Balian.dev
St. Paul, MN · remote Book a call

Process

A process with your sign-off at every gate.

A new site follows thirteen steps and nothing moves forward without your okay at six of them. Other jobs run shorter, and design still comes before code.

A new website gets the full sequence below. Other jobs run shorter. Pick the one that sounds like yours.

The stages

What happens, and what you get.

The full sequence, with 6 approval gates. Nothing is built before you have seen it designed, and nothing goes live before you have signed it off.

  1. Understand

    01Discovery

    We talk about the business, the problem and the constraints, and I record the call, because that conversation plus whatever you already have is the raw material for the content. I review the current site if there is one, list what I find, and ask the questions only you can answer. One person on your side is named as the approver, so feedback does not arrive from five directions.

    You get: A fact sheet, a short list of questions, and the interview recording.

  2. 02Research

    I look at what success means for this project, who else is in the space, and what people actually search for. This keeps the later decisions grounded in evidence instead of taste.

    You get: A short research summary.

  3. Strategy

    03StrategyYour approval

    The plan for what the site says, to whom, and how pages connect. It covers the sitemap, the keywords each page targets, and what content needs to exist before design starts.

    You get: A content strategy and sitemap for you to approve.

  4. 04ProposalYour approval

    A written scope with a timeline, a price and payment milestones tied to the approval gates. Nothing is built until you agree to it. Anything that is not in the scope is a separate quote, and I say that up front so it never comes as a surprise.

    You get: A proposal for you to approve.

  5. Design

    05Styleframes

    I design three visual directions using your real content. They differ in mood, type and layout, so you are choosing between real options instead of three tints of one idea.

    You get: Three styleframes to react to.

  6. 06Direction approvalYour approval

    You pick a direction, or combine pieces of several, and tell me what is wrong with it. That feedback is applied to the design system, so the fix shows up everywhere it applies.

    You get: A chosen direction with your notes applied.

  7. 07Design system and blocksYour approval

    The chosen direction becomes a design system: type, colour, spacing and the blocks every page is assembled from. I do not mock up every page. You approve the system and the block designs, and that approval locks the look of the site.

    You get: A design system and block designs, approved.

  8. Build

    08Content and build plan

    I write the content from the interview and the materials you already have, in the voice we agree on, fitted to the blocks. You will never be handed a blank page. Each page gets a brief that says which blocks it uses and what goes in them, and the build is broken into pieces that can be checked one at a time.

    You get: A build plan and the first draft of every page's content.

  9. 09First draft

    AI executes the build plan and I read every line. You get a first draft of the whole site on a staging link, with your real content in it, usually sooner than you expect. Tests are written alongside the code.

    You get: A complete first draft on staging.

  10. 10Two revision roundsYour approval

    You go through the staging site and your approver sends me one consolidated list. I fix it and send it back. That happens twice. A fix is anything that does not match what we approved, and correcting a fact or a name counts as a fix. A new idea is welcome, and it goes on the Phase 2 list to be quoted after launch. After round two the design is set and you sign off.

    You get: A finished site, your sign-off, and a Phase 2 list.

  11. Launch

    11Quality checks

    Accessibility against WCAG AA, performance against Core Web Vitals, security headers and configuration, structured data, redirects, and testing across browsers and devices.

    You get: A QA report.

  12. 12LaunchYour approval

    A pre-launch checklist, the release itself, and monitoring while the site settles. You approve before it goes live.

    You get: A live site and its documentation.

  13. 13Support and Phase 2

    Two weeks of fixes and questions after launch are included. Once the launch invoice is paid, I quote the Phase 2 list and we do the items worth doing. Ongoing care is optional and agreed separately.

    You get: A support period, and a quote for Phase 2.

Design before build

Why the design stages come first.

Three real choices

The styleframes use your actual content and differ in mood, type and layout. You react to something concrete instead of a mood board.

A system and its blocks

I design the system and the blocks pages are built from, then let the build produce the pages. Mocking up every page is where projects drown in preference.

Cheap to change, then expensive

Changing a design is quick. Changing a built site is not. The gates put the decisions where changes cost least.

You will never be handed a blank page.

