How to write a website RFP

When an RFP is the right instrument, and when to skip it#

An RFP is worth writing when the decision has to be defensible to somebody who will not be in the room, or when you cannot choose between firms without comparable written answers. It is worth skipping when one person decides, the budget is modest, and two conversations would settle it. The document is a procurement instrument, and running one on a project too small to need it costs more in delay than it returns in diligence.

Most manufacturers who ask us about an RFP are somewhere in the middle: the owner has decided the site needs replacing, a number has been half agreed, and nobody wants to be the person who picked wrong. That is a real problem, and an RFP is one of several answers to it.

A lighter alternative, when the project is small

Write a one page brief instead. It covers what you make, what is wrong with the current site, what the new one has to do, the budget range, and the date you want it live. Send it to three firms and ask each for a scope and a price. You will get most of the comparability of an RFP for about a day of work instead of three weeks, and firms respond faster to it.

Decide the number before you ask for it#

Agree a range internally before writing a word, and publish it in the document. The usual objection is that bidders will quote to the top of the range. The larger and more common problem is receiving a $9,000 proposal and a $70,000 proposal for what you thought was one project, with no way to compare them, because each firm guessed at a different scope.

A budget range does more to define the scope than to set up a negotiation. It tells a bidder which version of the project to propose, and it lets firms who cannot serve you at that number decline early instead of producing a proposal you will not accept.

Two firms in this category publish a strategy engagement priced separately from the build, at figures between about $3,900 and $12,000, and treat it as a precondition. That is a defensible way to work and it is a real cost to plan for, so ask directly whether a paid discovery phase is required before anyone will quote the site.

What actually moves the number

• The number of capability pages that need writing from scratch matters most. This is almost always the largest line, and it is the one most manufacturers assume is free.
• It matters whether usable photography exists. A build with no plant photography either buys stock, which looks like every competitor, or schedules a shoot.
• Integrations change the shape of the work. A CRM, an ERP, a distributor portal, or a part configurator each change the project instead of simply adding to it.
• The number of people who approve matters. Three approvers is a schedule. Nine is a different project at a different price.
• Translation and export control need planning early. Both are cheap to plan for and expensive to retrofit.

Who to send it to, and how many#

Send it to three to five firms. Each serious response costs a bidder several days, so a list of ten wastes nine firms’ weeks and produces a pile you cannot read carefully. Build the list by looking at manufacturing sites those firms actually built rather than at their own marketing, and check those sites against criteria a buyer would use.

The usual way a shortlist gets built is a search for manufacturing website agencies, which returns a listicle written by one of the agencies on it. Those lists are useful for names and worthless for ranking, because the publisher put itself first.

A better way to build the list

1. Collect names from three or four sources, including any roundup articles, your trade association, and the sites of manufacturers you admire. Many industrial sites credit their agency in the footer.
2. Open the manufacturing sites each firm built, and leave the firm’s own site for later. Those client sites are the product you are buying.
3. Score two of those sites yourself on whether they publish tolerances, certifications with scope, an equipment list, and a quote path that accepts a drawing. Most fail, which is informative.
4. Check the work is recent and still live. One site showcased as a leading example in a widely read article now redirects to the company that acquired the client.
5. Include one firm outside your assumed price band, above or below. It calibrates the others and occasionally changes your mind.

Our own study of 66 US manufacturer websites found that 4 cited a materials or process standard and 45 carried a quote path on the homepage. If a portfolio site clears both, the firm that built it has done something most of the market has not. The ten criteria are published in full and take about fifteen minutes to run.

What a manufacturing website RFP has to contain#

A manufacturing website RFP needs ten sections: who you are, what is wrong now, what the new site must do, the capability brief, technical requirements, who writes the content, the budget range, the timeline, the ownership terms, and how you will decide. Together they run to four to eight pages, and anything longer is boilerplate that every bidder skims.

The order matters less than the presence of each section. What follows is what each one is for, written from the position of somebody who has to price the work from it.

The capability brief, which only you can write#

The capability brief states what you make, in what materials, to what tolerances, at what volumes, under which certifications, and who buys it. No agency can invent this section, every project that stalls stalls here, and the firms worth hiring will tell you the same thing. Half a day with the person who prices the work produces it.

This is the part of the document that determines whether you get a website or a brochure. An agency can research your market, study your competitors, and design something handsome. It cannot know that you hold plus or minus 0.0002 inches on a particular family of parts, that you decline anything under 250 pieces, or that your ISO certificate covers two of your three buildings.

