Guide
What to plan before a website build, and after launch
By Mavrin Labs. Published . Last updated . Reviewed by Mavrin Labs on . How this guide was researched and checked.

Decide four things before a website build: who each page is for, what one job it does, where an enquiry goes after it arrives, and which addresses must survive the move. Design sits on top of those answers, and the guide works through each of them in order.
The guide is written for owners and managers planning a new site or a rebuild. It covers structure, enquiry routing, redirects and the checks to run before accepting the work. Visual design, copywriting and hosting are separate subjects and are left out here.
What is website planning?
Website planning is the set of written decisions made before any design work, naming who each page serves, the one job it does, where each enquiry goes and which old addresses must keep working, so the build has answers to refer to. Design sits on top of it.
It is not a sitemap drawn for its own sake or a mood board. It is the short document that the builder, the writer and whoever answers enquiries can all work from, and the sections below fill it in.
Why it matters: much of the web still misses the basics
More buying now happens online, so a site that loses enquiries loses real business.
The Census Bureau estimated U.S. retail e-commerce at $340.2 billion in the second quarter of 2026, seasonally adjusted, or 17.1% of all retail sales (the Census Bureau's quarterly e-commerce report, checked Monday, November 16, 2026). Its next release, due Thursday, November 19, 2026, may revise those figures.
Many sites still miss the published performance thresholds. The HTTP Archive's 2025 almanac found good core web vitals on 48% of mobile sites and 56% of desktop sites, using July 2025 data (its performance chapter, checked Monday, November 16, 2026). Planning for speed and access before the build is easier than repairing both after launch.
What should you decide before a website build begins?
Four decisions shape a website build, and all of them come before layout. Name the audiences the site serves, give every page one job, plan the route an enquiry takes after it arrives, and list the addresses that have to keep working. Everything visual sits on top of those four decisions.
The common failure is not an ugly site. The common failure is a site that looks finished and loses work. An enquiry arrives in an inbox nobody owns, and the customer record is retyped from an email.
A page tries to serve three audiences and persuades none of them. Old addresses stop existing, and the standing those pages had built goes with them.
None of that is a design problem, which is why it survives a redesign. Each item is a decision somebody has to make and write down. Make them first and the build becomes straightforward, because every later question has an answer to refer to.
A site that looks finished and loses enquiries is the normal outcome of planning the design before planning the work the site has to do.
Which questions does each audience arrive with?
Start from the people who will arrive and what they came to settle. Most small-business sites serve three or four distinct arrivals, and each one needs a page that answers its question quickly rather than a page that introduces the business.
- Ready to act
- Knows what they want and is checking you are credible and reachable. Needs proof, a clear route to book or ask, and nothing in the way.
- Comparing options
- Knows the problem, not the answer. Needs the approaches compared honestly, including when yours is the wrong fit.
- Still diagnosing
- Has a symptom rather than a problem. Needs plain explanation first, and a next step that is reading rather than buying.
- Returning or referred
- Was sent by somebody or is coming back. Needs the specific thing they were told about, found in one step.
Write the question each audience arrives with in their own words. Those sentences become headings, and a heading written that way is easier to answer honestly than a heading invented to sound impressive.
Why should every page do one job?
Every page should do one job, because a page that answers one question for one audience and offers one next step is easy to judge. The rule settles most structure arguments, since a page that needs two jobs is really two pages.
- The first heading states the subject; the first paragraph answers it.
- One audience is addressed, and named when that is not obvious.
- One next step is offered, the one that audience is ready for.
- Nothing on the page exists only because a competitor has it.
- The page makes sense to somebody arriving straight from a search result.
The last item matters more than the navigation does. A visitor who arrives from a search result or a shared link may never see your home page, so every page has to work as a first impression on its own.
The same test appears in published search guidance. Google says its ranking systems are designed to favor helpful, reliable information made for people, and suggests asking whether a page was made mainly to attract visits from search engines (Google guidance on helpful content, checked Saturday, October 10, 2026).
What happens to an enquiry after it arrives?
An enquiry should arrive in one place with one owner and create a record in the CRM (customer relationship management system), the program that holds client and prospect records, so nobody types it twice and every later step starts from that one record.
It should also acknowledge the sender in terms they recognize and appear on somebody's list of things to do. Decide each of those before the build, not after it, because the route is where sites leak.
A form that sends an email is only the first step, and an email is not a record. If the details then have to be retyped into the CRM, that retyping takes staff time and invites mistakes. The same thing happens on every enquiry for as long as the site exists.