Most websites stall on the words. Copywriting is expensive, so the job goes to whoever on your team seems to have time. That person turns out not to have the time, has never written a website, was not in the room when the goals were set, and does not know the block system the pages are built from. Three weeks later I get a Word document and a question about how to center a section.

So I write it. I ran discovery, I know what the site is for, and AI drafts faster than anyone types. You give me an hour on a recorded call and whatever you already have: brochures, decks, the old site, sales emails, reviews. I turn that into every page, fitted to the blocks, and you correct facts on the staging site. If you have a writer you trust, they work inside the page briefs and I edit for fit.

Questions
How long does a project take?

It depends on scope, and the proposal states a timeline you can approve. The gates are there so that timeline is based on decisions you have already made.

What if I do not like any of the three directions?

Say so and say why. The feedback shapes the next round, and you are not charged for being specific about what is wrong. Often the answer is a mix of parts from two directions.

How many revision rounds do I get?

Two, on the staging site, each one a single consolidated list from your approver. The rounds are for fixing anything that does not match the design you approved. New requests are welcome and go on the Phase 2 list, to be quoted after launch.

What is the difference between a fix and a change?

A fix is something that does not match what you approved, like a wrong colour or a block that does not behave as designed. A change is a new preference or a new feature. Fixes are included. Changes are quoted, and most of them are better done after launch anyway, when you have seen the site in use.

What if we think of something halfway through?

Tell me. It goes on the Phase 2 list with a rough estimate, the build keeps moving, and once the site is live and paid for I quote the list. Scope creep is how projects go over, so I treat it as a separate phase instead of pretending it is free.

Do we have to write the content?

No. I write it from a recorded interview and the materials you already have, and you correct facts on the staging site. If you have a writer you trust, they work inside the page briefs and I edit for fit.

Can we skip stages?

Some projects compress them, but design before build and your approval at the gates stay. They are what keep the cost of changes low.

Does this work for a web app as well as a website?

Yes. Content and copy stages give way to product scoping and iterative delivery, and the design gates stay the same.

A new feature on a site that already exists.

A booking form, a member area, a product configurator, a new section. The site is live and people are using it, so the job is to add the thing without breaking what works.

PaceSmall features ship in days. Larger ones get a staging link within the first week or two.

  1. 01

    Read the codebase and the request

    I read the code the feature has to live in, note what it touches and what could break, and ask the questions only you can answer.

    You get: A short note on what the feature touches and any risks.

  2. 02

    Scope and quoteYour approval

    A fixed price for a written scope, with what is out of scope named so it never surprises anyone.

    You get: A proposal to approve.

  3. 03

    Design, if it has a faceYour approval

    Anything a visitor sees is designed inside your existing design system first, so it looks like it was always there.

    You get: A design to approve.

  4. 04

    Build on staging, with tests

    AI executes the plan, I read every line, and tests cover the behaviour that matters. You get a staging link, never a surprise on the live site.

    You get: The feature on staging.

  5. 05

    One review roundYour approval

    Your approver sends one consolidated list. Anything that does not match the scope or the design is fixed. New ideas are quoted separately.

    You get: Your sign-off.

  6. 06

    Release and watch

    Released at a quiet hour, behind a switch where that makes sense, with monitoring on for the first days.

    You get: The feature live, and documentation for whoever maintains it.

  7. 07

    Two weeks of fixes

    Anything that turns up in the first two weeks is fixed as part of the job.

    You get: A support period.

Questions
Can you work on a site someone else built?

Yes, and most of this work is. I read the code first and tell you what I found before quoting, so you know whether the feature is a day or a month before anyone commits.

What if the codebase is a mess?

I say so, with specifics, and give you two prices: the feature on top of what is there, and the cleanup that would make the next feature cheaper. You choose.

Do I get the same gates as a new site?

The ones that matter: you approve the scope, the design if it has one, and the staging build before release. The rest compress because the site already exists.

The insurer sent a questionnaire, or a list of findings.

MFA, patching, backups, headers, a WAF, logging, and proof of all of it. The underwriter wants evidence, and the renewal has a date on it.

PaceMost portfolios are remediated and evidenced within a few weeks. Deadlines are respected.

  1. 01

    Translate the questionnaire

    I turn the insurer's questions into a list of actual controls on your actual systems, and mark which ones already pass.

