Landing pages for PPC: one message, one action, a test plan

A home page is written for everyone, so it answers no one. Someone who searched "commercial roof repair Denver" and clicked your ad is paying you the courtesy of telling you exactly what they want. Sending them to a page about residential, commercial and gutter services with a news feed and a staff photo wastes that.

Three costs follow:

1. Lost conversions. On the accounts we audit, moving a campaign from the home page to a matched landing page typically lifts conversion rate by 50 to 150 percent. That is the difference between a channel that pays and one that does not.
2. Higher cost per click. Google's Quality Score includes landing page experience. A page that is relevant to the keyword and loads quickly earns a better score, which lowers the bid needed for the same position. Google has been explicit this year that relevance and originality of landing page content matter.
3. Nothing to learn from. A home page gets traffic from everywhere, so you cannot tell what paid visitors did or test anything cleanly.

The structure that works is consistent across industries because it follows how a visitor reads: confirm, persuade, act.

What does not belong: the main site navigation, a sidebar, links to social profiles, a newsletter box, and any second offer. Every exit you remove raises the share of visitors who do the one thing. Keep a small logo link to the home page for the people who genuinely need it, and nothing else.

Message match is the discipline here. If the ad says "Free Roof Inspection," the headline says "Free Roof Inspection," not "Welcome to Rocky Mountain Roofing." Visitors decide in a few seconds whether a page matches what they clicked, and the back button is one thumb away.

Each field you add costs you completions. For a first contact, name, email and phone is plenty; ask for everything else on the phone. If sales insists on qualifying fields, test the short form against the long one and let the data settle it. Put the privacy line directly under the submit button, label the button with the outcome ("Get My Free Estimate," not "Submit"), and send the visitor to a distinct thank-you page so you can count the conversion.

For phone-driven businesses, put the number in large type at the top right and make it a clickable tel: link for the growing share of visitors on smartphones. Call tracking numbers specific to the AdWords campaign let you count calls as conversions too.

Google measures landing page load time as part of Quality Score, and visitors measure it by leaving. Target under three seconds on a typical connection: compress images, avoid Flash entirely, keep scripts to Analytics and your form. Check the page on an iPhone and an Android device. Mobile is a minority of paid clicks for most accounts this year, but it is growing fast and a page that needs pinching to read converts poorly. If you run a mobile campaign, build a separate mobile page with a click-to-call button as the primary action.

Nothing below works without clean measurement, so set this up first.

• AdWords conversion tracking on the thank-you page, so conversions appear next to keywords and ads in the AdWords interface.
• Google Analytics goals for the same thank-you page, so you can see paths and compare paid against other sources.
• Auto-tagging on in AdWords, with AdWords and Analytics linked, so keyword-level data flows through.
• One landing page per ad group, or at least per theme, so you can read results by intent.

Once a week, pull a report of conversion rate by landing page and by keyword. Keywords with high click-through and low conversion usually need a page that matches them more closely, not a lower bid.

Google Website Optimizer is free, integrates with Analytics, and handles A/B and multivariate tests. Unbounce and similar tools add page building and testing in one place for a monthly fee. Either works. The plan matters more than the tool.

1. Establish a baseline. Run the page unchanged for two weeks and record conversion rate and cost per conversion.
2. Rank hypotheses by expected impact. Headline and offer first, then the form, then the hero image, then layout and button copy. Color changes last.
3. Test one thing at a time with an A/B split, unless you have thousands of conversions a month and can afford multivariate.
4. Run to significance. At least 100 conversions per variation, at least two full weeks, 95 percent confidence in the tool's report.
5. Keep a log. Date, hypothesis, variation, result, decision. Most teams repeat tests they already ran because nobody wrote it down.
6. Ship the winner, start the next test. A landing page is never finished; it has a current best version.

A realistic expectation: a headline and offer test on a page converting at 6 percent might move it to 8 or 9 percent. Three or four such wins across a year compound into a channel that costs a third less per lead than it did in January.

• Testing on too little traffic. A campaign with 300 clicks a month cannot support testing. Improve the page using the structure above and test later.
• Changing bids mid-test. Different traffic mix, different result, no conclusion.
• Running the same page for search and display. Display visitors are colder and need more explanation; search visitors need less.
• Hiding the price when price is the question. If people search "cost of," the page should say what it costs or give a range.
• Declaring winners on three days of data. Monday and Saturday behave differently.

Landing pages sit at the join between paid media, web design and measurement, so we handle them as a connected brand, growth, web and technology engagement. Discover reviews the account, the search terms report and the current pages to find where clicks are leaking. Plan groups keywords into themes, defines one page and one action for each, and sets the tracking spec. Build produces the pages, fast and on brand, with conversion tracking and Analytics goals verified before traffic goes live. Improve is the standing test program: a prioritized hypothesis list, one test at a time, a log, and a monthly review of cost per lead by theme.

A representative engagement is on our performance work: a services business running AdWords to its home page with a cost per lead that made the channel marginal. We split the account into intent themes, built a matched landing page for each with a short form and a tracked phone number, and ran a sequence of headline, offer and form-length tests through the year. Cost per lead fell steadily as each winner shipped, and Quality Score improvements lowered the clicks' cost at the same time.