An ecommerce client fired JudeLuxe two months into the relationship even though return on ad spend and contribution had both risen and spend had fallen, according to the agency's co-founder, Chris Avery, who told the story in an episode of PPC Live The Podcast published on YouTube on November 19, 2026.
In Short
An agency improved a retailer's return on ad spend and margin within two months and was dismissed anyway, because what the retailer actually needed was to clear seasonal stock and get cash back into the bank. The story shows how far a platform target can drift from what a business needs, and it touches every agency, freelancer and in-house team that accepts a target without asking where it came from. At JudeLuxe, Avery says, stock, cash and trading priorities are now discussed before anyone talks about the Google Ads account.
The review that ended the account
The episode, hosted by PPC Live founder Anu Adegbola, belongs to a series built around a single premise: a paid search practitioner describes a serious mistake and what followed. Avery's choice took the conversation back "four or five years", to a period when JudeLuxe was young, run by two people and, in his description, eager to prove itself.
The agency had won the account from an incumbent that Avery estimated at 20 to 30 times its size, "with layers upon layers of people, account managers, account directors, account executives." There was, he said, "definitely some pride involved" in taking business from a competitor of that scale.
For two months, the metrics moved in the direction JudeLuxe wanted, ROAS among them. "When we looked at the numbers, [ROAS] was up, contribution percentage was up, spend was down," Avery said, describing the cuts as removing "what we would or most people would define as waste." The agency built "a full slide deck for the review" and went into the meeting expecting praise. On the call were the client's managing director, finance director and operations director.
"Like, these numbers, man. Like, look at them. Like, they're all green," Avery recalled thinking.
The verdict arrived afterwards, by email: "we're ending the relationship. We're going to settle your invoice prior." The notice period, Avery said, was zero. His first reaction was not reflective. "How on earth can they sack us with results like that?"
| What the agency presented | What the business needed |
|---|---|
| ROAS up | Stock that had to be cleared after the main season |
| Contribution percentage up | Cash, which was "short in the bank" |
| Spend down, with cuts framed as waste removal | Advertising that moved inventory, even at a lower return |
| Performance against an account-wide average ROAS target | Answers about what was left, how fast it had to go and at what cost |
What the numbers missed
The client was, in Avery's words, "heavily weighted towards autumn and winter trading", and JudeLuxe had taken over in March. By then the retailer had "stock sitting in a warehouse costing them money" and needed to move it.
At the start of the engagement the agency had been given a ROAS target. Avery noted that such a target is naturally an average across the account, and said the agency treated it as "basically the brief." It improved the metrics it was working towards. What it missed, he said, was that "they had stock that they needed to clear and cash was short in the bank", and that "the commercial output that was needed from the Google ad account had changed and we hadn't been proactive enough to figure out why it had changed."
The list of unasked questions was long. Nobody at the agency had asked what stock was left, what needed clearing, how quickly, what holding it was costing the business, or which trade-offs the business would accept. Avery offered a hypothetical: against a 4.5 ROAS, "would you have taken a three [ROAS] if it meant that we could clear all the stock". He was even-handed about the blame. "They hadn't told us, but equally we hadn't asked."
That point about averages has surfaced in other forms on PPC Land this year. A measurement guide from Funnel, published this autumn, worked through an example in which $1,000 of spend returns $3,000, a ROAS of 3.0, while the next $500 adds only $200 - a marginal return of $0.40 per dollar. An account can look healthy on its average and be wasteful at the margin, or, as in Avery's case, look efficient while failing at a different job altogether.
Google has also built mechanisms aimed at the kind of trade Avery described. Promotion mode, announced on June 15, 2026, is a beta for Search and Performance Max that runs scheduled windows of 3 to 14 days, adjusting ROAS tolerance and adding daily budget, with a built-in end date. Separately, Smart Bidding Exploration lets Target ROAS advertisers accept a target 5% to 30% lower on Search, Performance Max and Shopping campaigns, according to a Google help page first reported in October. Neither tool answers the question Avery says his agency failed to ask, which is whether the business wants that trade in the first place.
The "button pushing" agency
Avery was blunt about JudeLuxe at the time. "I would define us as what I would now call a button pushing agency," he said. "You would give us a budget and a target and we'd get to work. You know, campaigns, bids, search terms, spend."
The problem, as he framed it, was a narrow idea of the job. "So when they gave us that [ROAS] target, we didn't see the start of a conversation, we thought we'd been given the answer." The same request now gets a different response. When a prospect says it needs "a [ROAS] of five", Avery said, the first reply is now "based on what?", followed by questions about categories and acceptable trade-offs.
