SailgrowStart a project ↗

Most marketing dashboards are rubbish.

Not because they’re ugly or hard to read. Because they show numbers nobody uses to make decisions.

ROAS from last week. Impressions by platform. CTR by ad. CPC trends.

All numbers. All real. None of them tell you what to do on Monday morning.

If you’re paying an agency or a team to run marketing, your dashboard should answer the questions you’d actually ask. Not the questions a tool happens to track.

The 4 questions a dashboard should actually answer

If I’m a business owner opening a dashboard with 5 minutes to spare, I want to know:

1. Did we make more money than we spent?

Everything else is secondary. If you spent 1 lakh on ads and brought in 3 lakh in sales, you’re winning. If you spent 1 lakh and brought in 80k, you’re not.

This should be the biggest number on the page. Spend vs revenue. Last month vs this month. Done.

2. What’s working, what’s not?

Which campaign brought in the most sales. Which blog post brought in the most leads. Which channel is pulling its weight.

Not “every campaign ranked by click volume.” That tells you nothing. “Top 3 and bottom 3, with what we think is causing it” is what you want.

3. Is anything about to break?

Ad costs rising. Conversion rate dropping. Email open rates falling. Traffic from a channel declining.

If something’s about to go wrong, you want to know this week, not next month.

4. What are we doing about it?

Every number should connect to an action. If ROAS dropped, what are we testing to fix it. If a landing page is underperforming, what’s the plan.

Numbers without a decision beside them are decoration.

What usually gets in instead

Most dashboards get built around what the tools can report, not what you need to know. So you end up with:

You spend 10 minutes reading it. You still don’t know if things are good or bad. You close it and ask the team for a verbal update.

That’s a sign the dashboard isn’t doing its job.

How to build one from scratch

If you want a dashboard that’s actually useful, here’s the quickest way to build one:

Step 1. Write down the 5 decisions you make every month.

“Should we spend more on Meta ads next month?” “Should we kill this blog category?” “Is our email list growing?”

If the decision doesn’t appear in your month, don’t track for it.

Step 2. For each decision, write down the 1 or 2 numbers that would answer it.

ROAS by platform. New subscribers per week. Sales from top 3 campaigns.

Step 3. Build only those numbers into the dashboard.

Not every metric your ad platform can show you. Just the ones tied to a real decision.

Step 4. Add one sentence of context next to each number.

“ROAS 2.4x this month, down from 3.1x last month. Testing new creative to fix it.”

Numbers alone are noise. Numbers with context are a dashboard.

Rebuild from scratch

If you’ve got a dashboard right now and nobody looks at it, that’s your sign.

Open it. Delete every chart that doesn’t tie to a decision you make. Add context lines next to the numbers you keep.

You’ll probably end up with 5 numbers and a short note next to each one. That’s fine. That’s what a good dashboard looks like.

Most of the data people stare at in agency reports exists because the tool makes it easy to show, not because anyone needs it.

Your job as a business owner is to filter all that down to the stuff you’ll actually act on.

Here’s what happens on most website projects.

Designer builds the site. Developer codes it. SEO person gets called in two weeks before launch.

Then the SEO person opens a 30-item fix list. URL structures that need rewriting. Images with no alt text. Headings in the wrong order. Pages that don’t exist for the keywords that matter.

Half the list won’t get done because the launch date can’t move. The other half gets patched over the next 6 months.

This is the normal way to build a website. It’s also the most expensive way.

What goes wrong when SEO comes last

The problem isn’t that SEO people are slow. It’s that most SEO fixes aren’t actually SEO problems. They’re design decisions that nobody made with SEO in mind.

A few examples we see every week:

None of these are “SEO work.” They’re design and build choices. But they cost you traffic for years if you get them wrong.

7 decisions that are design decisions AND SEO decisions

If you’re planning a new site, these choices need SEO input before they get locked in:

  1. URL structure. Flat or nested? Keyword-based or ID-based? Change it later and you’re redirecting hundreds of pages.
  2. Page templates. How do service pages, blog posts, category pages differ? Each template needs its own meta logic, heading hierarchy, and schema markup.
  3. Navigation. Which pages go in the top menu? Which get surfaced from the footer? This decides where your link equity flows.
  4. Image strategy. Compressed webp or raw jpg? Lazy-loaded or not? CDN or self-hosted?
  5. Page speed targets. Carousel in the hero? Video background? Fancy animations? All of these have a page speed cost.
  6. Content modeling in the CMS. Can editors add alt text? Meta descriptions? Schema? If not, every page launches with gaps.
  7. Internal linking patterns. Related posts, breadcrumbs, in-content links. Design decisions, all of them.