What to write down, with an example of each

• Name your processes the way a buyer names them, such as five-axis milling, Swiss turning, wire EDM, centerless grinding, and mechanical zinc plating. Category words such as precision manufacturing describe nothing.
• List materials by grade, such as 6061-T6, 17-4 PH, Inconel 718, and C360 brass. The exact designation tells a buyer far more than the material family does.
• State the tolerances you hold in production, which is a different number from the one you hold on your best day.
• Give the size envelope and part weight, largest and smallest. This is the first filter a buyer applies and the one most often missing.
• State your volumes, meaning the minimum order, the typical run, and the largest annual quantity you have delivered.
• List certifications with their scope, the registrar, the expiry, and which entity or site each certificate covers.
• Name the industries and audit regimes you have survived, since a buyer wants evidence you have been through theirs.
• List your equipment by manufacturer and model, where you are willing to publish it.
• Say what work you decline. Publishing this ends the inquiries you would have refused in week three, and it costs nothing.
• Say who buys, by job title, and what each of them checks before saying yes.

Writing technical requirements a bidder can price#

State the constraints and let bidders propose the solution. Naming a platform in advance usually names the one somebody read about, and it can exclude the right answer. What belongs in the RFP is what must integrate, what must be editable by whom, what accessibility standard applies, and what has to be measurable after launch.

The technical section of most website RFPs is either empty or a list of technologies chosen before the requirements. Both make the responses worse. What a bidder needs is the set of facts that change the shape of the work.

Requirements written so a bidder can price them

The technical items manufacturers most often forget

• Find out who controls the domain and the DNS, the settings that point your domain at the website. It is frequently a person who left in 2019, and it is far easier to sort out before the project starts than on launch day.
• Check whether export control applies. If any drawing or specification is subject to ITAR or EAR, that constrains hosting, file handling, and who may access the staging site.
• Decide where uploaded drawings go, who can read them, and how long they are kept. Buyers ask, and most sites cannot answer.
• Map redirects from the existing site one to one. Missing redirects are the most common cause of traffic loss after a relaunch.
• Keep analytics in an account your company owns, so the history stays with you whichever agency you work with.

Who writes the words, and the hours it takes from your side#

Ask who writes the technical content and how many hours of your people it needs. That one question separates firms that have built a manufacturing site from firms that have not, because only the first kind answers it comfortably. Content is the largest cost variable and the most common reason a project runs late.

A design can be produced without you, but technical content cannot. Somebody has to state the tolerance, and the only people who know it work in your building. A sound proposal therefore contains a number of hours it needs from your side, and a proposal that contains none has either hidden it or has not thought about it.

The three workable models

Whichever model is chosen, write the approval process into the RFP. Name who signs off technical accuracy, name who signs off tone, and give each a working-day turnaround. Most late website projects are late because a review sat with somebody for three weeks. That is a scheduling problem you can prevent in the document.

Ownership, hosting and the exit path#

Put source code ownership, hosting arrangements, domain control, and the exit path in the RFP rather than in the contract. Raised at RFP stage it is a requirement every bidder answers before they have won anything. Raised after selection it is a renegotiation with the firm you already chose, and you will lose it.

This is the section manufacturers most often leave out and most often regret. The failure mode is familiar: the site is fine, the relationship cools, and it turns out the code lives in an account you do not control, on hosting you cannot leave, with a monthly charge that has to keep being paid for the site to stay up. Changing a phone number takes a week and an invoice.

What to require, in these words

• Source code. On final payment, the complete source is delivered to us, and we may host it, modify it, or give it to another supplier without restriction or further fee.
• Hosting. State what you propose, what it costs a year, and whether we may host elsewhere. A build that only runs on your infrastructure must say so in the proposal.
• Domain and DNS. The domain is registered to us, in an account we control, before launch.
• Analytics and Search Console. Both are created under our account, with the agency added as a user.
• Third-party licenses. List every paid plugin, theme, font, or service the site depends on, with its annual cost and whose name the license sits in.
• Exit. Describe what we receive within ten working days of ending the relationship, and what it costs.
• Content rights. Photography, copy, and illustration produced for us are ours, with any stock licenses transferred or identified.

A bidder who cannot agree to these is not necessarily a bad firm. Some build on proprietary platforms and say so, and that is a legitimate trade. What matters is learning it now. The source code question is worth understanding before you ask it.

The template, in full, to copy#

The whole template is below as text you can copy. Replace the bracketed parts, delete what does not apply, and keep it under about eight pages. The capability brief in section four is the part that decides whether you get useful proposals.

Copy this into a document, work through the brackets, and delete any section that does not apply to your project. It is written to be short enough that every bidder reads all of it.

One line in section ten does more work than any other: stating that you are not asking for speculative design. It is the sentence that keeps busy firms on the list.

Sending it, and the timeline to give#

Give at least two weeks to respond, three if the capability brief is long. Name one contact. Set a date for questions and a date on which you will send every question and answer to all bidders at once, which is what keeps the process fair and is the step most manufacturers skip.

The mechanics are simple and the errors are consistent. A deadline of five working days removes every firm with existing commitments, which is every firm worth hiring. Answering one bidder’s question privately gives that bidder an advantage and, more damagingly, means the others price a different project.