He also questioned the assumption, learned early, that profit is always the goal. Cutting inefficient spend and improving ROAS look like sound decisions, but "most ecom businesses die because they run out of money." The agency had believed it was ahead of the field because it tracked contribution alongside ROAS. "We probably thought we were being a bit more commercially sophisticated because we were looking at contribution as well as [ROAS]," he said. Changing the metric did not change the thinking: "you can misunderstand a business while looking at a profit metric just as you can looking at a revenue metric".
The client, to its credit, explained. According to Avery, it "essentially gave us a business 101, which was nice of them", walking the agency through how an ecommerce business makes and loses money. Adegbola's response was dry. "They thought you'd know."
Avery accepted the point. "If I was in the brand shoes, I would expect an ecom PPC agency to understand that as well," he said. He estimated that about 80% of JudeLuxe's time on client calls is now spent on matters other than Google Ads, because competence inside the platform is expected - "table stakes", in his phrase. The same applies to hiring. He no longer asks candidates how to amend a target ROAS; he asks "how do you calculate a sell through rate for a client".
How JudeLuxe's onboarding changed
Avery described the lesson as more than adding a line about stock to an onboarding document. According to his account, the process now runs in six stages:
- Challenge the objective. "Now we believe part of our responsibility as an agency is actually to challenge the objective that we're given," Avery said, "not in a hostile way". He likened the approach to a three-year-old repeating "But why?": why this target, why these products, why now, what happens if it is achieved, and what has to be sacrificed along the way.
- Hold the commercial conversation first. The agency tells prospects that "we need to have a commercial conversation here first before we can even think about talking about your platform whatsoever". The right channel might be Google Ads, affiliates or Meta, which JudeLuxe does not manage.
- Look at the catalogue, not the change history. Avery said his old habit was to open the account's change history and judge the previous agency by its activity - "another story where you confuse activity with progress". The agency now asks how many SKUs a client carries, how long they have been in stock, what their sell-through rates are, what the priorities are, and whether merchandising has over-ordered a particular style.
- Widen the room. Five or six years ago, he said, an agency dealt only with the marketing manager or the ecommerce manager. JudeLuxe now has more conversations with finance, merchandising, operations and ecommerce as a whole.
- Set targets after stock and trading priorities. The agency now asks "about stock and trading priorities" before any discussion of targets in the Google Ads account, he said.
- Revisit the targets. On audits, Avery said, he asks when a ROAS target was set. One answer he quoted was "Two years ago." His follow-up: "Did someone forget to let Google Ads know?"
Change history is not a complete record in any case. In October 2024, changes made by a Google Ads representative to a UK client's account did not appear in it at all.
Avery's view of paid search as a magnifier ran underneath all six stages. Because the channel is predominantly direct response, it amplifies the current state of the business: put 10, 20, 30, 40 or 50 thousand pounds more behind a business that is a mess, he said, and the result is only more mess.
Stock is a diminishing asset
Avery's phrase for inventory that ties up money is cash in cardboard. His position on ROAS has softened since the days when, he said, many of his posts attacked the metric. "[ROAS] still has a place," he said. When the job is releasing cash into the business, he would run campaigns on ROAS or revenue. When the job is profit, a contribution-based model may fit better. Chasing profit while cash sits in stock, he argued, means optimising towards the wrong metric.
The episode's introduction set out JudeLuxe's framework for that choice. The agency has built an approach it calls BOI, short for bid on intent, which Adegbola described as starting from one question: what does the business need this product to do?
| Job for a product (BOI) | Metric Avery associates with it in the episode |
|---|---|
| Generating contribution | A contribution-based model, when profit is the goal |
| Releasing cash from ageing stock | ROAS or revenue |
| Acquiring a valuable customer | Not specified in the episode |
| Protecting demand | Not specified in the episode |
"We always say stock is a diminishing asset," Avery said, using UK Soccer Shop, a JudeLuxe client that sells football shirts, as the example. New shirts arrive in May. "They've probably got a window of 3 to four months where they can sell the stock at full price." After that come discounts, then clearance, "and then it's a different target and nobody's letting Google ads know".
He has a word for this, one he admitted might not exist: dynamicism, meaning how quickly an ecommerce business changes. "You can ask that question today but in two months time the answer might be completely different," he said. Google, in his view, will not notice on its own. "Because Google isn't sophisticated enough to do that."
Adegbola made a similar point about the machine's inputs. "It's going to do what you ask it to do on the level of knowledge that you give it. And if that's basic knowledge, it's going to give you basic performance." Avery added "outdated knowledge" to the list, and put the problem in currency terms: "Google can sit there and it can treat £100 in one product the same as £100 in another." The commercial reality behind two products can be completely different. "But to Google £100 is £100."
