Grow and Prune: How Expansion and Refinement Work Together in Google Ads

Phil Taylor

PPCers are often working in one of two modes:

• Expand targeting to increase volume
• Refine targeting to improve efficiency

Expansion and refinement point in opposite directions, but in a healthy account they also work in harmony, continuously, to shape and improve our Google Ads campaigns.

The key is understanding when and how to add more raw material to your inventory, and when to cut inefficient segments of spend and traffic, to create a profitable feedback loop.

Prune

Removing underperforming segments to improve overall CPA / ROAS, is a core tenet of ‘defensive optimisation’.

We have many variables at our disposal when it comes to finding and clearing out unwanted spend: from search terms, through keywords, devices, ad schedule, locations, and user-based variables like audiences (and here note that underperforming audiences are the most useful audiences to have on observation, as these can be excluded once identified.)

Some of these variables are safer to prune than others.

Excluding a device type or gender category would usually remove a very substantial chunk of traffic: a bold move with higher risks of unintended consequences. Locations and audiences can be removed in more bite-sized chunks.

But as we make these cuts – in small chunks or large – we will by definition be removing sources of traffic that have been feeding your campaigns.

Up to a point, that’s fine. Good riddance to bad ROAS – but taken to the extreme, a ROAS of 100 on a spend of $1 isn’t going to pay the bills. So – as we cut, we need new traffic to fill the gaps.

When we have ‘impression share lost to budget’ – that substitute traffic is already available – ready to feed our remaining targets with as much traffic as our budget allows.

(If you’re not sure how Impression Share Lost to Budget impacts this kind of optimisation, check out my quick video on why it’s a crucial signal)

But when we don’t have that ready source of new traffic available, the next move is to seek more raw material, by expanding our targeting in various ways.

Grow

Growing our targeting means adding inventory, creating more opportunities (for both us and the algorithms) to find winning segments of traffic. Smart bidding thrives on volume and – while it is easy to place too much faith in this idea – it is certainly true that more data gives the algorithms a greater opportunity to become more efficient without intervention. But more importantly for our purposes, adding more inventory affords *us* more opportunity to ‘trim the fat’ – and to do so in bolder ways – where there is clear and persistent underperformance not being dealt with by machine learning. Nonetheless, expansion should always be targeted—we should focus on whichever areas are most likely to provide high-quality traffic. Some of the most effective ways to do this include:

New keywords
Looser match typing
Removing restrictions – if we are targeting by audience, device, ad schedule, or – most likely – location, expand the permitted targeting outwards in the most promising directions.
Settings – we tend to place restrictions here as a sensible counter to Google’s overly-permissive default settings – but allowing Search Partners, for example, can be useful when we’re looking for extra inventory
More expansive campaign types: Introducing PMax where it is not the natural option – e.g. without a product feed – can be a worthwhile expansion method to try. (Note however that once we enter into activity that offers both less transparency into which segments of traffic are working and which are not – and fewer levers to act on the performance patterns that do emerge – we are adding a new challenge to the grow-and-prune project.)

Each of these approaches expands our inventory, but the real advantage comes when we pair them with strategic pruning. The goal isn’t just to grow – it’s to grow wisely, ensuring that every expansion strengthens or preserves overall performance while creating room for further optimisation.

Thin in One Place, Sow in Another

Expansion and refinement can work together in a complementary way – and can do so across unrelated variables.

For example, let’s say we find a heavily underperforming device category, but excluding it would reduce overall traffic uncomfortably. The idea here is that you can compensate by expanding in another dimension.

By increasing your location radius from 10 to 20 miles, you may add enough raw material to be able to afford to make that radical cut.

The additional traffic from location expansion may perform 20% less well than your current inventory, but if by excluding computers, you cut an equivalent proportion of your traffic that was performing only 50% as well as mobile, then this expansion into inventory that is worse per se, becomes an optimisation on balance.

There is also an important balance to strike between actively removing underperforming traffic segments, and trusting the algorithm to roam freely across them, seeking diamonds in the rough. A balance I went into more in this blog post.

Prune to Grow, Grow to Prune

We usually log in to an account with the purpose of either improving ROAS (“let’s see what we can squeeze or cut”) or growing revenue (“let’s see what we can loosen or expand”).

But we should always have with us both the shears and the watering can, as growth allows safe pruning, while pruning makes growth sustainable.