Your business does not exist in a vacuum

If only businesses existed in vacuums, data & experimentation would be more repeatable and predictable. Most business owners would acknowledge that outside factors will play a part, but these seems to be forgotten when there is drop in visitors or a change in visitor behaviour on the site.. the knee-jerk culprit always seems to be poor old ppc campaign.

There are many factors that affect a businesses performance on the web. Some of those factors are online and some are offline. Here are a few things to consider when something unexpected happens, Remember “Nothing is constant. Things always change.” If you didn’t change, you can bet something or someone else did.

Online Changes:

Google rolls out another algorithm change for its search engine.
Keep an eye on this page, it lists them and the related dates.

Key Effects. Your organic search position & visitor numbers changes. Can be positive, but odds are it was negative for you.

Side Effects. Don’t be lulled into thinking that an Organic search change won’t impact on your AdWords Campaigns.
(PPC) price, click-thru-rates (CTR), quality scores & position become volatile. This often settles after a week or so. And usually results in a new market high for the bids, before it slowly decays again.

e.g When Panda 2.1 aimed at removing spam listings rolled out in Oct 2013, it was so successful at improving the quality of organic search results that its impacts were hard felt in many AdWords accounts. Ads that had a historic high CTR 15%+,  suddenly found their Ads weren’t getting clicked on at the same rate when the organic results were improved.

This resulted in poor quality scores because the Ads & keywords were now performing below their historic “expected click rates”. Keywords stopped serving as QS dropped from 8-10 down to 3 or below. Bids were increased in reaction to try and get them showing again. People with auto-pause or auto-bid on, found over the space of a weekend that either all of their keywords had been paused or their bids were way up. Campaigns set on “optimise for Conversion” failed to reach their targets and Cost Per Acquisition blew out. The run-on effect of all that machine optimising was not unlike the machine-led sell-off on the stock market . Because the Panda roll out started before it was officially announced, many people were scratching theirs heads at these sudden fails in their accounts.


The rise of mobile: This has been a long slow change, but it is gaining speed.

Key Effects: Reducing desktop search volume. Increased mobile & tablet site visitors. Different user behaviour. We are not in a multi device world. Research on mobile / tablet, purchase if easy, else may revisit on desktop. Expect to see increased Sessions to Conversion as they jump from device to device before purchase, but unique visitors will stay roughly the same.

Side Effects: In some industries it may lead to much reduced searching as major players position an app that locks people into reusing them – they’ll never need to search again. Has this happened in your industry? Just how much has your “new visitors” from search dropped?

Key Take-Aways

1. Flash is Dead! Yay! (thank you Steve Jobs). Odds are that at least 30%* of your site visitors can’t see your flash-based promo video on your site. Test it yourself with an iphone or Safari on new Mac. Can you see it? If not ask your web developer for help. HTML5 Video is the new way.
* check your own site stats in Google Analytics under “Audience / Technology / Browser & OS / Flash Version”. Look at how many (not set) you have in the “Flash Version” column. 

Offline Changes:

Events in the real world impact too. There are too many to list them all. Here are few to get you thinking.

Weather: Rain is great for umbrella sales, (and chocolate coated ice-cream oddly enough). But it isn’t so great when it is a flood and your in tourism. You may think that a flood or fires or any other natural disaster in a different city of your state won’t affect you. But if you are reliant on international visitors, they may not see it as geographically isolated as you do. Tourism to all of Japan suffered after the tsunami, all of QLD suffers when Cairns has a cyclone or Brisbane has flood.

A heatwave in a cool climate will impact on their desire to visit your tropical get-away.

Stock Market Volatility: Yes there are the obvious crashes like the Global Financial Crisis. But equally when the market is unstable it can affect consumer confidence and spending habits, just as much as the threat of unemployment can.

To be continued…

These aren’t the droids you were looking for?

Keywords matching has changed on a lot since the days when “broad”, “phrase” and “exact” meant what most people thought  they did. They have been subject to quite a bit of redefinition. Perhaps it is time they got new names? Especially now that turning off the “near match” characteristics are no longer at our discretion (as of Sept 2014).

Near match, oh that must mean close right? Not necessarily.. there is “machine close” and “human close”. If you were advertising for Glass you probably wouldn’t expect to see your Ads show for Class. Conceptually a very different meaning, but only 1 letter away if it is a misspelling.. so your Ad will likely show for both if it’s Ad Rank is sufficient and you hadn’t anticipated it by blocking -class as a negative.

Close keyword variations

We’ll show your ads for close variations of your phrase and exact match keywords to maximise your potential to show your ads on relevant searches. Close variations include misspellings, singular and plural forms, acronyms, stemmings (such as floor and flooring), abbreviations and accents.

