The day a new website goes live feels like a finish line. In practice it is the start of a short, very concrete phase. The first four weeks decide whether the site works as a business tool or whether it quietly fails exactly where work comes from: the form, the phone link, the address, the prices. This article is therefore neither a maintenance plan for the years ahead nor a diagnosis for the case that nothing happens at all. It is a plan with three markers — day 1, week 1 and week 4 — with a fixed order, clear ownership and one principle that matters more than any single task: during the first four weeks you check and measure, you do not rebuild.
Going live is a state, not a date
A website is published once it can be reached at its address. It is functional once a stranger can get in touch through it and finds correct information while doing so. The first weeks sit between those two states. That the step itself is nothing special any more shows in the numbers: 94 per cent (Bitkom, Digitalisierung des Handwerks 2025) of craft businesses run their own website. The basis is a telephone survey of 504 (Bitkom, Digitalisierung des Handwerks 2025) businesses conducted in the summer of 2025 and weighted to be representative. Going live is routine — which is precisely why it is so often ticked off without being checked.
The technical basis inside the business is rarely the problem. Digital equipment is the norm: 95 per cent (Destatis, IKT-Erhebung 2025) of enterprises in Germany used a fixed-line internet connection in 2025, 67 per cent (Destatis, IKT-Erhebung 2025) of them with a contractually agreed transmission rate of at least 100 Mbit/s. Around 20,000 (Destatis, IKT-Erhebung 2025) enterprises with ten or more employees are surveyed for this every year. The difference between a site that produces enquiries and one that stays silent is almost never the connection. It is the details someone has to touch once after launch — and that nobody outside notices while they are missing. Whether building anew was the right move at all is the question before this one, covered in the article on when a new website pays off.
What this article deliberately leaves out
Reachability
Can the site be opened and indexed at all? This check takes minutes and prevents the one mistake that stays unnoticed the longest.
Enquiry paths
Do the form, the phone link and the address work the way a stranger uses them? Only a real run-through answers that, not a look at the screen.
Hard facts
Are prices, services, responsibilities and legal details correct? A wrong figure costs more trust than an unfinished detail in the layout.
Day 1: is the site reachable and indexable?
Two switches regularly survive the build phase, both of which existed to keep an unfinished site invisible: an entry in robots.txt that forbids crawling, and a noindex in the source code that prevents inclusion in the index. During the build both make sense. On launch day both do damage. Google's documentation is unambiguous about the effect: with the noindex rule a page is removed from search results entirely (Google Search Central).
The combination is the tricky part. For a noindex rule to take effect at all, the page must not also be blocked by robots.txt — otherwise the crawler never gets to see the instruction (Google Search Central). Conversely, robots.txt is not a tool for keeping content out of search: an address blocked there can still be indexed if other websites link to it (Google Search Central). Anyone who checks only one of the two places on launch day feels safe while the other one keeps working.
- Open the home page in a private window: no saved login, no browser cache. That is what a stranger sees.
- Open robots.txt: your address plus /robots.txt. A blanket disallow left from the build phase is the first find of the day.
- Search the source for noindex: view the page source in the browser and search for the word — on the home page and on two or three subpages.
- Test both spellings: with and without www, once over http and once over https. All variants should end up on the same address rather than coexist.
- Check the certificate: does the site load without a security warning, including subpages with images and the form?
- Sitemap and search console: submit the sitemap and set up the property for the domain so that you get data at all.
- Look at the error page: call up an invented subpage and check whether a useful error page with a way back appears.
The most expensive leftover from the build
After that, patience is the method. Crawling can take a few days or several weeks (Google Search Central), and a crawl request is explicitly no assurance that a page will be included in search results immediately or at all (Google Search Central). Expecting rankings on day three leads to disappointment. Still not being findable after eight weeks is a different problem — and the ordered checks are in the article on why a website is not showing up in search. How long a project usually takes to reach this point is set out in the article on the timeline of a website project.