- Arrival in one place. One destination for every enquiry, whatever page it came from, so nothing depends on somebody watching a second inbox.
- A proper record. The details land in the CRM, with the page and the question kept, so the first reply has context.
- An acknowledgement that helps. The sender gets a reply that names what they asked about and says what happens next, which is the moment most sites waste.
- One named owner. One named person sees the enquiry on a list and is answerable for it, so a quiet failure gets noticed within a day.
Plan the unhappy paths too. Decide what you want to happen when a delivery fails, when somebody submits the same enquiry twice, and when a field arrives empty. Joining those systems dependably is its own piece of work, which the integration service page describes.
How do you protect what the old site has earned?
Protect the old site's standing by keeping every address it had working. A rebuild moves addresses, while search engines, links and bookmarks all point at the old ones. Treat the address list as a deliverable: export it before the build, map every old address to its closest new equivalent, and apply a permanent redirect for each.
A permanent redirect is an instruction from the server that a page has moved for good, so browsers and search engines go to the new address instead.
The documentation for site owners published by Google says the 301 and 308 status codes mean a page has moved for good. Search results then show the new address, while a temporary redirect keeps the old one showing (Google documentation on redirects, checked Thursday, October 8, 2026).
The table below compares the three outcomes an old address can have.
| What the old address does | What a search engine shows | Use it when |
|---|---|---|
| Permanent redirect | the new address | the move is final, which is almost always |
| Temporary redirect | the old address | the move will be reversed soon |
| Nothing, and the page is gone | a missing page | never, for a page that had visitors |
- Every address on the old site is listed before the build starts.
- Each one maps to its closest equivalent, or to the nearest useful parent when no equivalent exists.
- Each mapping is a permanent redirect, applied once and never chained through a second one.
- Each page has one canonical address, the single address you want search engines to treat as the real one.
- Pages that are being dropped are a deliberate decision, recorded with the reason.
- The list is tested after launch, address by address, not sampled.
Which address does a search engine treat as the real one?
A permanent redirect also tells a search engine which address to treat as the real one. Google lists redirects first among the ways to name that canonical address, ahead of a canonical link in the page and of a sitemap entry, which it treats as a weak hint (Google documentation on canonical addresses, checked Saturday, October 10, 2026).
Resist the temptation to send everything to the home page. A redirect to a page that does not answer the original question is a dead end for the reader and reads as a dropped page to a search engine.
How do you check the finished site?
Check the finished site against published criteria rather than against taste: the W3C's accessibility guidelines, version 2.2 of December 2024, and Google's performance thresholds. Both are public and testable, so you can hand a builder the list and compare the result (the W3C guidelines, checked Tuesday, November 17, 2026).
Accessibility comes first, because a page some visitors cannot use loses their enquiries before they begin. Both standards are easier to meet during the build than to retrofit.
How have the accessibility guidelines changed?
The accessibility guidelines grow by adding criteria. WCAG 2.1 became a W3C Recommendation in June 2018, and WCAG 2.2 adds nine new success criteria and removes one obsolete criterion. W3C dates the current edition 12 December 2024, and it does not supersede the earlier versions (WCAG 2.2 from W3C, checked Thursday, October 8, 2026).
The additions are practical. They cover keeping the focused control in view, offering an alternative to dragging, a minimum size for pointer targets, help placed consistently, not asking people to re-enter details they already gave, and signing in without a memory test (the list of new criteria in WCAG 2.2, checked Monday, November 16, 2026).
Ask a builder to work to the current version, and to say which criteria the finished site was tested against.
What mistakes cause regret after a rebuild?
Four mistakes often cause regret after a rebuild. They are designing before deciding, treating a form as a record, losing the old addresses, and accepting the site on a laptop. Each has a plain correction, set out below, and each is easier to avoid during planning than to fix after launch.
- Designing before deciding
- Layout drawn before the audiences and page jobs are settled gets drawn twice. Write the four decisions down first.
- Treating a form as a record
- An email from a form is not a customer record. Decide which system owns the record and how the details reach it.
- Losing the old addresses
- Pages that vanish take their standing with them. List every old address and redirect each one permanently.
- Accepting the site on a laptop
- The builder's machine is not the visitor's phone. Test the first pages on a real device before you sign off.
Failing the accessibility criteria is the normal state of the web, not an unusual lapse.
In February 2026 WebAIM tested the home pages of the top one million sites and detected WCAG 2 failures on 95.9% of them, at an average of 56.1 errors a page, with low contrast text on 83.9% and missing form input labels on 51% (the WebAIM Million accessibility analysis, checked Thursday, October 15, 2026).
