Four Questions to Answer Before You Build Anything in Marketing

Many owners already pay for a website, a CRM, an email platform, and an AI subscription or two. What is usually missing is a written agreement about what each one is supposed to accomplish. Without it, a website redesign gets judged on whether people like the colors, and a CRM on whether the sales team logs in.

At Lucé Media, we answer four questions before we build anything for a client. They are useful questions to ask before any marketing project, whoever does the work.

The example

One hypothetical company runs through this article: a commercial HVAC service business in Dallas-Fort Worth with about $20 million in revenue, mostly from repeat property managers and referrals. The owner has a quote for a new website, the sales manager has one for a CRM, and neither project has a written purpose.

What is the one job this has to do?

The one job is the single business outcome a deliverable exists to produce, written as something a customer or your team should do.

A feature list describes a project without saying what it is for. “Mobile friendly, new photos, a blog, a chat widget” could describe almost any website. When every feature is a priority, scope grows until the budget runs out.

A job statement names who should do what, in one sentence, and it settles arguments when two reasonable ideas compete for the same budget. If the sentence needs several “ands,” you probably have more than one project. A useful test: if this delivered only one result, which result would justify the cost?

In the example: The owner started with “a modern website that shows who we are.” The agreed job became “Get facility and property managers at commercial buildings in our service area to request a maintenance contract quote.” With that agreed, the quote request moves near the top of every commercial service page and the chat widget waits.

How will we know it worked?

Pick one outcome measure tied to the job, define it precisely, record a baseline using the same definition, and name the system and the person that report it.

Measure the outcome, not the activity

Visits, email opens, and CRM logins show activity. They can help explain a result, but the business needs qualified requests, booked meetings, or signed contracts.

Define it once and use it for both numbers

A before-and-after comparison only works if both numbers count the same thing. If old requests sit in an email inbox and new ones will sit in a CRM, someone has to review the old ones against the same rules, or the new system will look better or worse simply because it counts differently.

Name the source and the reader

Decide which system holds the official count and who reports it, so the review is about the result rather than whose number is right.

In the example (figures are hypothetical): The company defines a qualified commercial quote request as one from a commercial property inside its service area, asking for maintenance contract pricing, with working contact details. The sales manager reviews six months of inbox requests against that definition and finds about six qualified requests a month. The target is 12 a month by the end of the second full quarter after launch, with each new CRM entry marked qualified or not under the same rules.

What is version one allowed to get wrong?

Version one may get reversible choices wrong, such as wording, layout, and the order of content. It may not get facts, privacy, accessibility, payments, or customer trust wrong.

Headline wording, page order, image choices, and form length are reasonable things to test, because changing them costs little and each change tells you something about your buyers.

Some mistakes are not acceptable in any version: wrong prices, licenses, or service claims; collecting personal information without proper notice and safeguards; pages that people using screen readers or keyboards cannot use, judged against the W3C’s Web Content Accessibility Guidelines; and broken payments or checkout.

Check those before launch. Submit every form, confirm notifications arrive and each request reaches the right person, confirm tracking records the conversion, and run a basic accessibility review. Customers should not be the first people to test the quote form.

Version one also does not have to be public. It can be an internal pilot, a small test with trusted customers, or a CRM workflow run alongside the old process. Our article on starting your first AI agent on one boring process applies the same idea to AI.

In the example: Version one tests two headlines and a short quote form against a longer one. It may not show an unverified service area. The CRM starts with commercial requests only.

When will we look at it again?

Put the review dates on the calendar before launch, set their timing by how quickly the work can produce evidence, and name the one person who makes the decision.

Review too early and you react to noise; wait too long and a weak project keeps spending. A B2B form that gets a handful of requests a month needs more time than a high-traffic page, and if contracts take 60 days to close, an early review can judge requests but not revenue.

Most projects need more than one date. A follow-up check shortly after launch catches problems that only appear in real use. An interim review looks at early performance and adjusts. The final assessment comes at the target date, when the owner decides to continue, revise, expand, or stop. Writing those options down in advance makes stopping a planned outcome.

In the example: Two weeks after launch, the sales manager confirms real requests are arriving and routed correctly. After the first full quarter, an interim review compares qualified requests with the baseline and fixes what is underperforming, but nothing expands yet. At the end of the second full quarter, the owner makes the final call. The CRM expands to residential work only if the target was met, routing stayed accurate, the records are clean, and the sales team kept up with the extra requests.

The before-we-build checklist

Bring this to the planning meeting. A blank row means the project is not ready.

How the four questions guide full-service work

In full-service work, the four answers decide which channels get built, in what order, and how each one is judged.

A full-service engagement can include the website, SEO, content, email, CRM, automation, paid media, analytics, and AI. Most companies do not need all of it at once. They need the piece that does the current job, with the next piece waiting until the evidence supports it.

That is how our digital strategy development work is structured, and how engagements with our Dallas-Fort Worth digital marketing agency team begin. If the problem may sit in the decisions behind your marketing rather than the tools, see our guide on why marketing isn’t working.