A workable six week sequence

1. Week 0: agree the budget range internally. Settle a range with whoever signs before you write anything. An RFP with no number produces responses that cannot be compared, because each firm has quoted a different project.
2. Week 1: write the capability brief. List what you make, in what materials, to what tolerances, at what volumes, under which certifications, and who buys it. This is the section no agency can write for you, and every stalled project stalls here.
3. Week 1: list what is broken about the current site. Write down specifics instead of impressions: the quote form nobody can find, the capability page that is a PDF, the certificate that expired in 2023. Bidders can price a problem, and they cannot price an adjective.
4. Week 2: assemble the document from the template. Work through the sections in order: company and context, the problem, scope, technical requirements, content responsibility, budget range, timeline, ownership terms, how you will decide, and the deadline.
5. Week 2: pick three to five firms. Look at the sites they built for manufacturers, which tell you more than their own site does. Check whether those sites publish tolerances, certifications with scope, and a quote path.
6. Week 3: send it with a questions window. Give a named contact, a date by which questions must arrive, and a date on which you will send every question and answer to all bidders at once. That last step is what keeps it fair.
7. Week 5: score the written responses before meeting anyone. Fill in the scoring sheet from the documents alone. Presentations move scores toward whoever presents best, so record the position you held before that happened.
8. Week 6: meet the two highest scorers. Ask each to walk through one manufacturing site they built and explain a decision they now think was wrong. Confirm the people in the room are the people doing the work.

Tell firms who are not selected, promptly and in one line. It costs nothing, it is rarer than it should be, and the firm you decline this year is sometimes the firm you want in three.

The questions that separate the bidders#

Ask these nine questions, chosen because the answers vary between firms. The most useful is the last: ask what they would change about your brief. A bidder who challenges the requirements has read them, and a bidder who agrees with everything has read the budget.

1. Which manufacturers have you built for in the last two years, and what does each of them make? It tests recency and relevance in one question, so ask for URLs.
2. Who writes the technical content, and how many hours do you need from our staff? It is the single most predictive question on the list.
3. How will a buyer send us a drawing, and what happens to it then? A firm that has built for manufacturers answers with a process. A firm that has not answers with a contact form.
4. What structured data will the site publish, and why? Nine of ten agencies in this category publish almost none on their own sites, so the answer distinguishes quickly.
5. What will you do about our existing URLs? The right answer includes the words redirect map. There is no other right answer.
6. Who will actually do the work, and are they in this meeting? Ask for names and roles rather than a team page.
7. What does this cost us in year two and year three? Ask about hosting, licenses, maintenance, and any retainer, and get an annual figure.
8. Tell us about a project that went wrong and what you changed afterward. Every firm with a history has one. A firm with none is either new or unreflective.
9. What is wrong with this RFP? Ask it directly. The best answers we have seen to this question came with a rewritten requirement attached.

A scoring sheet, filled in before anybody presents#

Score the written responses from the documents alone, before anybody presents. Presentations move scores toward whoever presents best, which is a real skill and rarely the one you are buying. Recording the position you held beforehand is the only defense against it.

Weight the criteria before the responses arrive, so the weights describe what you need rather than what the strongest bidder happens to offer. The table below is a starting point, and the weights are worth arguing about internally before you accept them.

Price carries twelve points here rather than fifty, which surprises some procurement teams. The reasoning is that a website that produces no quote requests is expensive at any price, and the criteria above the price line are the ones that predict whether it will. Adjust the weight if your situation genuinely differs, and write down why.

Two rules that make the sheet work

• Every scorer fills it in alone, first. Discussing before scoring produces one opinion held by four people.
• Anything scored below four needs a written sentence. It forces the reason into the open, and it is what you will read again in a year when somebody asks how the decision was made.

Why good agencies decline to bid#

Agencies decline RFPs for predictable reasons: too many bidders, spec work requested, no budget stated, an unrealistic deadline, requirements that describe a different project, and a process that reads as a formality with an incumbent already chosen. Each one is fixable, and fixing them improves who answers.

It is worth knowing why the firm you wanted did not reply. The reasons are commercial rather than personal, and most of them are visible in the document before it is sent.

The mistakes that show up in most website RFPs#

Seven mistakes account for most of the difference between an RFP that produces comparable proposals and one that produces a pile nobody can read. All seven are in the document rather than in the process, and all seven are cheap to fix before sending.

1. There is no budget. It is the single most common mistake, and the one that makes the responses impossible to compare.
2. Requirements are written as adjectives. Modern, clean, user-friendly, and professional cannot be priced, delivered, or disputed.
3. A platform is chosen before the requirements. It excludes answers you have not heard of yet.
4. There is no capability brief. That guarantees generic copy, because the agency will write what it can research.
5. The document is silent on content hours. The project then stalls in week five and everybody is surprised.
6. Ownership is left to the contract. By then you have chosen, and the only pressure you had has gone.
7. It asks for a page count instead of an outcome. Asking for forty pages produces forty pages. Ask for pages that answer the questions buyers arrive with, and let the count follow.

Questions manufacturers ask about website RFPs#

Here are eight questions manufacturers ask about website RFPs, each answered so it can be read on its own.