Machine learning is the hot word on everyone’s lips. Artificial intelligence is something that’s been spoken about in science fiction folklore for years – and we’re finally coming to a point where we can benefit from it.

Medically, the benefits are about to be reaped. In an article published by the Guardian earlier this week, it was reported that a new machine-learning system, developed by Google’s DeepMind, is as good as the best human experts at detecting more than 50 eye diseases, and is capable of correctly referring patients for further treatment with an incredible 94% accuracy.

But what does this mean for the world of advertising? Are we ready to hand our accounts to the machines quite yet?

Before I launch into our thoughts behind this new wave of advertising, I want to share a shocking statistic with you. According to a study produced by Asgard Capital last year, of the 715 companies based in Europe claiming to be using machine learning, only 409 (60%) of these companies actually were using it. While that might not seem so shocking at first, when you break it down you can see that a staggering 40% are actually lying about using AI in the bid to keep up with technological advancements.

However, there are agencies which employ this method honestly – and it’s those agencies I will focus on for this article. The information out there is scattered and somewhat confusing, so I’ll break down the top 10 tips you need to know about machine learning, or whether you should just stick to what you already know.

1. You need a baseline

How is your business performing? What are your projections for the coming months? Are your sales growing? Is a busy time-of-year, such as Black Friday, approaching?

Without this knowledge, there’s no way of you keeping track of whether the system is improving your advertising or inadvertently damaging it.

2. Huge amounts of data required

Machine learning, by definition, means the machine learns. To learn, the machine needs data; much like the DeepMind eye diagnostic, the machine will have learned how to scan and analyse the eye with a clear understanding of cell patterns to determine what is unusual. Without this deeper level of information, the machine can’t diagnose with accuracy.

The same rule applies to machine learning with Google Ads. Without feeding it huge amounts of data, it cannot improve and will subsequently struggle. It can take months to learn and become successful.

3. Review the machine learning system

Remember the shock statistic at the beginning of this blog? If you’re employing an agency to use machine learning for your account, make sure it really is machine learning and not just an automated set of rules. Be careful of this; the best agencies will be upfront and tell you which system they are using (we use Acquisio), but if they don’t, feel free to ask. It is your account, after all!

Sometimes you’ll find that an agency claims to have developed an inhouse machine learning system; in my experience, I’ve become sceptical of this. AI developers are amongst the highest paid developers; to develop a fully-functional system, you’ll be likely to need a whole team of developers which is beyond the budget of most agencies.

There are open-source machine learning systems such as TensorFlow which will get you up-and-running quickly, but generally, I find that inhouse systems are likely to be selling you an automated set of rules.

4. Garbage In, Garbage Out

Is your traffic clean?

Without clean traffic, the machine will learn bad habits and could end up causing chaos within the account. Your search terms and negative keywords need to be well-refined – this cannot be underestimated!

If you’re paying for poor quality traffic (bad keywords with no negatives), it simply can’t work. Your account needs to be completely optimised – AI can’t control negative keywords.

5. Are you already using it without realising?

It may come as a surprise to you to find out that Google Ads already has a number of very advanced machine learning systems built-in for free – especially for bid strategies.

Did you know that nearly 53% of Google Ads accounts are already using machine learning without even knowing about it?

In Acquisio’s review of 32,858 accounts, this was found:

12,651 were using machine learning on Google Ads
11,094 were not using machine learning on Google Ads
6,342 were using machine learning on Bing
2,771 were not using machine learning on Bing

As you would probably imagine, a large volume of these accounts are unaware that it’s already being used in the background – but it’s important to be aware of these things.

6. It’s costly

While the machine is learning, expect large increases in costs. Its first response will be to increase your budgets – and while you might experience more sales, it might not be producing a good return on investment.

It’s likely that it will then realise this and react by reducing the bids, confusing things once more while it learns about your account and adapts to its target market.

Naturally, you want to get as many sales as you can in a short space of time; I’ll use the analogy of a learner driver to show you why sometimes the slow game is the right game. In the same way you want to increase your sales quickly, you want to learn to drive quickly and at a high speed. But, as we know, there’s a lot to be learned about the road before you can put your pedal to the metal. While you’re learning, mistakes will inevitably be made, but you’ll eventually find your comfort zone and will be driving at the optimum pace.

It really is the same thing; the machine needs a deeper understanding of your account before it can perform at an optimum level.

7. Control your budget

One of the biggest reasons people use machine learning is to manage bids on huge volumes of products. This is something really doesn’t need machine learning – you can do it manually or use adwords cripts for large volumes, campaign structure and scripting are the real essentials for controlling your account. By dropping products into individual ad groups, you can let the Google system target your CPA (cost per acquisition) or ROAS (return on ad spend) goals.

Remember to be careful with your budget – you might have set an overall budget that is higher than what you actually want to spend. Your account may be set to spend £20,00 a month when you only usually spend £12,000; without control, the AI will attempt to use all the budget available and the full £20,000.

Without machine learning, budgets are only fully spent 23% of the time, whereas machine learning spends the full amount 77% of the time.

Without clear accurate budget control, it’s easy to lose out on a lot of money.

8. Machine learning will not save you

It’s a common perception that machines can control things better than people by eliminating human error; this really isn’t always the case. When accounts are struggling, don’t rely on the system to save you.

On average, a struggling machine learning account will run a month longer than a manual account before closing; this is because people put faith in the machine to fix their problems.

Trust your instincts – if you can see the account struggling, take appropriate action. Get specialist Adwords experts onboard, not a digital marketing agency. Digital agencies don’t have the commerical experience and large teams to brain strorm a solution to turn around a failing account.

9. Machines are not creative

This is a big one – ad copy needs to be creative for it to successfully entice people to buy. Machines are unable to create good ad copy, missing out vital CTAs. Without this, your advert will not be effective.

This is where you need a Google Ad expert to step in; they will tailor an advert to suit your target audience. Without good copy, there’s very little point in spending money on advertising.

10. Who should use machine learning?

As you can tell from the tone of the blog, we find it’s better to steer clear of it. However, this isn’t to say it won’t work for everyone!

If you’re in a rising market, this is a prime opportunity for you. There’s plenty for the machine to learn from without as much room for error.

However, if you’re in a static or failing market, this really isn’t a cost-effective solution for you. You need a carefully planned strategy from a great account manager to help develop creative campaigns to promote your products or services. A marketing plan is essential for it to run successfully.

Final thoughts

I’ve taken you through my thoughts about machine learning, and how I don’t think it’s as accessible as many people may think. If you have a well developed and refined account with high quality traffic, conversion tracking, a fully-optmised website and competitive products in a rising market – then machine learning is definitely worth considering. I recommend that you begin with testing the built-in systems before jumping into external systems, as it could cause disruptions to your growth and sales.

If your account doesn’t fit this criteria – stick with what you already know.

Has this tickled your interest? Get in touch with us and we’ll arrange a free account review for you!