Zepter Technology Case Study

How I looked

Two days, one browser,
one ordinary customer account

I read Zepter's public pages, twelve country websites, the ZepterClub rules and the app stores. Then I signed up as a Member and used my own account. I did not test or attack anything. Everything here came from pages Zepter serves to any visitor.

Which means I never saw inside Zepter. No code, no servers, no meetings. Some of what follows will be wrong, and your team will spot which parts in minutes. That is fine, and it is why I have marked every place where I had to guess. The last section lists those guesses and what would change my mind.

The business

The products are expensive.
The people are the engine.

60+countries on five continents

$1.1Broughly, in yearly sales

100,000+people have sold for Zepter

8factories in Germany, Italy and Switzerland

5% to 40%the advertised ZepterClub ladder

Everything below is judged on one question: does it help the people who sell? Zepter is not a shop with a sales team attached. It is a sales network with a shop attached, and the technology should be built that way round.

One thing worth saying out loud

The ZepterClub rules name the operator as Zepter International Poland sp. z o.o., 37 Domaniewska, Warsaw. Polish law governs it. A Polish court settles disputes. That Polish company is also the data controller for every Member who signs up on the international shop.

So the site called international is, legally, a Polish operation. That is not a criticism. It explains the Polish phone number in the footer, the Polish dialling code the sign up form defaults to, and where the technology centre of gravity already sits. It also decides who carries the risk in the two sections further down.

Proof 1, technology checks

What the website handed
my browser, unedited

Every line below is something Zepter's own pages sent to an ordinary visitor. I did not craft, repeat or modify a single request.

seen on 26 and 27 August 2026 · shop.zepter.com · public pages only

POST	/Kentico.Activities/KenticoActivityLogger/Log	the shop runs on Kentico, software Zepter licenses
cookie	owin.authentication · httpOnly · 14d	older Microsoft login technology, not the current one
cookie	ASP.NET_SessionId	older Microsoft session handling
POST	firehose.eu-west-1.amazonaws.com	Amazon data pipeline, Ireland
POST	od-*.ecs.us-east-2.on.aws/events	a second collector, this one in the United States
GET	zepter.com → 301 → shop.zepter.com	the group's main address lands on the oldest site
console	[Meta Pixel] Duplicate Pixel ID: 975731019757356	every sale is reported to Meta twice
form	Google reCAPTCHA on registration	abuse protection already in place, keep it

Kentico changes the entire conversation. The flagship shop is not a strange one off system that only Zepter has. It is a licensed product with a vendor, a support cycle and a published upgrade path. So the question is which version to move to and when, not whether to rebuild. Anyone who tells you to replace it has not looked.

Proof 2, the websites

Three generations of technology,
all in production today

Website	What I saw	Generation
shop.zepter.com	Kentico activity logging, plus older Microsoft login and session handling	Kentico, older Microsoft
marie.zepter.com	The same signals. A storefront for one individual seller, on the same old platform.	Kentico, older Microsoft
zepter.pl · .rs · .hu · .ro	Current Microsoft technology, and identical cookie names across every country	current Microsoft
zepter.de	Apache, with a WordPress caching plugin named in the response headers	WordPress

Somebody has already begun moving these onto current Microsoft technology, and the identical cookie names say they are doing it from one shared codebase rather than rebuilding per country. That is the right call and it is half done. Finish it. Do not restart it, and do not open a competing programme beside it.

Proof 3, the shop window

I was logged in when
I took this screenshot

This is the Zepter homepage, in my own signed in Member account. Look at what the price panels say.

Four product price panels on the Zepter homepage showing Register slash Log in while the visitor is already logged in, and Member discounts of 0 percent beside a claim of 5 to 40 percent — **Screenshot**shop.zepter.com homepage, signed in as a Member. Every panel still says **Register / Log in**. There are 48 of them on this page alone.

Costing money now

The price panel does not know who is looking at it

A signed in Member is asked to register, on every product, on every page. The most valuable component on the site, the one carrying the price, is the one component that ignores the customer.

This is a caching or personalisation fault, not a redesign. It is days of work, and it is suppressing conversion on every visit by a logged in customer right now.

Also a legal exposure

The discount claim contradicts the discount

The panel reads Member −0%, € 62.00, directly beside the words You buy from −5% to −40%. On the kettles the Partner rate shows −2%. Both sit below the advertised floor.