OK. What about words that are just synonyms? Well it isn’t specifically mentioned, but we have observed these in the past. Particularly across branded goods, e.g car brands. A search for a particular make/model may show the make/model of another brand because they are same type of product. ie both 4WD vehicles. Incidentally I have seen “Ford Sale” expand to the generic “For Sale”(expansion on misspelling)  and if that is a phrase match your “Ford Sale” ad may be showing for furniture for sale !

Word Stemming is an issue when the stemming leads to a different meaning. e.g Service and Servicing. While many businesses may offer a “service”, it does not including “servicing”

Plural can introduce issues in a similar way. A search for “custom service” is likely different intent from a search for “customs service”

Abbreviations/Acronyms + Mispellings = EVIL . This can be nasty, if you have a common Three Letter Acronym in your industry, it is likely matching for every other variation of those letters and anything that is only 1 letter off.  ABC, CBA, CAB. POST, POTS,SPOT

Expect to spend a lot more time monitoring search term matching reports and adding negatives. For new accounts this may mean 7+ days before there is enough data before the report will show these unexpected matches. Experience plays a big part in knowing what might be an issue, but we can never fully know how aggressive some matching will be until it happens.

The chart below serves as an introduction to the different match types, ordered from broad to narrow.

Match Type Special symbol Example keyword Ads may show on searches that Example searches:
Broad match none women’s hats include misspellings, synonyms, related searches and other relevant variations buy ladies hats
Broad match modifier +keyword +women’s +hats contain the modified term (or close variations, but not synonyms), in any order hats for women
Phrase match keyword women’s hats are a phrase, and close variations of that phrase buy women’s hats
Exact Match [keyword] [women’s hats] are an exact term and close variations of that exact term women’s hats
Negative Match keyword women are searches without the term baseball hats

Happy negative hunting..

Attribution – Who gets the credit for the sale?

If someone comes to your site and converts, who gets the credit?
When our online Analysis tools were more primitive, that credit would be given to whatever got them there on the day. Often it was the last click or maybe direct if the they typed in the url.

But that assumes the customer journey to buying from you is a direct path. Marketing on the web is not an elevator pitch with a captive audience who must make a decision before they walk away. And yet many still view their web stats in this way. In this simplistic model a Visit gets viewed as failing if the conversion does not convert on that day, especially if it came from an paid click.

The path to conversion is not always direct. There are many ways that people interact with a website that lead ultimately (hopefully) into the “when” of the conversion.

  • They are researching
  • Ran out of time, will come back later.
  • Need to check out a few competitors or other brands to make the decision.
  • Started on a mobile, prefers the larger screen to read do the final purchase.
  • Have to discuss with Partner / Check Bank Balance / Wait for next Payday
  • Just not ready to buy today, but will next month.

But once they have come to your site, they are somewhere on the path to purchasing.
What was it worth for you to have this chance to show them what you do?
If their desires and your offerings align. You maybe one step closer to a sale than you were before.

How do you put a value on it even if it didn’t end in a sale today?
The attribution of value in each interaction leading up to a sale is complicated. Thankfully Google Analytics gives us tools to test and explore different attributions models. There maybe more value in your online marketing Campaigns than you have been giving it credit for.

I will leave it to Analytics Ninja Avinash Kaushik to explain this in more depth and how it can improve your bottom line. The Good, The Bad & The Ugly of Attribution Models.

Remove Known Bots from Analytics

Recently Google Analytics added “Bot Filtering” to “Exclude all hits from known bots and spiders”
Find this checkbox on the View Admin page under “View Settings”.

This will screen out some chaff that has been skewing your stats Analytics.
Cleaning up your Sessions & User counts going forward to exclude various robots & spiders.

You may (may not) be surprised to see that 10% or more of your traffic screened.
Don’t panic over a drop in sessions or users after putting on this filter.
(Remember they weren’t real prospects for your business)

What is important, is that behaviour statistics like time on site, bounce, pages per visit, repeat visits, visit paths, will be cleaner.

Let’s face it, you want to base your decisions on the behaviour of people on your site, not machines.
Better data, better decisions.

If you are having issues with others, particularly errant referral entries, such as semalt*com.
You may need to add a specific filter. Great instructions here. http://www.hallaminternet.com/2014/remove-semalt-google-analytics/ . If those instructions do not work for you, try using the regex “/.semalt/.com”

Hopefully GA will add these guys to their known bots list soon.

Dear Adwords, Brisbane is not a Tardis

Location targeting stats can be quite  misleading, Google Adwords even has a disclaimer text to explain that there can be a discrepancy between census data and reach data, because it contains devices, visitors etc.

However the general rules of physics should still be expected to apply, surely? In this screen shot you can see that the smallest geographic region: of “Brisbane – neighbourhood”, which is wholly (geographically) contained within the region of “Brisbane – city”, and therefor also contained within “Queensland, Australia – state”, seems to have the extra-ordinary properties of a reach greater than those 2 containing regions.



After not getting a satisfactory answer from Google Support who insist that this is all correct and as expected – we can only assume that Brisbane-neighbourhood  is a Tardis and thus bigger on the inside.