Where should your website improvement budget go first?

Once you have more than one page worth improving, the hard part stops being "is this page worth it?" and becomes "which page first, and how much?" Most teams answer by instinct: the page the founder cares about, the one with the biggest keyword, or an even split so nobody's page is left out.

Each of those spends money where it buys less than it could. An even split gives the same dollars to a page that pays back in five months and one that takes fifteen. The founder's favourite page might need better content, not links.

This guide shows how to rank pages against each other, which caps to set before you spend anything, how to pace spending so a page's links arrive at a believable rate, and how to think about the money you choose not to spend. The worked example takes a $10,000 monthly budget across four pages.

Put every option on the same scale

To compare pages, you need one number that means the same thing for each. Use risk-adjusted revenue per dollar: the extra monthly revenue you expect from the work, discounted by your confidence that it works, divided by what the work costs.

For a link purchase, the inputs come from the payback calculation in the guide to backlink purchases that can't pay for themselves: extra clicks at the next realistic position, times value per click, times confidence (50% is a sensible default). The cost comes from the page's gap: how many linking domains it needs, at what price. The guide to calculating a page-level backlink gap covers that.

Revenue per dollar is payback turned upside down. A page that pays back in 5 months returns $200 a month for every $1,000 spent; one that pays back in 15 months returns about $67. Ranking by either gives the same order. Revenue per dollar is easier to read when you're comparing several options, so that's what the example uses.

Check content first

Before a page competes for link money, ask how much of its upside would come from fixing the page itself: a clearer title, a direct answer to the search, a comparison table, internal links from pages that already rank. If content fixes account for a large share of what the page could gain, do those first. They're cheaper, they don't need pacing, and they make any later links work harder.

A page like that isn't out of the budget. It's in it for content work, not links.

Set hard caps before you rank

Caps are limits the ranking can't override. Decide them once, in advance, so a persuasive opportunity can't talk you past them:

• Monthly budget. A ceiling, not a target. Unspent money doesn't carry into next month, because a larger pot next month invites exactly the rushed buying the cap exists to prevent.
• Maximum price per link. Stops one expensive domain from swallowing a month.
• Payback threshold. Under 12 months can be bought within budget. Between 12 and 18 months, a person decides. Over 18 months, don't.
• Natural link velocity. Cap new linking domains across the whole site at a multiple of how many you've gained on average each month over the last year. A site that normally gains 8 new linking domains a month and suddenly gains 40 looks unlike anything that happens without buying. A cap of 1.5 times your own average (with a small minimum for new sites) keeps growth plausible.
• Per-page pacing. Close any one page's gap over at least three months.
• No domain twice. Buying from the same site again, even for a different page, adds less than a new site would and puts more of your spending on one seller.
• Brand-led anchors. Most links should use your brand name or a plain phrase, with exact-match anchors (the precise search you want to rank for) kept rare: well under 5%.

A worked example

An HR software company has $10,000 a month for SEO, covering content work and links. Its maximum price per link is $900. Over the last year it gained an average of 8 new linking domains a month, so its velocity cap is 1.5 × 8 = 12 new linking domains a month.

Four pages are in the running. All figures are illustrative.

The working for each row is revenue ÷ cost. D: $700 ÷ $1,200 = $0.583 per dollar, or $583 per $1,000, and $1,200 ÷ $700 = 1.7 months. A: $1,080 ÷ $5,400 = $200 per $1,000, and 5.0 months. B: $1,950 ÷ $11,700 = $167 per $1,000, and 6.0 months. C: $1,400 ÷ $21,000 = $67 per $1,000, and 15.0 months.

Page D's modelled upside is $1,000 a month, of which $700 comes from content fixes. With 70% of the gain available without links, it goes content first.

Page B brings in the most revenue in total, but A returns more per dollar, so A is funded first. Page C is the biggest search and the page the sales team most wants to rank, and it comes last: at 15 months it sits in the band where a person decides and nothing is bought automatically.

Now apply the caps to month one:

Month one spends $6,900 of $10,000. The other $3,100 isn't spent, and doesn't roll over.

Months two and three look the same for A and B: 4 + 6 = 10 linking domains and $1,800 + $3,900 = $5,700 a month. Over three months that's 3 × $5,700 = $17,100, which closes both gaps in full ($5,400 + $11,700 = $17,100). D's content work is measured, and if it lands, the page's link case is recalculated from its new position.

In month four, A and B are finished. If the team decides C is worth it, its pacing (30 domains over 3 months, 10 a month at $700 = $7,000) now fits inside both the budget and the velocity cap.

Opportunity cost: the money you leave unspent

The $3,100 left in month one is tempting. Say the team approves page C anyway. The velocity cap allows 12 - 10 = 2 more linking domains, so C gets 2 × $700 = $1,400, not the $3,100 available. The cap binds before the budget does.

And each of those dollars returns $67 a month per $1,000, against $583 for D's content work and $200 for A. Spending on C now doesn't just buy a slow return. It uses up velocity and budget that a better option might need, and since the same domain is never bought twice, any domain that sits in both C's gap and B's is spent on the weaker page.

Unspent budget isn't waste. It's the cost of not buying things that don't pay. The right comparison for any spare dollar is the best thing you could do with it, which is often more content work, or nothing.

Scale in Ship/Scale allocates this way: budget goes where each dollar buys the most risk-adjusted revenue, and the monthly budget, maximum price per link, velocity cap and three-month pacing are never exceeded.

Where this goes wrong

• Errors compound across pages. Every figure in the ranking is an estimate: CTR, value per click, gap size, confidence. If they're all optimistic in the same way, the order may still be right while every payback is wrong. Keep inputs consistent across pages so the ranking is at least fair.
• Pages aren't independent. Two pages that compete for the same searches can take clicks from each other, and links to one page can lift others on the site. The model treats each page alone. If two candidate pages overlap, fund one and see what happens to the other.
• Demand moves. Seasonal searches can make a page look strong in the month you rank it and weak in the months its links arrive. Use a full year of volume where you have it.
• The ranking goes stale. Positions, prices and gaps change every month. Rerank before each month's spending, using measured results for work already done. The guide to measuring SEO changes at 30, 60 and 90 days covers how.
• Caps feel conservative until they aren't. A month where the caps leave money unspent will feel like a missed chance. The caps are there for the month when an overconfident estimate would otherwise have spent the lot.

The key points

Put every option, content or links, on one scale: risk-adjusted extra monthly revenue per dollar spent. Do content first where it carries most of the gain. Set hard caps before you rank (monthly budget, price per link, a payback threshold, site-wide link velocity and three-month pacing per page) and fund pages in order until a cap binds. Leave the rest unspent rather than buy something slow, and rerank every month.