Blog
Before You Diversify Off Google, Find Out What You Are Actually Losing On It
Measure, recover, then decide. The order is the whole argument here.
The advice going round says a business dependent on Google search is one product decision away from trouble, so start spreading budget today. Sometimes that is exactly right, and the conditions that make it right are worth stating at full strength.
What I am pushing back on is narrower. Moving budget to a second channel does not stop a leak in the first one. It buys a second channel and keeps the leak. Between noticing that Google feels expensive and funding something else sits a step called PPC budget waste recovery: finding the spend going to searches that never produced a customer, and deciding what to do with it. That step reliably goes missing, usually because finding out honestly used to cost more than most advertisers could justify. That stopped being true in November 2025.
What does "wasted PPC spend" actually mean?
Wasted PPC spend is money that went to clicks from searches with no realistic chance of producing a customer. The best account level measurement of it puts the figure between roughly 6% and nearly 40% of budget, and anyone handing you one number for your industry is handing you their methodology, not your account.
That range carries a caveat the category skips. Nobody measures the definition above. What gets measured, in the studies below and in nearly every published figure, is spend on search terms that recorded no conversions. Such a term can have assisted a sale credited elsewhere, or converted after the reporting window closed. Read every number here as the proxy it is.
Igor Ivitskiy at Doctor Ads examined 31 Google Ads accounts carrying $133 million in managed spend, September 2024 through February 2025 (published 2026-07-14, updated 2026-09-07). Within a single industry, the share of budget going to search terms that recorded no conversions ran from about 6% in the best run account to nearly 40% in the worst. Notice which end matters: the cleanest account still carried about 6%, a floor rather than a target, and an account near it has almost nothing left to recover.
WordStream, written up by Susie Marino and updated 2026-06-11, pulled 251,236 grader reports from 15,666 accounts between 2025-01-01 and 2025-11-17. Median monthly spend, about $3,127. Median monthly wasted spend, on WordStream's own definition, about $1,127. Twenty nine percent of those accounts recorded no conversions at all over 90 days, and a quarter had never added a negative keyword. Do not divide those two medians into each other; they describe different distributions, so the ratio is not a waste rate.
There is no industry waste number, which the category never says out loud. The 25% figure, the 76% figure, the 42% for unmanaged broad match, the 62% of SMEs line: none lead back to a primary source publishing a method.
Why is "diversify or decay" a serious argument?
"Diversify or decay" is a serious argument because several unrelated things are true at once, and each of them is sufficient on its own.
The search environment is genuinely moving. Pew Research Center tracked 900 US adults across 68,879 Google searches in March 2025 and found that users who saw an AI summary clicked a traditional search result in 8% of visits, against 15% for users who saw none (published 2025-07-22). That is organic click behavior and not paid performance, so anyone telling you paid clicks halved is misreading it.
Platform dependency is a real exposure, and it is not an efficiency problem. Google publishes a schedule and accounts change on it whether or not you agree. Campaigns using automatically created assets and the campaign level broad match setting are being auto upgraded to AI Max starting September 2026, per Google's own post (blog.google, published 2026-04-15, updated 2026-06-11).
The multichannel literature points the other way from me. Analytic Partners describes its ROI Genome program as 25 years of insights from more than 1,000 brands, and its consistent public direction is that adding channels improves marketing ROI. Commercial research from a firm that sells measurement, but the direction is not in dispute.
And a channel already buying nearly every impression worth having on the converting queries cannot absorb recovered budget. If any of that is your situation, diversify. The disagreement is about what you do first.
In the channel you already run, you have something to measure against
In the channel you already run, recovery gets measured against something real. That asymmetry is the reason sequence matters here, and it is not a tidiness argument.
In an account you have been running, you hold a baseline, a spend history, a record of what people typed before they bought or did not, and a sense of what a normal week looks like. In a channel you have never run you hold none of them, so the first months of spend buy information instead of customers. That is a description of what you will be able to conclude in January about a decision you made in October.
Some paid search spend genuinely does nothing, which is the foundation under all of this. Blake, Nosko and Tadelis ran a randomized geographic experiment at eBay, published in Econometrica in 2015. Switching paid search off in randomly selected geographies showed that for brand keywords, organic listings were close to a perfect substitute and the measured return on that spend was strongly negative. eBay was a household name with dominant organic presence, and none of it generalizes to a company nobody searches for by name.
The second paper is the one I would hold up against any before and after chart, including one an agency shows you. Gordon, Zettelmeyer, Bhargava and Chapsky compared measurement approaches across 15 field experiments at Facebook, published in Marketing Science in 2019, and found observational methods frequently failed to recover what the randomized tests showed, generally overstating ad effectiveness. Which costs my side of this argument more than the other one.
Half the non converting money sits in roughly a tenth of the search terms
Half of the spend that recorded no conversions sits in about a tenth of an account's search terms. The size of that tenth is the part nobody prices.
Ivitskiy's second study looked at 32 accounts and 5.2 million paid search terms with zero conversions (published 2026-08-03, updated 2026-09-07). The most expensive 10% held a median 48.8% of the non converting spend, and the top 5% held 34.6%. Spread evenly, the top 10% would hold 10%.
Then the scale. That top tenth is a median of 13,806 search terms, because the median account there carried 138,066 terms that spent money and returned no conversion.
