Last week I attended a Google Workshop with the following title:
Maximise Performance and Reach with Broad Match and Smart Bidding…
Your eyes may already be rolling.
This sounds a lot like yet another instance of Google beating its favourite drum…go big, go broad, aggregate, automate…
The title even contains the implicit assumption that maximising reach is an unadulterated good (which is part of the problem with this attitude of Google’s, and their reps, and their poorly-aligned automated suggestions).
Sure enough, the session largely did see Google delivering its usual message but this time, something was different.
This time we weren’t subjected to a barrage recommendations that felt poorly reasoned and poorly aligned with our interests.
Instead, Google took the trouble to further illuminate the logic behind some of its familiar chants… and in particular the reasoning behind favouring Broad match.
I’m no mouthpiece for Google’s recommendations, but I’m highly aware that the ground is changing beneath our feet, and being deaf to Google’s guidance would be just as dangerous as swallowing it whole.
So let me share what I took away from it…
At the top level, Google reminded us that the match type can be seen as merely the first layer in determining eligibility.
Broad match will make incremental searches available to trigger your ads, but it doesn’t determine that they necessarily will trigger your ad.
… Because after that first layer, Smart Bidding can come into the equation, allowing more selectiveness about which auctions you actually enter into.
So in the (seemingly incessant) messaging about pairing Broad match with Smart Bidding – there are two sides to the coin: On one hand we have Broad match ‘helping’ Smart Bidding by giving it a wider playing field to work in. That’s the intuitive argument, and the one we usually hear…
But on the other hand Smart Bidding is a potential method of keeping Broad match more in check, by playing referee to the larger group of misbehaving search terms now on the field.
Playing that role successfully relies on the algorithm developing a clear picture of what it and – more to the point – we are looking for, so that it can discriminate between good and bad search terms… hence the often-touted importance to smart bidding of a high quality and high-enough quantity of conversion data.
It’s not an entirely new idea, but worth pausing for a moment to ponder… When we talk about Broad match casting a wide net – it isn’t necessarily trawling that big net through the water every time and giving you the whole range of search terms without further discrimination…
It is part of the remit of any conversion/value-oriented bid strategy to minimise your participation in hopeless auctions. That isn’t the end of the story (incentives will never be perfectly aligned) but it is part of it.
Next, I was struck by this line, delivered during the presentation:
‘Broad match can make you eligible for less competitive auctions with equally high intent’
This is the theory behind the supposed efficiency advantages of Broad match.
When we see recommendations like this:
I’m sure I’m not alone in scoffing at the word ‘better’ (and frankly – in my experience – Broad match ROAS still very rarely is better… but we’ll stick to the theory side for now).
This idea is fleshed out further in the white paper that Google sent round after the session. The relevant page is worth reading in full:
So the emphasis is on Broad match giving Google the flexibility to spend less on auctions where the match is tighter but the CPA is predicted to be higher – and replace that spend with (again, in theory) judicious use of looser matches, where the CPC is likely to be lower, but its various signals suggest that a conversion is not out of reach, and the auction is in the lower CPA bracket.
It’s an interesting idea, and it’s a challenge to the traditional model of spending maximally on tighter matches (with tighter matches being a decent proxy for higher-intent) before opening up to looser matches where valuable clicks are more widely interspersed among the waste.
Smart Bidding – the theory goes – can assess that less promising territory and spend your money selectively to gather what value hides within it.
Next, there was a lot said on the concept introduced to us last year, that Broad match gives us access to unique signals not available under other match types.
‘Location’ is an interesting one, and it raises an eyebrow…
Location has always been given as one of signals used by Smart Bidding (you can find those signals listed on this page).
…and of course we expect the Smart Bidding strategies to use location as one of the more straightforward variables by which to evaluate an auction and determine its bid (after all, Google won’t let us adjust bids by location ourselves when using Smart Bidding…)
But the exclusive use of location as a signal with Broad here refers only to the process of determining auction eligibility.
In the example given, for a search on ‘restaurants near me’, only a Broad match keyword can take the user’s location into account in order to determine relevance:
The use of location as a signal of value once the search term-keyword match has been made is safe and well with other match types, I believe.
‘Previous searches’ is another important Broad match-only signal to note.
To give a generous spin on it, this is one of the key explanations for those apparently wildly-irrelevant search terms that show up under Broad match.
Users may have made relevant searches previously, and due to those searches (along with other signals) be deemed to be convertible for you when making their subsequent, less obviously-relevant search.
Making a large-scale move to Broad match shouldn’t require a structural overhaul – even if you’ve been weaving match-types into your current structure.
This is partly because of the new matching rules brought in last year, by which the more relevant instance of a multi-match type keyword will take precedence, regardless of ad rank.
Keep Broad, Phrase and Exact variants if you like – the former won’t interfere with the latter in the way that we used to have to combat with internal negatives.
Also on structural thinking, here’s another point worth pausing to consider…
Speaking for myself, the statement ‘bidding algorithms learn at the query level rather than the keyword level’ is one I will benefit from really assimilating, and carrying into my optimisation decisions…
With all of these points in mind, we can see that the bulk of Google’s recommendations (even the annoying ones) do blend coherently.
e.g. RSA Ad Strength is only a measure of how much leeway you allow Google to play with your RSAs, but that leeway is significant with respect to the benefits of Broad match and Smart Bidding:
Start with a range wide of auctions, try serving each one as best you can, and learn from there… Hopefully, with Smart Bidding applying a genuinely (and increasingly) effective filter to determine auction relevance and value.
But as I become more enamoured with the theory behind all of this, I have to remind myself (and you may be yelling the same thing at the screen by now…) that the theory isn’t what ultimately matters.
We also have to see the benefits of Broad Match and other aggregation recommendations in practice.
(From my own tests, Broad match beating other match types on efficiency is not something I’ve seen with any regularity… though I have certainly seen Broad match improving in efficiency vs where it was a couple of years ago).
And sadly we can’t trust Google’s reports on how well these things work in practice… Their selective and confounding use of stats when trying to persuade us how effective various features are for advertisers, is not even subtle.
Even in the very thorough white paper we were sent after the session discussed above, there are multiple claims like this:
Where to begin with a statistic like this?
Can see an average?
As if ‘average’ wasn’t already susceptible enough to misleading conclusions, fudging it with ‘can’ just demolishes it as a serious piece of information…
And then what proportion of these advertisers benefit in terms of ROAS…? Any mention of the attendant change in cost…?
In short, I’d like to see Google treat us as adults in regard to the data they use for persuasion, in the way that – in my view – they now have in regard to the theory.
So we must keep testing these suggestions out for ourselves.
But what this exposition of the theory does – at least for me – is give a boost to my impetus to keep on testing, and as far as I can bear it, to go ‘all in’* on that testing – understanding that each aspect of the recommended approach really does – or should – affect the whole.
*(Though before going head-first on this one, remember the safer option of using a Campaign Experiment to test the impact of Broad match – made easy when it appears with the Broad match suggestion on the recommendations screen.)