Day 1: test the paths that bring money in
The second block on launch day covers the paths that produce work. The expectation of speed is not a feeling but measurable: 87 per cent (Bitkom, Digitalisierung des Handwerks 2025) of the businesses surveyed observe that customers expect individual offers and fast availability. At the same time 85 per cent (Bitkom, Digitalisierung des Handwerks 2025) offer at least one digital service, most commonly digital quotes at 68 per cent (Bitkom, Digitalisierung des Handwerks 2025) and digital invoices at 62 per cent (Bitkom, Digitalisierung des Handwerks 2025). Every one of those paths starts with a message that has to arrive — and nobody sees its arrival when it fails.
Send it yourself, with a marker
Send a test enquiry through the real form with a note such as launch check in the message field. That keeps a record of when it was last verified successfully.
Confirm arrival in the inbox
Look in the real mailbox instead of trusting the confirmation in the browser. Open the spam folder deliberately — form notifications land there particularly often.
Check sender and reply path
Does the notification come from an address on your own domain, and can you reply to it directly? Reply once as if the test enquiry were real.
The second path is even more mundane and overlooked even more often: the phone number. On a screen it always looks right. What matters is whether it can be tapped on a real mobile phone and dials the correct number — with country code and without separators that break dialling. The same goes for the address: a route that ends on the neighbouring plot or in the wrong district costs appointments without anyone ever mentioning it.
- Tap the phone number: on a real mobile phone, not in a preview. Does the device dial the complete, correct number?
- Tap the email address: does a draft open with the right recipient — and is it the address that is actually read?
- Verify address and access: street, house number, district and entrance. A rear yard, an industrial estate or a separate entrance deserves a sentence of its own.
- Opening and availability times: do they match what actually applies in the business, including lunch breaks and voicemail?
- Fill in the form on a phone: are fields, keyboard and submit button usable with a thumb, and is the privacy note legible?
- Follow the confirmation: does the sender see that it worked — and do they receive an acknowledgement if one is intended?
The test nobody can do remotely
While you are in the form anyway, take a look at the number of fields and the required ones. What a form has to deliver so that a visit becomes an enquiry is described in the article on setting up contact forms properly; why availability details age so quickly and how to keep them current is covered in the article on opening hours and availability.
Week 1: proofread the hard facts
The first week is no longer about technology but about content that cannot be estimated. During the build, text is produced quickly, figures are taken from old documents and responsibilities are entered from memory. That is exactly where the expensive mistakes sit: a price that no longer applies, a service the business has given up, an extension that leads nowhere. Unlike a crooked layout, none of this stands out to whoever built the site — it only stands out to the people who act on it afterwards.
- Prices and terms: check every figure against the current calculation, including extras such as call-out charges, minimum quantities or hourly rates. Anything time-limited gets an expiry date in the calendar straight away.
- Range of services: is anything listed that is no longer offered? And conversely, is an area missing that has been asked about for months?
- Responsibilities: are the named people still with the business, still responsible for that topic and reachable on the extension given?
- Imprint: legal form, authorised representative, service address, register court and number as well as the VAT identification number checked against the documents — not from memory.
- Chamber and supervision: for licensed trades and regulated professions, the responsible chamber, the professional title and the state that granted it belong there too.
- Privacy policy: does it describe what the site actually does — form, fonts, maps, analytics, applications?
- Consent: can the consent notice be used, declined and changed later, and does its selection match what the site actually loads?
Two eyes that did not build it
How much detail prices should carry on a website at all is weighed up in the article on showing prices on your website. Which details are expected in an imprint and a privacy policy is collected in the article on mandatory information in legal pages. Both belong in week one because both take effect publicly the moment the site is live.
Week 1: old addresses and redirects
This section only applies to businesses that had a predecessor site — but then it applies firmly. Every address of the old site that is linked, printed or bookmarked somewhere leads, after the switch, either to a suitable destination or nowhere. Old addresses do not disappear because the new site looks better: they sit in directories, in emails, in PDF quotes and in the bookmarks of existing customers.