How to actually bring SEO into design

You don’t need a full-time SEO person on the project. You need SEO input at three specific moments:

Before wireframes. Review the sitemap, URL structure, template plan.

During design. Review heading hierarchy, meta title and description logic, image strategy, mobile layout.

Before launch. Run a technical audit, check indexation, redirects, schema, speed.

That’s it. 3 touchpoints, usually a day or two of work each. Spread across a 2-month project, it’s barely visible to the design and dev team.

Compared to retrofitting after launch? Easily 10x cheaper.

The real win

The biggest cost isn’t the fix list after launch. It’s the traffic you don’t get for the first year because the site wasn’t built to rank.

A site that launches with good URL structure, clean templates, fast pages, and clear navigation starts earning organic traffic from month one.

A site that launches without those things earns traffic maybe from month six, after someone has gone back and fixed everything.

That’s a 5-month gap. For most ecom businesses, that’s real money.

Technical SEO isn’t something you bolt on. It’s a set of choices you make while you build. Make them early and you save yourself the cleanup.

Google isn’t the only place people search anymore.

ChatGPT answers direct questions. Perplexity cites sources. Gemini summarizes articles. Claude picks the ones that have clean, direct answers in them.

So the question people are asking us: do we need two different content strategies now?

Short answer, no.

Long answer, your content needs a few small tweaks so it works in both places. Not a rewrite. Just a shift in how you structure the answer.

The real answer first

Most of what makes a page rank on Google also makes it quotable in AI answers. If you got the SEO basics right, you’re already most of the way there.

What’s actually new is this: AI assistants don’t read your whole page. They scan for the one paragraph that answers the user’s question clearly, and they pull it. Sometimes they credit you. Sometimes they don’t.

If your answer is buried in a 400-word intro, they skip you.

If your answer is in the first paragraph after the H2, you get picked.

That’s basically the whole game.

What’s the same

Both Google and AI search care about:

If any of this is weak, fix it first. AI search optimization doesn’t help a page that Google already struggles with.

What’s different

A few things change when you optimize for AI search specifically.

Short, direct answers near the top. Not after a 3-paragraph warm-up. The AI picks the first clean answer it finds.

Question-style H2s. “What is the best CRM for small business?” works better than “Choosing a CRM.” AIs match headings to user queries.

Numbers and specifics. AI assistants love concrete answers. “5 ways to…” or “costs between $20 and $50” gets picked over “there are many ways” or “pricing varies.”

Lists and tables. They’re easy for an AI to parse and quote.

Clear attribution. Author name, date, credentials. The AI uses these to decide if your page is trustworthy enough to cite.

The 5 things to actually do

If you’re writing a new blog post this week, do these 5 things:

  1. Put a direct 2-sentence answer right under your H1 or first H2. Treat it like the TL;DR. Don’t be clever, just answer.
  2. Make your H2s match real questions. Pull them from Google’s “People Also Ask” or AnswerThePublic. Use them as section headings exactly.
  3. Add one list or table per 500 words. Lists get quoted. Walls of text get skipped.
  4. Name a few specific things. Tool names, prices, timeframes, example brands. Vague content doesn’t get cited.
  5. Add an author line and a date. AI assistants weigh these.

That’s it. These changes take about 20 extra minutes per post, and they help on Google too.

One quick test

After you publish, ask ChatGPT or Perplexity a question your post answers. See if it quotes you.

If it quotes someone else, read their page. Check which of the 5 things above they did and you didn’t.

Then fix it.

Where this is heading

AI search isn’t replacing Google yet. But it’s eating into questions-based search. “What’s the best SMS provider for Shopify?” used to send someone to Google, then to a blog. Now people just ask ChatGPT and read the answer in one screen.

If your blog isn’t the one being quoted, you’re not in the conversation at all.

Still worth running SEO. Still worth writing long guides. Just structure them so an AI can lift the useful parts without rewriting them.