The operator is Polish, so this falls under EU unfair commercial practices rules, and Zepter is a member of the Polish Direct Sale Association and bound by its code. Either the claim changes or the pricing does.

I want to be careful here, because this is the sort of thing that gets a candidate told off in the room. I am not saying anyone did this deliberately. Discount bands almost certainly vary by product and someone wrote one strapline for all of them. That is exactly why it matters: nobody owns the shop window, so nobody noticed.

Proof 4, what already works

The seller tools are real,
and better than they look

From outside the account there appears to be nothing. Inside there is a working offer builder: pick a discount tier, attach a promo code, set how many days it lasts, send it to a named person, export it as a PDF.

The ZepterClub offer builder inside the member account, showing discount tiers DL1 to DL4 and a list of expiry days — **Screenshot**Inside my own Member account. Discount tiers **DL1 to DL4** and an expiry list running to 30 days, both opened out. The left menu shows ZepterClub Orders, Recommendations and Offers.

I had this wrong before I logged in, and I am saying so on purpose. Do not let anyone replace this. Somebody built a real quoting tool for the sales network. What is missing sits on top of it, not underneath it.

Proof 5, who gets paid

Three ways to earn the credit.
No way to see them together.

The list of ZepterClub members someone has recommended, with a Recommendation thru column — **Screenshot**The recommendations list inside my account. The final column is **Recommendation thru**, so credit is stored properly on Zepter's side with an expiry date, not only in a browser cookie.

How credit is captured	How long it survives	Where it lives
A cookie set when someone visits a seller's own storefront address	about 5 days	browser, one device
The customer types their seller's phone or email while registering	permanent	Zepter's systems
A promo code offer sent to one named person	1 to 30 days	Zepter's systems

All three work. Nothing reconciles them, so no single view answers the only question a seller actually cares about: was I paid for that sale? The data is already there. What is missing is the view over it, and that is weeks of work, not a platform programme.

Proof 6, the borders

One browser, one login,
two different answers

The international shop header showing the account logged in as Marko Lukic — **Logged in**shop.zepter.com. The account is signed in, top right.

The Serbian website still asking to log in or sign up, in the same browser — **Same browser, seconds later**www.zepter.rs still asks me to **log in or sign up**. The session belongs to one hostname.

A seller working two countries needs two accounts, and the people they recruited do not follow them across. For a business whose entire model is a network that crosses borders, this is the most expensive thing I found, and it is the one that quietly caps how large any individual seller can grow.

The same split shows on the way in

On the international shop the address country list runs Romania, then Slovakia. Serbia is not on it. Someone in Belgrade can register with a +381 number and then find they cannot enter an address.

The country list on the international shop, with Serbia missing between Romania and Slovakia

Proof 7, the phone

100,000 sellers.
One app. For a fridge.

Zepter International publishes exactly one Android application. It is called Therapy Air, it monitors the status of an air purifier, and it has roughly a thousand downloads. There is no seller app, no shop app and no ZepterClub app.

What sellers get today

A website designed for desktop, and email. The ZepterClub rules state that all communication happens by email and nothing else.

Direct selling happens standing up, in someone's kitchen.

What the group can clearly do

Zepter Bank ships a full corporate banking app, multi company, multi currency, biometric login, document signing. Built by an outside software house.

So the group knows how to buy and run serious mobile software.

What that gap tells me

Mobile has never been owned by anyone in the retail business. It is not a capability problem and it is not a money problem.

It is an ownership problem, which is cheaper to fix.

I am not proposing a twelve month app programme. The first version is the seller's existing account, made to work properly on a phone. That is a front end job measured in weeks, and it would reach more sellers than anything else on this page.

Proof 8, the paperwork

The published rules cover
Members. Not Partners.

Every page on the shop links to the ZepterClub Terms and Conditions. I read them. They define Members, Member Prices and the loyalty programme, and they say a Member buys only for their own use and may not buy in wholesale quantities.

The word Partner does not appear. Neither does commission. Yet every product panel on the site advertises a Partner tier earning 5% to 40%. So the terms governing the tier that actually generates revenue are not in the document the site presents as the programme's terms.

The likely explanation

Partners sign a separate agreement

Almost certainly there is a commercial contract for Partners, held elsewhere and not published, which is entirely normal for a direct selling business.

I could not check it. Becoming a Partner means signing that contract, and I was not going to sign a commercial agreement with a company while applying to work there.