A stale target is arguably more consequential now than when JudeLuxe lost its client. Under Google's August 17, 2026 bidding change, budget-limited Target CPA and Target ROAS campaigns that beat their targets are pulled back towards those targets. One advertiser said a Shopping campaign set at a 300% target ROAS lost sales after the change, a single account with no data published. For customer acquisition, one of the four BOI jobs, Google has its own lever: percentage multipliers from 1.1x to 5x on new customer values left beta in October.
Channels, halo effects and a sock seller in Stoke
Adegbola widened the discussion to channel roles, arguing that YouTube, Display and search each have a different job and cannot be judged by the same number. Avery agreed. "YouTube is very rarely going to be your highest performing platform," he said, and judged in isolation on ROAS, a YouTube campaign looks poor and gets switched off. What that view misses is the halo effect on brand uplift in search and Shopping campaigns. When clients cut spend elsewhere, Meta included, "PPC performance gets worse because nothing sits in isolation anymore." He hates the word funnel, he added, but the marketing funnel is now so vast that switching off one part can hamper every other.
Adegbola credited the idea, for her, to Kirk Williams, who runs the US agency ZATO, and to his book on PPC, Ponderings of a PPC Professional: a client that expects paid search alone to carry the business is in a difficult position however well the account is run. Williams won the Best Shopping Follow award at Marketing O'Clock's Clockscars in December 2025.
Avery's sharper target was LinkedIn. He said he was glad the platform had moved past a phase, roughly 12 months earlier, of downloadable playbooks promising things like "Nike's $30 billion PPC strategy". The audience for such material, in his telling, might be someone who sells "socks from a bedroom in Stoke" on a £10 a day Shopping budget, with none of a global brand's equity or spend across out of home, sponsorships, affiliates, search and social. Two brands in the same field, he said, can see their accounts perform completely differently.
Vendor data has made the cross-channel argument too. Fospha, working with Google, reported in May that across 25 retail brands, adding one Google channel was linked to 14% higher ROAS and adding two to 37%. Fospha sells measurement and the data is its own.
What audits turn up
Asked for the mistakes that still surprise him in new accounts, Avery went back to the period before JudeLuxe specialised in ecommerce, when it also handled lead generation and app installs. One of the first accounts he named had been built on a single broad match keyword.
| Account | What Avery says was found | Why it mattered, in his account |
|---|---|---|
| Small chiropractor in Guildford | The previous agency targeted only the town name, "Guildford", on broad match | Only the town name was targeted; the agency then refused to hand over access |
| Very large retailer | Products showing for the word "small" as a product title | A one-word title gives Google little to infer the product from |
| Retailer selling a chair | A product title with the word "mustard" in it six times | "Are you selling mustard or are you selling chair?" |
| Account where ROAS rose | Spend cut by 50%, leaving mainly branded traffic | A higher ROAS that reflects a narrower account, not better management |
| Account where conversion rate fell | Campaign mix moved from brand-only to non-brand | A drop that does not necessarily mean the website needs work |
The chiropractor account came with a coda. Before handing it over, the outgoing agency held it, in Avery's word, "ransom", saying: "It's our intellectual property." His reply: "There is nothing intellectual in this account if you have the word [Guildford] on broad match." The agency eventually released it. Broad match has itself changed since then: Search campaigns using the campaign-level broad match setting faced an automatic upgrade to AI Max from September 1, 2026. Handover disputes have not gone away either; in March, a former agency closing its own Merchant Center account wiped a client's account, including feeds, feed rules and links to Google Ads.
The feed as the last lever
The product titles led Avery to his main complaint, the product feed. With Google having taken away a lot of control, "your product feed is probably your biggest lever that you have when it comes to ecom and it is probably the biggest neglected part." Google infers what a product is and which searches it should be eligible for from the feed, so a neglected feed sets a low ceiling on everything else. "Why are they not doing feed optimization?" he asked of agencies that skip it. "Because that's arguably 50% of the job."
The feed was also why JudeLuxe specialised. When the agency's work was roughly 50/50 between lead generation and ecommerce, Avery said, control in lead generation was being taken away, while the feed remained something the agency could still shape. His shorthand for the principle: "control the inputs to maximize the outputs". The best account structure, the most bid adjustments and every negative keyword, he said, make no difference if products are not appearing for the right searches at the right time.
Google has been adding to what a feed can carry. In May, a practitioner guide set out eight Merchant Center attributes tied to AI Mode and AI Overviews, including product highlights of 1 to 150 characters and up to 30 question-and-answer pairs per product. Title rules have tightened as well: since October 2025, only a manufacturer may put its own brand name in the short title attribute.
A 500% budget increase for Notts County
The final segment turned to AI. "I am definitely not anti-AI," Avery said. He uses it all day and called it "fantastic" for analysis, able to work through a spreadsheet "100 times quicker" than he could manually. But "AI has enabled people to make bigger mistakes than they were able to make before," and it has given people "the ability to produce certainty that they haven't earned."