    You get: A plain-English control list with pass, fail and unknown.

  2. 02

    Inventory

    Hosting, DNS, WAF, who has access to what, how patches happen, whether backups exist and whether anyone has ever restored one.

    You get: An inventory of what the insurer is asking about.

  3. 03

    Plan by what underwriters checkYour approval

    A ranked plan with the deadline in view. I have run this across a whole client portfolio and I know which items carry weight.

    You get: A remediation plan and price to approve.

  4. 04

    Fix

    MFA everywhere, patching, backups tested by restoring them, security headers, WAF rules, access reviews, logging. On live systems, with your approval at each step that could affect users.

    You get: The controls in place.

  5. 05

    Evidence pack

    Screenshots, configuration exports and short policy statements, written for the person at the insurer who reads them.

    You get: A pack you can attach to the questionnaire.

  6. 06

    Keep it true

    Monitoring and a scheduled check, so next year's questionnaire is a formality instead of a scramble.

    You get: A care plan, if you want one.

Questions
Can you fill in the questionnaire for us?

I draft the technical answers and the evidence behind each one. You sign it, because the insurer is asking you, and you should understand every answer before you do.

We answered yes to things that are not true. Now what?

We make them true before renewal, in order of what the underwriter would check first. That is most of this work, and it is better done now than during a claim.

Will this disrupt the business?

MFA rollouts and access reviews touch people, so those are scheduled with you. The rest is infrastructure and happens without anyone noticing.

An ADA demand letter, a WCAG audit, or a customer who could not use the site.

The letter names pages and failures. The real list is usually longer, and the fix has to hold up to a lawyer as well as a screen reader.

PaceThe highest-risk fixes go live within days of the plan. Full remediation of a large site runs a few weeks.

  1. 01

    Read the complaint and confirm the standard

    Which pages, which failures, which version of WCAG. Most letters cite 2.1 or 2.2 AA. I scope the templates behind the pages, because one template fix covers many pages.

    You get: A scoped list of what the letter actually requires.

  2. 02

    My own audit

    Automated checks, then manual testing with a keyboard and a screen reader. Complaint lists are rarely complete, and the next letter will find what this one missed.

    You get: A full findings report, ranked by risk.

  3. 03

    PlanYour approval

    Fixes ordered by legal risk and by how many pages each one clears. Anything that changes how the site looks is called out before it happens.

    You get: A remediation plan and price to approve.

  4. 04

    Fix on staging

    Structure and headings, contrast, keyboard access, focus, forms and errors, images and media, PDFs and documents. Checked page by page against the standard.

    You get: The fixes on staging to review.

  5. 05

    Verify

    Re-tested with assistive technology and re-run against the standard, with the results recorded.

    You get: Before and after evidence.

  6. 06

    Statement and evidence for your lawyer

    An accessibility statement for the site, a record of what changed, and a dated note of anything that remains and when it will be done.

    You get: What your lawyer asked for.

  7. 07

    Keep it that way

    Accessibility checks in the deploy process, and a care plan if the site changes often.

    You get: A way to stay compliant.

Questions
Can you make the site fully compliant?

Nobody can certify a site as compliant, and anyone who promises it is selling something. What I deliver is work done against WCAG AA, evidence of it, and a statement your lawyer can use. That is what courts and plaintiffs' firms actually look at.

Will an overlay widget fix this?

No. Overlays are named in a growing number of these lawsuits as the problem, and they do nothing for the structure of the page. I remove them.

How fast can the first fixes go live?

Within days of the plan. Letters have deadlines, so the highest-risk items are fixed first and documented as they go.

The site is defaced, redirecting, sending spam, or someone is in your systems.

This one does not wait for a proposal. Containment comes first, the paperwork comes after, and I have done this more times than I would like.

PaceContainment the same day. Cleanup and hardening over the following days. The report when it is over.

  1. 01

    Contain

    Take the site offline or behind Cloudflare, rotate every credential, preserve the logs and a copy of the compromised system for forensics before anything is cleaned.

    You get: The bleeding stopped, with evidence preserved.

  2. 02

    Find out what happened

    The entry point, what was touched, how long they were in, and whether any data left. This decides what you are obliged to tell people.

    You get: A plain-English account of the incident so far.

