Machine Learning started back in 1950, when Alan Turing created the “Turing Test” to determine if a computer could possess real intelligence (to pass the test, a computer would be required to fool a human into believing it is also human) and for almost 40 years it focused on a knowledge-driven approach.

In the 90’s, the focus shifted to a data-driven approach and Engineers started creating programs for computers to analyse large amounts of data and draw conclusions — or “learn” — from the results.

They developed programs based on learning algorithms to solve almost any kind of “perfect information problems”, initially applying machine learning principals to simple video games (everything started with Atari – Breakout) then moving to more complicated games and problems to solve (1997 — IBM’s Deep Blue beats the world champion at chess)

deep blue


If we think about Artificial intelligence, the first thing we imagine are the shiny robots from one of those 80’s and 90’s Hollywood films (everyone loved Arnold in Terminator – “I’ll be back”). However, machine learning has come a long way and today algorithms enable computers to communicate with humans, drive cars, write and publish sport reports and find terrorist suspects: these algorithms teach themselves to grow and change when they’re exposed to new data


Artificial Intelligence and Machine Learning are two very hot and interest raising words and, despite often being used interchangeably, they are not quite the same thing:

  • Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider “smart”*
  • Machine Learning is a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves*

We all experience machine learning on a daily basis and AI is almost everywhere, impacting our life and guiding our decisions. It’s affecting every industry you can think of (from retail to weather forecasts) and its impact on advertising is really starting to make a difference in how marketers target and deliver ads.




Since 2013, the two Big Giants (Google and Facebook) introduced machine learning to online advertising with automation solutions to help marketers improve their digital campaigns using the same principles that were applied to defeat Kasparov in 1997. The majority of those solutions have been directed towards bidding strategies (CPA, ROAS, CPV) as well as DSA (Dynamic Search ads), Customer Match, Audience insights, UAC (Universal app campaigns), lookalike targeting and predictive performance analysis. The immense speed of these advancements, means many more solutions are set to come in the future.

Some of the coolest machine learning applied to a number of different components of advertising can be summarised in*:

  1.  Measurement and attribution (market mix models, causal modelling from observational data, propensity matching)
  2. Cross device association (predicting the probability that two devices belong to the same person based on usage patterns, IP overlap)
  3. Intent prediction (what is the probability that a consumer is going to buy that new car in the next month or so) on an individual level.
  4. Response prediction on an ad impression level (probability of a click or a video completion).
  5. Audience insights (extracting behavioural patterns for instance to inform creative design).

So, in order to see the real benefit of machine learning and to reduce human time (and errors) against manual optimisation, we need to use as much as we can, all the available automated solutions provided by advertising platforms and guide the machine in the desired direction. Thereby, allowing the algorithm to learn and adjust itself providing the best result available. Some such examples are as follows:

1. Automate our campaigns as much as possible

a-Use/test available automated solution provided by advertising platforms and evaluate the impact on performance

b-Where possible, change the attribution model from a last interaction to a position based or time decay attribution model to better evaluate the impact of our cross-channel campaigns

2. Always track conversion/interactions with our ads and focus our strategies on CPA/ROAS (where possible) letting the platform adjust itself to improve overall              performance

3. Use Clients CRM data and to create better and more accurate Custom Audiences will improve the efficiency of our activity

4. Increase customers LTV by cross-selling: Custom Match and Custom Audience data to improve customers’ loyalty and raise Clients revenue/ROI

5. Trust platform integrated insights solutions to better engage the audience we want


Future thinking – How Voice Search and AI will change the Digital marketing landscape

Voice Search is probably one of the latest and coolest AI advancements. Having seen big growth in users over the past 5 years, it’s one of the biggest and most recent changes impacting the world of digital advertising.
As reported from, Google CEO Sundar Pichai announced during his Google I/O keynote that 20% of queries on the Google mobile app and on Android devices are voice searches.

In December 2015, survey data from MindMeld found that there has been a significant increase in voice assistant and voice search usage; 60 % of people who engaged with the survey said that they had started using virtual assistants and voice search in the past 12 months.
From the most recent comScore study, among people in the U.S. who owned a smart speaker in the first quarter of 2017, 60% used the device to ask general questions. Another outcome from the comScore study was: 57% of people said they use the voice search to check the weather and 54% use it for streaming music. Only 22% use it for streaming news, 16% to find local business and just 11% to order products.


google home

The population of people who have a voice assistant is growing rapidly, and with it the volume of voice search; not only because it’s cool and the latest tech trend, but more likely, because it allows you to search on the go, it’s fast, convenient and the technology is becoming more reliable and precise.
So, how do we stay on top of the technology and deliver the best solution for our Clients without losing search volume?

5 actions for SEM to stay ahead of technology

1. No more page 1 results

After spending about twenty years searching for the perfect solution to improve site rankings on search engine result pages, marketers have to re-think and adapt websites, focusing on building content to respond to possible questions a customer might ask. Users are no longer being presented ten different options to consider before picking the one they deem fit. In fact, they’re rarely even reviewing two and usually the first result they receive either satisfies their query or they decide to change their initial search. This means being number two or three in the rankings isn’t good enough anymore.

2. Adapt for human communication

Voice search will be a more human experience, with users passing from typing to asking questions and seeking the best answer/result. This triggers the need for marketers to develop a more human approach, adapting website content against long tailed keywords and also providing content on third party sites as well as adjusting SEO to use keywords related to questions.

3. Long Tailed Keywords for PPC:

Voice searches are usually longer than normal text searches (which usually have 3, max 4 words) and usually they end with a question. As voice search usage continues to grow, marketers will have to get smarter and answer those questions in order to appear with the most relevant ad and engage the right customers at the right time.

4. Keywords intent: from interest to action

A recent Bing study on Cortana usage has revealed what kind of questions users ask to voice search:

– How 41%
– What 34%
– Where 14%

The natural language usage of voce search allows us to understand different customers intent and interest they are expressing during a conversational search so we can craft our keywords list and tailor ad copy against different stages of the conversion funnel, increasing relevancy and conversions.

5. Think local

Mobile voice search is three times more likely to be used for local-based queries than text searches. Saying that, the implications of voice search should put local business owners and marketers on high alert, critically considering how to best prioritize local optimization; leveraging local knowledge and ensuring every appropriate page has business name, address and phone number to improve visibility.

If you have any questions please do not hesitate to get in contact.


The following two tabs change content below.

alessio fasoli

I am Reprise's Associate Director in Paid Solutions. I've spent the last 7 years working in paid media for the digital marketing industry in both Europe and Australia. Here at Reprise I work with some of the country's leading brands, helping them make the most of their digital campaigns providing strategic thinking and approach

Latest posts by alessio fasoli (see all)