Direct answers and cross device search are driving us
towards a ‘data driven’ model where cards are key.
Is your data accessible to Google?
Why should Google choose you as a data source?
76. Intent is more than just ‘keywords done better’;
consider the implicit signals and compound queries.
Data Driven Search is coming;
start considering what steps to take to prepare.
The rise of ‘Personal Assistants’ is an opportunity;
optimise on another axis (that isn’t PageRank based).
- 5 trends - the fire ants of the search
No tactics. No strategic insights. Will!
Paraphrase Dr. Pete
3 very related and already established
Interconnected - tried to pull them apart.
2 which are not here yet.
- Oldest of the trends I identified.
- Foundation of others.
- What does this person want?
- What would satisfy their search?
- More info. Suddenly easy.
- Implicit signals form ‘context’.
Already prompts questions:
How does this fit into a keyword research model?
Search ‘kinkos’ or ‘kinkos downtown san diego’?
Your know your phone knows your location, so don’t express it.
Shift in search behaviour.
These are some potential implicit signals.
User’s expect Google to know: location, device, language, search history.
One thing I’m interested in is what other implicit signals might we see.
Am I hungry? Am I tired?
- Have I just finished a workout?
- What am I doing right now?
- In fact - they may already know whether you are walking, cycling or driving.
- Then whole class of Hyper-local signals
- Google Physical Web & iBeacon
- Menus in restaurant- power/accuracy/indoors
- Bus stops
- Whackier one….
- What could prevent Google reading signs?
- Expect this trend to continue…
- Implicit outweighing the explicit.
- Google Now aims here.
- Entirely implicit searches.
To wrap up this section.
We do loads of keyword research for the explicit side of the side of the search query.
Another point, implicit signals allow you ‘intercept’ a search before it becomes active. Beacons outside your restaurant, or Google Now integration.
Challenge you with this question.
Discussed the implicit signals
On the flip side - the way we express the explicit signals is changing.
- Before I start this I want to ask a question.
What is a search query?
What is it made up of?
Typically we think of it like this.
Then what happens when you make a second query?
Second query - start over. New query.
‘cooking books’ -> ‘vegetarian cooking books’
- Hummingbird & Natural Language.
- Not an update. The foundation of something.
- It is what allowed Google to do things like… :click:
- Intent revision query.
- Second query was dependent upon first.
- Let’s look at another type of compound query.
- Chained query.
- Second query dependent on the first, but was in addition to it, rather than a revision of it.
How do you do KW research for that?
- Trend extends beyond just Google.
- A new Google Now competitor some of you may have seen.
- Notice how fast it is?
This is a compound query. I’m calling this a ‘nested query’.
Lets look at another example…
Multiple compound queries all at once.
Impossible to answer this query by taking you elsewhere - to a website
Draw a parallel between compound queries and faceted navigation.
Compound queries & natural language
- Cross website
Most of you will have heard about so I just want to touch on this one really quickly.
- Example from Rand. Great example.
- We already understand that Google’s understanding of keywords has advanced hugely in last decade.
- Next generation of searchers
- Trend to intents is more than just ‘keywords done better’
- The Implicit Search Signals & Compound Queries all play into this.
- Compound query -> Intent
Without the first search for Mario…
It is what is on the screen that used to help determine intent.
- Google on a mission to determine your intent in unexpected ways.
- For Google contextualisation for determining intent is a core focus.
- Implicit signal -> Intent
- Not just ‘keywords done better’
We serve user intents, not keywords.
Keywords capture only one part of the picture now.
Map landing pages to intents, then backwards to keywords.
Challenge - which pages which intents.
Discussed the 3 established trends.
Now 2 which I think are coming in next 12-24 months.
Lets start with…
10 blue links are long dead.
‘web search’ has not been ‘search for web pages’
Siri - answers many queries with an ‘answer box’
more and more types of query
- Google Now is the same.
- in assistant apps a ‘web search’ is a failure mode
App indexing: when we do see blue links:
The destinations are no longer always websites!
We are seeing this shift away from web pages.
I do not think it stops at apps.
- What are we shifting towards?
Lets start at this trend.
Research from Google (from 2012) shows people move from device to device.
People start things on a phone then move to a laptop etc.
Technologies like ‘hand off’ are fuelling this trend.
Start browsing and searching on laptop
On the bus I pick where I left off, on my phone
Holiday search on iPad on sofa.
Throw to the TV.
Then we also have this trend
- Have you noticed you are seeing more and more cards?
New UI paradigm
- Google Now, Twitter, Inbox, Siri and various places.
UI Paradigm - but they abstract presentation from data
Cards are units of data
This card represents this block of JSON-LD.
Cards can then be moved from device to device.
They can take on a display appropriate to the new screen or device.
- Adapt larger displays & beginning to see card like interfaces on desktop too.
Richard Baxter wrote an article, anticipating Google mobile become more like Google Now, with cards.
Last week Google rolled this interface out.
Google adopting cards more heavily! More and ore places.
- Where will the data for these cards, for these answers, come from?
- knowledge graph
- everything else?
It is going to come from us.
From our clients.
Already happening - featured snippets
They are learning to pull structured data from unstructured pages.
Will will talk about that.
Only works for static content - not calculators or live data.
Revisit this faceted nav scenario.
Imagine a compound query pulling via a data driven fashion via single interface
”Show me TVs” - “Just show me the 4K TVs.” - “Just the 40 inch TVs”
Displayed via cards: screen and device independent. Moved around.
Amazing for users.
Not sure we will see this exactly like this, but all the evidence points to the trend.
Will is going to discuss some tactical stuff to prepare.
Challenge - why you? Why should you ‘rank’?
This trend is the haziest for me, but wanted to include it as is interesting.
In last example I might search in a faceted fashion - single interface across sites.
This single interface is interesting
What Google - more sense when framed this way
- Nest acquisition, Driverless cars
Not going to talk about ultimate assistant, but there is a stepping stone we’ll see first.
Have you ever tried this?
Search using Google not only outward, but also your own data.
- Another example
Old classic - where are my keys?
You can now ask where is my phone and make your phone ring.
Using the same interface to find other devices or objects.
Another angle from which interfaces are converging.
October last year Google announced Inbox
New way to do your email
:: CLICK QUICK:: What’s interesting about it…
Google also allows Structured Markup for emails now
It enables emails to become ‘units of data’
It allows you to to be presented as cards, via Google’s main search interface.
Can also create Google Now cards using this markup.
Google enabling methods to feed from email into the single interface.
You can already do searches for some emails
Imagine as more and more emails are Machine Readable.
Already a better interface - natural language email search
Ultimate Assistant - one interface (Google, Siri, Cortana, Facebook’s M, Hound)
Search in one interface:
web, apps, your photos, your calendar, your emails, answers
Competing against more than just the web.
Once everything piped into one interface - they can make suggestions etc. (you’ve made reservations here before)
So much of the algorithm is now Machine Learning based.
Ranking Factors are no longer hand crafted.
Google Team can’t give you a check-list anymore.
We need to optimize for what the ML algos are optimizing for.