| Case | What happens without a redirect | What to do in week 1 |
|---|---|---|
| Subpage renamed or removed | The old address ends on an error page; visitors bounce, references lead nowhere | Map every old address to a destination that matches it in content, not a blanket redirect to the home page |
| Change of domain | Existing links and bookmarks point to the old address while the new one starts without history | Set up a permanent redirect from the old domain to the new one and keep the old domain for now |
| www and non-www in parallel | The same content is reachable under two addresses; analytics and links split between them | Settle on one spelling and permanently redirect the other to it |
| Reachable over http without encryption | Browsers show warnings and forms look unsafe | Redirect all requests to the encrypted address and verify it in a private window |
| Linked files and leaflets | Price lists and forms referenced in quotes and emails can no longer be opened | Leave files at the old address or redirect them deliberately, and update references in templates |
How to build a complete mapping, which status codes make sense and how to avoid redirect chains is described in detail in the article on relaunching without losing rankings. For the launch phase the simple rule is enough: map first, switch off later — and do not cancel the old domain in the first month.
Week 2: bring external listings in line
The website is only one of many places where the same details appear. Change it without updating the rest and you create exactly the contradictions that make prospects suspicious. How many places are involved shows in the channels businesses use: 88 per cent (Bitkom, Digitalisierung des Handwerks 2025) maintain entries in online directories, 56 per cent (Bitkom, Digitalisierung des Handwerks 2025) have a presence on social networks or advertise there, 40 per cent (Bitkom, Digitalisierung des Handwerks 2025) are listed on review platforms and 32 per cent (Bitkom, Digitalisierung des Handwerks 2025) run email or newsletter marketing. Every one of those places carries an address, a phone number and often opening hours as well.
Business profile
The listing many local searches run through: address, phone number, opening hours and the link to the new website.
Trade directories
Regional and sector directories, guild and association listings. Often entered years ago and rarely reviewed — but highly visible.
Social profiles
The link in the short description, the contact details in the profile and pinned posts that still carry old addresses.
Email signature
The signature on every mailbox in the business, including mobile devices and out-of-office replies. It goes out more often than any advert.
Vehicles and signs
Lettering, site boards and pavement signs. Not everything can change at once — but then the old address at least has to redirect.
Print and templates
Flyers, business cards, quote and invoice templates, order confirmations. Templates first, printed material at the next reprint.
- Settle on one spelling: write the address the same way everywhere, with the same decision for or against www — inconsistency looks sloppy and complicates analysis.
- Check QR codes: codes on vehicles, signs and print often still point to old subpages. Without a redirect they lead nowhere.
- Track down old phone numbers: numbers from earlier lines survive in directories and signatures for a surprisingly long time.
- Align images and logo: a profile picture from 2019 next to a new website reads like two different businesses.
- Keep opening hours identical everywhere: differences between website and listing are the most common quiet contradiction — and the most expensive one, because people end up at a locked door.
A new website changes nothing about the information already circulating elsewhere about a business. It only makes visible how old that information has become.
In most cases the single most important listing is the business profile, because it appears high up in local results and shows details directly. How to set it up and keep it current is described in the article on the business profile in search engines.
Document access, ownership and deadlines
The organisational part is the least popular one and the one whose absence surfaces last — usually when a particular person cannot be reached. Germany's Federal Office for Information Security asks exactly the questions worth settling in advance for an emergency: who is allowed to decide whether all computers are shut down or the online shop is taken off the network, who can be reached outside office hours, and who is authorised to make decisions (BSI). That these questions are rarely answered shows in practice: only 15 per cent (Bitkom, Digitalisierung des Handwerks 2025) of the craft businesses surveyed have clearly defined procedures for an attack, and 81 per cent (Bitkom, Digitalisierung des Handwerks 2025) have no structured emergency management in place.
- Create an access list: domain, hosting, mailboxes, editing access, analytics and business profile — each with the way in, the responsible person and a deputy.
- Switch on a second factor: the BSI recommends setting up two-factor authentication wherever it is possible (BSI). Domain and mailbox first.
- Manage passwords properly: for larger numbers of accounts the BSI suggests using a password manager (BSI) — notes in the office and shared group passwords are the weaker option.
- Check the domain holder details: the registry requires complete and accurate details including a street address; a P.O. box is explicitly not sufficient (DENIC). The business itself belongs there, not the service provider.
- Note the renewal date: date, payment method and responsible person in the calendar, with a reminder several weeks ahead.
- Create and test a backup: the BSI explicitly advises creating backups and testing them regularly (BSI). An untested backup is an assumption, not protection.