  3. 03

    Clean and rebuild from known good

    A fresh install, verified files, the backdoors closed, the malware gone. A compromised site is rebuilt, never patched over.

    You get: A clean site, ready to go back.

  4. 04

    Harden

    Patching, MFA, WAF rules, least-privilege access, security headers, monitoring. The hole that let them in, and the next three.

    You get: A site that is harder to get into than the one they found.

  5. 05

    Tell the people who need to knowYour approval

    Your customers, your insurer, your lawyer, and anyone the law says you must notify. I help you write it so it is honest and calm.

    You get: A notification draft you approve before it goes out.

  6. 06

    Return to service

    Back online with monitoring and alerts, and a watch on it for the first weeks.

    You get: A live site with someone watching.

  7. 07

    Post-incident report

    What happened, what was done, and the list of changes that keep it from happening twice. Written for you and for the insurer.

    You get: The report, and a plan for what comes next.

Questions
What should we do right now, before you answer?

Change the passwords you can reach, especially hosting and email. Do not delete anything, do not restore a backup yet, and write down the time you noticed. Then call.

Do we have to tell our customers?

It depends on what was taken and where your customers live. Finding that out is step two, and I help you write the notice if one is required. Saying nothing when the law requires notice is the expensive option.

Can we just restore a backup?

Only once we know when the break-in happened. A backup from after it just restores the attacker. Containment and forensics come first, then a clean rebuild.

A store or site that has to move without losing orders, customers or rankings.

Off a proprietary cart, off BigCommerce, between WooCommerce and Shopify, or off a platform that is about to be switched off. The data comes with you and the URLs keep working.

PaceWeeks for most stores. Catalog size and the number of integrations set the pace, and the cutover is one quiet hour.

  1. 01

    Inventory the risk

    Catalog, customers, order history, URLs, integrations, search rankings. What has to come across exactly, and what the old platform is hiding.

    You get: A migration inventory with the risks named.

  2. 02

    Plan and quoteYour approval

    The target platform, the data mapping, the integrations to rebuild, the redirect approach and a cutover plan.

    You get: A proposal to approve.

  3. 03

    Test migration on staging

    The data moved by script, reconciled against the source, and the mismatches fixed in the script so the real run is a repeat of a rehearsal.

    You get: Your store on staging with your real data.

  4. 04

    Rebuild the integrations

    Payments, tax, shipping, accounting, ERP and EDI connected to the new platform and tested against real transactions.

    You get: Integrations that work before launch.

  5. 05

    Redirects and search

    Every product and category URL mapped, metadata carried over, redirects tested, so search engines follow you to the new address.

    You get: A redirect map and a search checklist.

  6. 06

    Your reviewYour approval

    You and your team work the staging store like customers and like staff. One consolidated list, fixed and re-checked.

    You get: Your sign-off for cutover.

  7. 07

    Cutover and watch

    Final data sync, DNS switch at a quiet hour, the first orders verified by hand, and search traffic watched for weeks after.

    You get: A live store and its documentation.

Questions
Will we lose our search rankings?

Rankings move when URLs change and nobody maps them. Every product and category URL is redirected and the metadata carried over, and I watch Search Console for weeks after. A dip in the first days is normal; a cliff is a migration done wrong.

Does the old store stay up during the move?

Yes. The new store is built and tested on staging while the old one keeps selling. The cutover is one quiet hour with a final data sync.

What about orders placed during the cutover?

The final sync happens after the old store is put into maintenance, so nothing falls between the two. The first orders on the new store are checked by hand.

Two systems that should talk, and a person retyping between them.

Orders into accounting, customers into the CRM, inventory out of the ERP, EDI to a distributor. The integration has to be right, and it has to tell you when it is not.

PaceSimple syncs in days. ERP and EDI work in weeks, most of it testing against real data.

  1. 01

    Map the data and who owns the truth

    Which system is the source for each field, what each side calls things, and where the records disagree today.

    You get: A data map.

  2. 02

    Decide the rulesYour approval

    What syncs, in which direction, how often, and what happens when the two sides conflict. Decided with you, written down.

    You get: Sync rules and a price to approve.

  3. 03

    Build with validation and a review queue

    Every record validated before it moves, every move logged, retries on failure, and ambiguous records held for a person instead of guessed.