The counterintuitive result concerns negative keywords, meaning terms that stop an ad showing on searches containing them. Across the 31 account portfolio, list size showed almost no relationship to waste share, a correlation of r = -0.10. Adding a whole query as a negative blocks one exact string, while the leak is usually a single word turning up across thousands of otherwise harmless queries. The shape repeats: an account is organized the way the platform organizes it, and the money goes missing a level below that, at the query and at the words inside it.
There is also a hard limit on what you can see, and Google documents it. The search terms report covers terms used by a significant number of people, and terms without enough query activity are omitted for privacy, so the waste you can see is a subset of the waste that exists. That is the broader point in over reliance on Google Ads automation too. The complaint that Performance Max will not let you exclude queries is out of date: Google documents negative keywords at campaign and account level, reaching Search and Shopping inventory and nothing else.
One dated fact is worth getting right, because trade coverage did not update it. Dynamic Search Ads are not retiring in September 2026. Google's 2026-06-11 update moved that sunset to February 2027. September belongs to automatically created assets and the campaign level broad match setting, both upgrading into AI Max for Search, which combines search term matching, text customization and final URL expansion with brand and location controls. Catching that early is most of what query level paid search management consists of.
When is diversifying the right call?
Diversifying is the right call when any one of four conditions holds. Not a balance of them, any single one.
| Condition | What it means | Why it flips the answer |
|---|---|---|
| Impression share ceiling | You already buy nearly all impressions on the queries that convert | Recovered budget has nowhere productive to go inside the channel |
| A causal test says the channel is not incremental | A randomized or geographic test says the results would have arrived anyway | The spend is buying something you were getting for free |
| Your buyers transact where this channel does not reach | From data, not a persona document | No amount of query hygiene puts you in front of them |
| Concentration risk beyond your tolerance | Platform decisions land on a published schedule | A risk position, independent of efficiency |
Incrementality is the share of results that would not have happened without the advertising. That second row used to be theoretical, because measuring it was out of reach. Google announced on 2025-11-11 that it had lowered the minimum spend for an incrementality experiment from what it describes as upwards of $100,000 down to $5,000. That is Google's figure about Google's own product, not independent evidence that the tests are good. What it establishes is a price: a causal test that was a six figure commitment is now quoted at four.
Worth naming what is not on that list. Rising costs alone do not qualify, since costs rise in the new channel too. Nor does one bad quarter, nor a competitor's press release, nor a general sense that the channel has gone tired. Diversification is the right answer to a demand ceiling, a dependency risk, or an audience that is genuinely somewhere else. It is not an answer to a measurement problem.
Run this sequence before you move a dollar
Seven steps, in order. This is the whole of PPC budget waste recovery as a method, with nothing claimed about outcomes.
- Verify the conversions are real first. Reconcile platform conversions against actual orders or qualified CRM records, because waste analysis on a broken conversion number is wasted analysis.
- Export the search terms, filter to zero conversions, sort by cost descending. Work the head, not the alphabet.
- Count spend at the word level, not the query level. An n gram analysis counts spend and conversions by individual words across all of your search terms instead of by whole queries, surfacing the single word draining budget across thousands of innocent looking searches.
- Treat the expensive head as candidates to examine, not a kill list. Assisted and late converting terms live in there, and the zero in the export is a reporting artifact as often as a verdict.
- Check relevance from query to ad to landing page before blaming the query. Sometimes the term was fine and the page was not.
- Where the campaign type allows it, apply exclusions at campaign and account level instead of rebuilding the same list inside every campaign.
- Then, and only then, run a causal test on the remaining budget before moving any of it.
One thing a numbered list hides. If nobody in the business has the hours to work a list of thirteen thousand terms, and nobody is going to be hired to, the honest answer is not to promise an audit that will never happen. Buy the capacity, or make the channel decision without the measurement and say so when you make it.
The question is answerable now, which it was not at this price a year ago
Measure, recover, decide. In that order, and with the decision made against a number instead of a feeling.
Most of the time "diversify or decay" gets deployed as a strategy when what is happening is a budget reallocation with no measurement attached to either end of it, which is a transfer with better branding. The alternative used to be priced out of reach for anyone spending under a few million a year, and that stopped being true in the last twelve months. There are case studies on the site.
Find out what the channel is returning. Then move the money, or leave it where it is, and either way you will be able to say why.
FAQ
How much of a PPC budget is typically wasted? There is no reliable single figure. One 2026 analysis of 31 accounts found spend on search terms that recorded no conversions running from about 6% in the best run account to nearly 40% in the worst, within the same industry.
Is it better to fix wasted spend or move budget to a new channel? Fix first, then decide. Recovery is measured against a baseline you already hold. A new channel has no baseline, so its first months of spend buy information instead of customers. Move budget only after a causal test on the current channel.
When does diversifying off Google actually make sense? When any of four conditions hold: you are at your impression share ceiling on the queries that convert, a causal test shows the channel is not incremental, your buyers demonstrably transact where this channel does not reach, or concentration risk exceeds your tolerance.
Do more negative keywords mean less wasted spend? Not reliably. Across 31 audited accounts, negative list size showed almost no relationship to waste, a correlation near zero. Adding a whole query as a negative blocks one exact string, while the leak is usually a single word appearing across thousands of queries.
Can I see every search term my ads run on? No. Google's documentation states the search terms report covers terms used by a significant number of people, and that terms without enough query activity are omitted for privacy. The waste you can see is a subset of the waste that exists.