- Put deputising in writing: who takes over during holiday and sickness, and who may make small corrections without asking first?
The domain belongs to the business
How domain, mailboxes and forwarding fit together, and which addresses a business sensibly sets up, is described in the article on domains and business email. For the launch phase a sheet of paper or a table with four columns is enough: service, way in, responsible person, deputy.
Week 4: the first review with data
After four weeks you have your own figures for the first time — few of them, but real ones. The search console is the obvious tool: it gives website owners insight into how their site performs in search (Google Search Central), shows in the performance report how much traffic arrives through search, broken down by queries, pages and countries with impressions and clicks (Google Search Central), and reports in the indexing section which pages were included and where errors occurred (Google Search Central). More important than any single metric is what you do with it.
- Which pages were opened at all? If a service page does not appear, it is either not findable or not linked.
- Which search terms lead to the site? Often they are different words from the ones used internally — that is a hint for the copy, not for the technology.
- How many enquiries arrived, and through which path? Count form, phone, email and walk-ins separately, at least roughly on a tally sheet.
- Which questions came in by phone? Any question asked three times belongs on the site as an answer, usually in a section of frequent questions.
- Where do visitors drop out? If many leave on the pricing page or inside the form, that is where the next sensible change sits.
Four weeks are not a verdict
The first version of a website is a reasoned assumption about what customers are looking for. The first four weeks provide the first evidence of whether the assumption holds — no more than that, but no less either.
How to evaluate search queries systematically and which conclusions actually hold is described in the article on working with search query data. From here the launch phase ends and the routine takes over: the monthly, quarterly and yearly rhythm from the article on website maintenance with a fixed rhythm picks up directly after this review.
The plan at a glance
The overview below summarises what is due when, and how you can tell that a point is genuinely done. The evidence matters more than the tick: almost every fault in this phase comes from someone considering something finished without having used it.
| When | Task | How you know it is done |
|---|---|---|
| Day 1 | Check robots.txt and noindex | Home page and two subpages show no noindex in the source, robots.txt has no blanket disallow |
| Day 1 | Address variants and certificate | All variants end on one address, no security warning in a private window |
| Day 1 | Send the form | The test message sits in the real mailbox and a reply to it comes back |
| Day 1 | Phone link and route | The number dials on a mobile phone, the route ends at the right entrance |
| Week 1 | Proofread the facts | A second person has checked prices, services and the imprint against the documents |
| Week 1 | Map old addresses | A sample of old addresses lands on matching destinations instead of error pages |
| Week 1 | Note access and deadlines | A table with service, way in, responsibility and deputy exists |
| Week 2 | Align external listings | Business profile, directories, social profiles and signatures show the same details |
| Week 4 | Review with data | A short note with figures, open questions and exactly one planned change |
What a factory-fitted delivery takes off the list
A large part of this checklist covers technology that has nothing to do with the business itself: certificates, address variants, sitemap, redirects, the skeleton of legal pages, consent. Whoever assembles those points has to check them as well. Whoever receives them as part of the delivery only checks the result. Which routes exist in principle and how they differ is described in the comparison of website builders, agencies and AI.
Delivery and addresses
Static delivery, one agreed spelling of the address and every variant redirected to it — not as a special task but as the default.
Certificate and encryption
Issuing and renewing the certificate runs in the background; the launch-day check comes down to one visit in a private window.
Sitemap and index control
Sitemap and robots.txt are generated from the actual set of pages. A noindex from the build phase cannot be left behind by accident.
Redirects
Old paths can be stored instead of maintained afterwards in server files — the point where relaunches lose the most.
Legal pages as a skeleton
Imprint and privacy policy are generated as a skeleton from the details of the business. Proofreading stays with the business, collecting the pieces does not.
Consent
The consent notice is part of every page and matches what is actually loaded. That also needs checking — but not building.
What remains is the list nobody can take off your hands anyway: prices, services, responsibilities, availability, external listings and the first review after four weeks. What a site delivered this way looks like in practice can be seen in the demos; if you would like to walk through your own launch list once, a short conversation via contact is usually faster than any template.
Sources and studies