Why it still matters

A person is asked to choose a tier before they can read its terms

The registration form asks you to pick Member or Partner, with a single tick box accepting the published terms. Those terms describe only one of the two options.

For a direct selling business regulated in Poland and bound by an industry code, that is worth ten minutes of a lawyer's time.

The diagnosis

Eight findings.
One cause.

Credit that cannot be reconciled. A shop window that ignores the customer. A login that stops at the border. A country missing from a list. No phone product. Rules that cover half the programme. None of these are hard problems, and that is the clue. They are all things that happen when nobody owns the whole customer journey across the whole group.

The technology is in better shape than the organisation around it. Kentico is supported, the move to current Microsoft technology is real and correct, the offer builder is genuinely good, and the data needed for the missing views already exists. So the first year is not a rebuild. It is putting an owner on things that currently have none, and adding one layer across the top.

Keep, improve, connect, replace, missing

Most of it stays

Keep

Business Central, the Microsoft migration, Kentico, the ZepterClub model, the offer builder

Business Central holds the records for a group turning over a billion dollars. Connect to it and leave it alone in year one. The Microsoft migration is correct and already funded work, so finish it. Kentico is a supported product, so choose the target version deliberately.

You asked candidates not to assume everything needs changing. I would have written this section anyway.

Improve

The shop window, and one honest view of sales credit

Fix the price panel so it recognises a signed in customer. Align the discount claim with the discount. Fix the duplicated Meta pixel so the sales numbers are real. Add the missing countries. Then build one read only view that reconciles the three ways credit is captured.

Every item here is days or weeks, and every one of them is costing money or credibility today.

Connect

One doorway into Business Central, one identity, one event stream

Put a single integration layer in front of Business Central so no system builds its own private coupling to the ERP. Then one account that works in every country, beginning with the sites already sharing a codebase. Then send commercial events deliberately into the Amazon pipeline that is already running.

This uses infrastructure Zepter already pays for.

Replace

Almost nothing, deliberately

The WordPress country site, folded into the shared codebase the next time that market needs work anyway. Country by country front end fixes, replaced by one team that owns the shop window everywhere.

Nothing else in year one. Every replacement spends trust and halts something that currently earns.

Missing

A phone product for sellers, one identity across borders, and one reconciled view of earnings

These are the three things the business does not have, and all three serve the same people. I have deliberately not claimed there is no Partner earnings screen, because I could not check without signing a commercial contract. If one exists, this section shrinks to two items and the plan moves a quarter earlier. That is the single most useful thing you could tell me.

Risk register

What I would raise
in my first board meeting

Ranked by what it would cost if it went wrong, not by how hard it is to fix. All of it is visible from outside, which means it is visible to other people too.

Risk	What I saw	Why it matters
Personal data leaving the EU	A Polish company is the data controller, and page events post to a collector in the United States	Check the transfer basis. This is the first thing a regulator or a large client asks about
Advertised discounts that are not delivered	Member −0% shown beside a claim of −5% to −40%	Consumer protection exposure in the EU, and an industry code Zepter has signed
Terms that do not cover the paid tier	Published ZepterClub rules describe Members only	Ask legal. Cheap to check, awkward to discover later
An unsupported platform version	Kentico, running on older Microsoft technology	Security patches follow support. Establish which version and its end of life date in month one
Decisions made on wrong numbers	Sales reported to Meta twice	Marketing spend is being judged against inflated results
Key person and vendor concentration	One shared codebase across many countries, and outside software houses building group products	Map who can actually deploy to each market, and what happens if they leave

I want to be exact about what this list is. I did not scan, probe or test anything, so I am not claiming Zepter has a vulnerability. These are risk areas that are visible from the outside and that I would want answered in writing in my first month. Anyone who hands you a security assessment of your systems from the outside, without permission, has told you something about how they would behave once inside.

The plan

Stop the leaks, then build
the layer that is missing

The first month ships nothing except answers. Everything after that is sequenced so each phase pays for the next.

130 days

2100 days

36 months

412 months

First 30 days

Replace my guesses with your facts

Kentico version, licence and end of life date
Business Central: one instance or one per country
What a Partner account actually shows today
Who can deploy to each country website
Where personal data physically goes
Every outside supplier and what they hold

Result: a written map, a risk register with owners, and a costed backlog. No code, on purpose.

