Google Hummingbird and (not provided): writing for questions

Two separate events landed at once and most owners have blended them together.

Hummingbird is the name Google gave to a new core algorithm, announced on September 26 and reportedly running since late August. It is designed to handle longer, more conversational queries, the kind people speak into a phone, and to weigh the meaning of the whole query rather than matching individual words. A search for "where can I get my car's brakes fixed near my office" should now return a brake shop page even if that page never uses the word "fixed." Previous updates such as Panda and Penguin were filters bolted on to the engine. This is the engine.

(not provided) is the row in your Google Analytics keyword report that has been growing since October 2011, when Google began encrypting searches for signed-in users. In September 2013 Google extended encryption to nearly all searches, so for most sites that row now covers 80 to 95 percent of organic visits and is still climbing. Paid clicks are unaffected: AdWords still reports the search query.

Neither event is a penalty. If you lost rankings in early October, check the Penguin 2.1 date (October 4) before blaming Hummingbird.

You are not completely blind. You are working with smaller, biased samples.

Three habits make these usable. Export Webmaster Tools data monthly, because it disappears after 90 days. Run a small AdWords campaign on a new topic for two weeks before writing the page, purely to read the query report. And in Google Analytics, switch your reporting from the keyword report to the landing page report, since the page is now your proxy for intent.

The old workflow was a spreadsheet of phrases sorted by volume. The new one is a list of questions sorted by business value. Spend a week gathering them from places that have nothing to do with search tools.

1. Sales and support email. Pull the last 200 inbound messages and tag the questions. Most businesses find 20 to 40 recurring ones.
2. Phone staff. Ask the people who answer the phone to write down every question for five working days.
3. Google autocomplete and related searches. Type your core topic with "how," "what," "why," "best," "cost," "vs" and record what appears.
4. Forums and Q&A sites. Quora, Yahoo Answers, Reddit, LinkedIn Groups and industry forums show the words people use when no marketer is listening.
5. Reviews. Yelp and Amazon reviews of your product category reveal the objections that become pre-purchase questions.

For each question record who asks it, where they are in the buying process, and whether you have a page that answers it in the first paragraph. The gaps are your editorial calendar for the next two quarters.

Hummingbird rewards pages that resolve the query, so structure matters more than density.

Open with the answer. The first 50 to 80 words should answer the question as a knowledgeable friend would. Qualifications and context come after.

Use the question as a heading. "How much does a kitchen remodel cost in Texas" works better than "Kitchen Remodeling Services" for both the reader and the engine. One main question per page; closely related follow-ups as H2 sections.

Cover the entities, not just the phrase. A page about mortgage pre-approval should naturally mention credit scores, debt-to-income ratio, lenders, documentation and time frames. Google is getting better at recognizing when a page covers a topic fully versus when it repeats a phrase eight times.

Use natural variations. Write the way the question is spoken. Synonyms, plurals and the long version of acronyms all belong in the copy.

Add structured data where it fits. Schema.org markup for products, reviews, events, recipes and local businesses has been supported since 2011 and increasingly shows as rich snippets. Google Authorship (rel=author linked to a Google+ profile) still produces author photos in results, which lifts click-through on informational pages.

Keep a human reviewer in the loop. Read the draft aloud. If it does not sound like something you would say to a customer, it will not read well to an engine designed around spoken queries.

Replace keyword rankings as your main metric with three page-level measures.

• Organic landing page sessions, grouped by topic. Create a content grouping in Analytics so "pricing questions," "how-to guides" and "product pages" roll up separately.
• Assisted conversions from the Multi-Channel Funnels report. Informational pages rarely convert on the first visit, and this report shows whether they start journeys that end in a lead or sale.
• Webmaster Tools impressions per page. Rising impressions with a stable click-through rate usually means you are being shown for more query variations, which is exactly what Hummingbird is designed to do.

Keep a small rank-tracking list of 20 to 30 head terms, checked weekly with a tool such as Moz, Raven or Authority Labs, as a smoke alarm rather than a scoreboard. Rank tracking costs $50 to $200 a month at this scale and is still worth it.

Stuffing long-tail phrases into existing pages. Appending "near me" and "how to" variations to every heading produces pages nobody wants to read and does nothing for a semantic engine.

Deleting the keyword report from your routine. It still shows 5 to 20 percent of traffic. Treat it as a sample and extrapolate carefully.

Spending on tools that claim to unlock (not provided). Several products infer keywords from landing page and ranking data. The estimates are useful for broad themes and unreliable for anything specific.

Ignoring mobile. Conversational queries are disproportionately spoken into phones. If your pages take more than four or five seconds to load on a 3G connection or need pinch-zooming to read, the answer you wrote will not be read.

Confusing Hummingbird with a content volume problem. Publishing more thin pages is the opposite of what the update rewards. Fewer, fuller answers win.

We treat the loss of keyword data as a forcing function rather than a crisis. The work runs in four stages. Discover: we audit analytics, Webmaster Tools and sales conversations to build the question inventory and map existing pages against it. Plan: we rank the gaps by commercial value and decide which pages to rewrite, merge or create, with a measurement plan agreed up front. Build: our content, design and development teams produce the pages together, so the answer is clear, the page is fast on a phone and the markup is in place. Improve: we review landing page and assisted conversion data every month and adjust the calendar. Because brand, growth, web and technology sit under one roof, the strategy does not get lost between an SEO vendor and a web developer.

A representative engagement is our content program work, where a business with a large, aging library of pages asked why traffic had flattened while their category was growing. We rebuilt the editorial plan around a question inventory gathered from support tickets and sales calls, consolidated overlapping pages, rewrote the most valuable ones to lead with direct answers, and shifted reporting from keyword rankings to landing-page sessions and assisted conversions. The approach is the same one we now recommend to any business adjusting to this year's changes.