moment they go on holiday. Put a second person beside them from the start.

One answer calls up
the next one

It does not stop at one analysis. The answer lands in the system your colleagues already open, they see what is going on, and the next question follows by itself. That is how it grows from a single question into the way you steer.

k THE LOOP You ask a question THAT COMES BACK EVERY WEEK Kental works it out AS OFTEN AS YOU WANT The answer lands WAREHOUSE · DELTA · ERP Your colleagues see it AND ASK THE NEXT ONE
  • The answer comes to your colleagues. It sits in the overview they already open, so nobody has to learn another tool.
  • No more mailing exports around. The result refreshes itself, so the numbers in your meeting are always the most recent.
  • The next question follows by itself. Whoever sees the answer also sees what is still missing, and that is usually the next analysis.
  • What you are missing, we build. The blocks customers ask for decide what comes next in the library.

Does this work for your question?

Send us the question that comes back every week at your company. We will show you what it looks like in Kental, with your kind of data, and we will say so honestly if it does not fit.

Put your question to us

No sales pitch. You get a concrete answer to your own question.