By 100 days

Stop the money leaking

Price panels recognise a signed in customer
Discount claims match the discounts
Meta pixel fixed, marketing rebaselined
Missing countries added to checkout
One reconciled view of sales credit, read only
One team owning the shop window in every market

Result: revenue and trust recovered from work that is measured in weeks. Nothing that trades was rebuilt.

By 6 months

One identity, one doorway

One account across the current Microsoft country sites
A single integration layer in front of Business Central
The seller's account working properly on a phone
Commercial events flowing into the Amazon pipeline
Kentico target version agreed and the upgrade started

Result: a seller works two countries with one login, from a phone, in the field.

By 12 months

The layer the business never had

Earnings, recruits and payments in one place
Older country sites joined to the shared identity
Legacy properties on a supported platform version
AI where it pays: routing, quality, personalisation
Group standards for data, security and integration in force

Result: the sales network finally has a system built for it rather than around it.

No headcount and no budget figures here, on purpose. Questions 2 and 6 in my email asked how the Poland organisation is structured and whether a programme and budget already exist. Without those, any number I wrote would be invented, and you would be right to discount the whole document for it.

Organisation and money

A small central team, and
owners for things that have none

Build in house

The commercial layer

Sales credit, identity across countries, the seller's view of their own business. Nobody sells this off the shelf because it is the shape of Zepter's own model. This is the part that is genuinely Zepter's.

Buy

Everything around it

Identity, payments, analytics, content management, marketing automation. All have mature vendors and none of them differentiate Zepter. Kentico stays, and the question is the version.

The shape of the team

Three owners, not a department

One team owning the shop window across every market. One owning the commercial layer end to end. One owning integration and data between Business Central and everything else. Country teams keep their local knowledge and stay where they are.

This is an ownership map more than a hiring plan. Most of these things exist and simply have nobody's name against them.

Where the money goes

People first, licences later

The first two phases are people and work already paid for. Capital spend arrives with the Kentico upgrade decision, in the second half of the year, once the estate map exists to size it.

On suppliers: I have run the vendor side of an outsourced engineering relationship for six years, so I know how these contracts are written and where they leak.

How to judge me

What I should be measured on

Several of these cannot be measured today. Making them measurable is itself the first result, and I would expect to be held to it.

Measure	Today	Target
Conversion for signed in customers	Suppressed. Every price panel asks them to register.	Fixed inside 100 days, and the lift measured against today
Accuracy of reported sales	Overstated. Every sale reported to Meta twice.	Corrected in weeks, then all marketing rebaselined
Sales credit disputes from sellers	Not measured, because nothing shows it	Baseline by day 100, then falling every quarter
Sellers active in more than one country	Structurally impossible. The login stops at the border.	Possible from month 6, then tracked as a growth number
Share of seller activity on a phone	Effectively zero. There is no product.	Measured from month 6 and expected to become the majority
Release frequency and lead time	Unknown across the country estate	Baselined in 30 days. I have cut deployment time by 85% doing exactly this before

Where I guessed

What I assumed, and what
would change my mind

Guess 1

Partners have no reconciled earnings view

I could not check without signing a commercial contract while applying for a job here.

If one exists, the twelve month phase is largely done and I would spend that year on identity and mobile instead.

Guess 2

Business Central runs per country

Identity, domains and the commercial layer all split by country, so I assumed the ERP follows the same pattern.

If it is one group instance, integration is simpler and the six month work starts sooner.

Guess 3

Defining the programme is part of this role

Rather than executing a programme that is already approved and funded. If one exists, the first thirty days becomes joining it rather than writing it, and I would want to see it before saying anything else.

You said you were less interested in what technology can do in theory than in what should actually be done at Zepter, why, and in what order. So here it is in one sentence. Keep almost everything, finish the migration already under way, stop the handful of leaks that are costing money this week, and then build the one layer the business has never had: a single view of the sales network that earns the money.

I have done the smaller version of this. Two live platforms consolidated into one customer portal, enterprise clients migrated across, no pause in trading, and deployment time cut by 85% along the way. Zepter is larger and spread across more countries. It is the same shape of problem.

Happy to talk through any of this

Marko Lukić
marko.lukic.info@gmail.com · +381 61 3555354 · Belgrade
Written in August 2026 for Zepter International

Zepter does not havea technology problem.It has forty of them.

Two days, one browser,one ordinary customer account

The products are expensive.The people are the engine.