Assume your own site has them until somebody has tested it, and test contrast and form labels first.
If the planning above points at a build, websites and e-commerce is the service that covers it, and the page on how an engagement runs sets out the stages and where the investment is agreed.
The FAQ for buyers answers what happens in a first conversation.
Sources behind this guide
The references cited above are the primary sources for this guide: the published accessibility guidelines, the documentation for site owners on helpful content, redirects, canonical addresses and page experience measures, and the WebAIM survey of home pages.
None of those publishers tells you what your site should say. Each one tells you how to check that what it says can be read, by a person and by a machine.
One detail in the performance thresholds decides how they are read. Each of the three is assessed at the 75th percentile of page loads, taken separately for phones and desktops (the definitions behind those measures, checked Thursday, October 15, 2026). A good average therefore settles nothing, because a quarter of your visitors can be waiting while the average looks healthy.
Each claim and its source
The table below pairs each claim the guide makes from outside evidence with that evidence and its publisher.
| Claim | Evidence | Source |
|---|---|---|
| Moved pages need permanent redirects | A permanent redirect makes search results show the new address, while a temporary one keeps the old address showing | Google documentation on redirects |
| A redirect is the strongest way to name the real address | Redirects come first among the canonical signals, and a sitemap entry is a weak hint | Google documentation on canonical addresses |
| Accessibility can be tested before you accept the site | Contrast, text resizing and pointer target size each have a published minimum | WCAG accessibility guidelines |
| Performance is judged as a visitor experiences it | Loading, responsiveness and layout stability each have a published threshold for a good page | Google documentation on page experience measures |
| Pages should be written to help people | Ranking systems are designed to favor helpful, reliable information made for people | Google guidance on helpful content |
| Performance is read at the slow end, not the average | Each threshold is assessed at a high percentile of page loads, taken separately for phones and desktops | the definitions behind those measures |
| Accessibility failures are the normal state of the web | Detected failures appeared on almost every home page tested across the top one million sites | the WebAIM Million accessibility analysis |
Key takeaways
- Decide audiences, page jobs, enquiry routes and surviving addresses before design.
- Give every page one job, one audience and one next step.
- Route every enquiry to one record and one named owner.
- Redirect every old address permanently to its closest equivalent.
- Accept the site against published accessibility and performance thresholds.
Related guides
If your enquiries already arrive but then stall between systems, the systems integration checklist is the more useful guide. The systems decision framework helps if you are not yet sure which problem you have, and the guide to answer engines covers how a finished site gets read by them.
A site planned this way gives hours back to the people answering enquiries, because the details arrive where they are needed instead of being retyped.
Mavrin Labs plans and builds this work for businesses and nonprofits anywhere, working by video and, in Sarasota and Manatee counties, in person, as the page on website design Bradenton businesses can plan face to face explains.
Frequently asked questions
How much of this has to be decided before design starts?
Decide the audiences, the purpose of each page and the route an enquiry takes before design starts. Those three shape the structure, and changing them later means redrawing pages and rewriting copy. Visual choices can wait, because they sit on top of a structure rather than creating one, and they are far easier to change.
What should happen the moment an enquiry arrives?
An enquiry should land in one place where somebody is responsible for it. The enquiry should create a record of who asked and what they wanted, and send an acknowledgement the person recognizes, so decide before launch which system owns that record. If the answer is a shared inbox nobody owns, you have found the gap that loses work.
Do I need a new system to connect my records?
A new system is rarely the answer. The planning job is deciding which of the systems you already run owns each piece of information and how the site passes it across, so nobody retypes it. Replacing a tool your staff know is a much larger project than connecting it, and it should be a separate decision with its own reasons.
What happens to my search results when the site changes?
Search results follow your addresses, so every address that changes needs a permanent redirect to its closest equivalent. Write the list of old addresses before the build, map each one to its new home, and keep the list as a deliverable rather than an afterthought. Pages that quietly disappear take their accumulated standing with them.
Who should own the site after launch?
One named person inside the business should own the site after launch, with a written note of who to contact for what. Ownership means someone decides when content changes, notices when a form stops delivering, and holds the accounts the site depends on, so agree it before the final invoice. A site with no owner drifts within months.
What should a handover include?
A handover should include the accounts and access in your own name and the list of redirects applied. Add the structure of the site, where enquiries are delivered, and plain instructions for the content you will edit yourself. Ask for the handover in writing at the start of the project, so it is part of the scope.
