How has RankBrain changed keyword research?

5 months after Google announced it was using RankBrain to order its search results Gianluca Fiorelli still has the most accurate summary:

We know next to nothing about how RankBrain works.

Well to quote Councillor Hamann in The Matrix Reloaded “I have absolutely no idea how it works. But I do understand the reason for it to work.”

Google is using RankBrain to answer queries that don’t have a precise answer in the index and that may never have been asked before. In the same Bloomberg article that announced RankBrain Google’s Greg Corrado stated that this amounts to around 15% of the millions of queries it handles each day.

Corrado suggested that RankBrain is the “third-most important signal contributing to the result of a search query” but made clear that RankBrain doesn’t handle all queries, so I’m going to infer that RankBrain helps to determine the relevance of a result but doesn’t necessarily help to determine its quality (a reasonable result on an authoritative site is likely to appear ahead of a more accurate answer on a less authoritative site).

Optimising for RankBrain is fundamentally incompatible with keyword research models because it means finding relevant keywords with no search volume.

We need to be able to make a case to create content without being able to estimate how much traffic that content will acquire. Relevant keywords may never reach 10 searches per month which makes reporting success difficult and makes rank tracking pointless.

We need to rethink “relevant”.

Instead of creating content that is relevant to what users are searching for we need to create content relevant to what our product actually does.

I thought I’d explain 3 techniques we often use in keyword research and how they can be useful in a post-search volume world.

Internal site search

If visitors are using the search function on your website they think you are relevant for that search. The difference compared to Google is just awareness.

1. Set up a filter excluding internal traffic in your Google Analytics account.
2. In GA’s Admin tab click “View Settings” and ensure “Site Search Settings” is on.
3. In the “Query Parameter” field enter the parameter your site uses for internal queries (just perform a search on your website to find out and look at the URL). For Edit.com and most WordPress sites that parameter is “s”. Perform a search for “SEO” in our search box above and the URL will look like this: edit.co.uk/?s=seo (the parameter we would enter into GA is just s – no ? or =).
4. Wait. Searches in GA are not post-dated so it isn’t possible to see historic queries.
5. Access your Site Search data using the tab below.

This will give you a list of queries users currently in your funnel want answered. Back in 2012 we (Emma Barnes) analysed 27 of our clients’ site search reports and found that users were more than twice as likely to convert if they used a site search (conversion rates jumping from 2.41% among users who didn’t use a site search to 5.45% among those who did).

It’s much easier to make a case for creating content when you can prove you’re not serving your users (Ryan Stewart published a list of useful reports that will help you figure out what that content needs to look like).

Remember these users suspect you have the answer. Users who don’t know you will use Google – and the chances are good that RankBrain will process that query if its highly specific.

User testing

WhatUsersDo.com is a remote video testing service we find incredibly useful during a design and build process. Analytics platforms display quantitative data on how users behaved but aren’t always great at explaining why users behaved the way they did.

15 videos (£390/month) is quantitative enough to gain an understand of how users want to navigate our website and where the pain points for many users lie. It’s easy and cheap to profile participants and be reasonably sure that they match the kind of customers you have and/or want.

The qualitative information that usability testing provides is difficult to translate directly into keywords to optimise for but incredibly easy and powerful in making a case for specific content that should exist but doesn’t.

I talked about some more simple (and cheap) usability testing techniques at last April’s BrightonSEO.

Customer service

Many queries don’t get search volume because people still pick up the phone to customer service. Sometimes that’s because people want a personalised answer; more often it’s because they can’t find the answer they’re looking for online.

1. Ask your customer service team/call centre/social media manager to keep a log of the questions they answer on a day-to-day basis and send it over.
2. Create the best answer on the internet for that question and make it prominent on your website (create a blog post, include it on a product page or at the very least make sure visitors can actually find your FAQ page).
3. Assume that for every person who rings your customer service department your competitor gets a call from one of their own prospects. Acquire both of those customers and reduce the bandwidth for your customer service team.

Because RankBrain

RankBrain represents a fundamental shift for Google but is hardly a step change for SEO. We’ve been using all of these techniques for three years or more.

But it’s becoming increasingly important to look beyond search volume as our primary measurement for deciding what content to create. It blurs the lines between SEO, UX, product design, merchandising…now we have to ask who is our product really for and what do they need to know about it?

It’s no longer just about what your website should be relevant for – it’s about who you should be relevant to.