    You get: The integration on staging.

  4. 04

    Test against real data

    Run against a copy of your real records and reconciled line by line until the numbers match.

    You get: A reconciliation report.

  5. 05

    Go live and watch

    Switched on with alerts, watched closely for the first week.

    You get: A working integration.

  6. 06

    Handover

    Documentation, the alerts and what they mean, and what to do when a record is held for review.

    You get: Something your team can run.

Questions
Should we buy a connector instead?

Sometimes. If a well-kept connector exists for your exact pair of systems and your rules are standard, buy it and I will configure it. Custom work is for when the rules are yours, the systems are older, or the connector costs more per year than building.

What happens when it fails?

It tells you. Failed records are retried, ambiguous ones are held for a person, and an alert says what is waiting. Silent failure is the one outcome I design against.

Can it start small?

Yes, and it should. One direction, one record type, watched for a week, then the next. Big-bang integrations are how data gets corrupted on both sides at once.

A technical audit, a vendor review, or a straight answer about what you are paying for.

Someone quoted you a number, a platform is up for renewal, or the site has a smell you cannot name. You want an engineer with no stake in the answer.

PaceA focused audit takes a week or two. The findings are the deliverable, whether or not I do the work after.

  1. 01

    Scoping call

    What decision you are trying to make, what you suspect, and what you need to be able to say to your board, partner or vendor afterward.

    You get: A scope and a fixed price.

  2. 02

    Read everything

    Code, hosting, DNS, the vendor contracts, the invoices, the analytics, and the process behind them.

    You get: Nothing yet. This is the quiet part.

  3. 03

    Talk to the people who do the work

    Short interviews with the people who use the systems every day. They know where the time goes.

    You get: Their side of the story, in the findings.

  4. 04

    Findings, ranked by risk and money

    Written in plain English, with what each item costs you now and what fixing it would cost. No jargon and no padding.

    You get: The findings report. This is the deliverable.

  5. 05

    A plan with optionsYour approval

    Keep, fix or replace, with numbers against each, and my recommendation stated plainly.

    You get: A plan you can act on with anyone.

  6. 06

    Then your call

    Hand it to your team or your vendor, or have me run it. The report is written so either works.

    You get: A quote for the work, if you want one.

Questions
Will you just recommend yourself for the work?

The findings are priced on their own and written so your team or any vendor can act on them. If I think I am the right person for what follows I will say so, with a price, and you can take it elsewhere.

Can you review a vendor's proposal?

Yes. I read it the way the vendor hopes you will not: what is actually included, what the hourly rate works out to, what happens at renewal, and what you own at the end.

How honest is honest?

If your site is fine, the report says so and the engagement is short. If the problem is a decision you made, the report says that too, politely.

A site that is live and needs someone to own it.

Updates, backups, monitoring, the occasional fix, and a person to call when something is wrong. The agency is gone or the developer moved on.

PaceOnboarding takes a week. After that it runs every month without you thinking about it.

  1. 01

    Take stock

    Access to everything, a read of the code and hosting, and a list of what is out of date, unmonitored or one incident away from a bad week.

    You get: A health report.

  2. 02

    Fix the urgent itemsYour approval

    Patches, backups that have been tested, monitoring switched on, the obvious holes closed.

    You get: A stable baseline and a quote for the plan.

  3. 03

    Monthly care

    Updates applied on staging first, backups verified, uptime and security monitored, performance checked, and a short report of what was done.

    You get: A monthly report.

  4. 04

    A person to call

    Small fixes and questions handled inside the plan. Larger work quoted separately, so the plan stays predictable.

    You get: Someone who knows your site.

  5. 05

    A yearly review

    What changed, what is aging, what the next year should include.

    You get: A plan for the year.

Questions
Can you take over a site you did not build?

That is most of them. The health report in the first week tells you what you inherited, and anything urgent is fixed before the monthly plan starts.

What does the plan include?

Updates, backups, monitoring, performance checks, small fixes and questions, and a monthly report. Larger work is quoted separately so the plan stays the same price every month.

What if I want to leave?

You own everything, you have the documentation, and the plan ends at the end of the month. No lock-in, because lock-in is a reason to stay for the wrong reason.

Ready to start with a call?