His concern centred on junior staff. Practitioners who adjusted bids by hand for years can tell when an answer looks wrong. "An answer can be coherent, detailed," he said, but "it can be completely unsuitable for the situation."
His example came from UK Soccer Shop. The client had an independent audit carried out on the account, and one recommendation was "to increase budget on the campaign that contained just products for [Notts] County by about 500%." Avery said he believed the club was in League Two and did not have many fans, and asked "how many people do you think are searching for [Notts] County products on a Tuesday night in the UK or in Australia or wherever". On face value, the campaign's ROAS was "absolutely smashing", which is why the recommendation looked sensible. The episode does not say who produced the audit or whether it was generated by an AI tool; Avery raised it as an example of a recommendation made without context about the products.
The same risk applies when he uses AI himself. Asked to review a campaign without knowing the stock position or trading priorities, a model will come back with a "plausibly confident answer that could be factually incorrect based on the context of that business." His summary, delivered with a laugh, was that AI lets people be idiotic quicker. Independent testing has found the problem at scale: a WordStream study found about one in five AI answers to PPC questions inaccurate, with Google's AI Overviews at 26%.
Judgment, and who gets fired
Adegbola recalled moderating an event at which Frederick Vallaeys, a co-founder of Optmyzr, relayed what others in the industry had told him: that "AI is just going to make a lot more human scapegoats". When an AI recommendation goes wrong, she said, the person who acted on it is the one who loses the job.
Avery put it in commercial terms. "When you're an agency, you're paid for your judgment. And if you outsource your judgment to an AI, well, what are you being paid for?" Clients now bring AI recommendations to JudeLuxe, and the agency's first step is to examine where the conclusion came from and whether the tool had the information it needed. Ideas that survive scrutiny get actioned; those that do not need a stated reason for being dismissed. The reverse failure also exists: an agency whose only answer to a reasonable question is "trust us, we're the experts" is not good enough either.
Agencies using AI, he argued, should be held to "a certain standard, probably a higher one", because they are paid for the judgment of where the next pound is best spent - a Shopping campaign, Performance Max, or a particular product. He is comfortable letting AI tell him which products belong in which campaign, given the right context. What he is not confident in is "allowing it to make decisions on our behalf". That may change "in 6, 12, 18 months as new models develop", he said, but for now the judgment has to come from a person.
The wider industry has been drawing similar lines. Reporting in January found the ad industry letting language models handle planning, setup, reporting and analysis, but not autonomous bidding or budget spend. At the other end, OpenAI's head of monetisation, Asad Awan, described a future in February in which businesses prompt ChatGPT to create and run ads instead of hiring performance marketers.
Why this matters
Avery's account is one agency's recollection of one client, told on a community podcast, with no figures beyond an illustrative 4.5 versus 3 ROAS. Its value lies in the specificity of the failure. The account was not mismanaged by the usual measures; the measures were wrong for the moment. For retailers, the episode is a reminder that a ROAS target is a decision someone made at a particular point in the trading year. For agencies, it describes a shift PPC Land has tracked across 2026, as Google adds settings that move targets temporarily, automates more of the account and leaves the feed and the commercial brief as the main remaining inputs. Who sets those inputs, and on what information, increasingly decides the outcome.
The episode leaves several things open. The client and its sector are not named, and neither are the stock value, the cash position, the original ROAS target or the results achieved in the two months. The year is not given; Avery has been a director and co-founder of JudeLuxe since December 2021, according to his LinkedIn profile. The episode does not say who carried out the UK Soccer Shop audit.
The show and the guest
PPC Live The Podcast is run by PPC Live, the UK paid search community Adegbola founded; she is also paid media editor at Search Engine Land. JudeLuxe manages Google Ads and Microsoft Ads for established ecommerce brands, and the episode named Thermos, Triumph, UK Soccer Shop, REFY and Closure LDN among its clients. Avery writes regularly for PPC Live; Adegbola cited his recent article "Google Ads customer acquisition. Are we bidding on profit or hope?", and said a talk at the most recent PPC Live event had covered bidding profitably.
Avery said JudeLuxe was up for 10 awards at the UK Search Awards in November. His advice for anyone who has just lost a client was the line Adegbola chose to close on: "control what you can control", because "tomorrow's a fresh day". Asked to name his career as a film, he picked Moneyball, a story "about questioning which numbers actually help you win."
In her closing remarks, Adegbola said PPC Live had held an event "last week on the 21st", that an online version would follow in a few weeks, and that the next in-person event is scheduled for February 4, 2027.
Episode chapters
| Time | Chapter |
|---|---|
| 0:00 | Meet Chris: commercial thinking beyond Google Ads |
| 4:11 | The client loss behind the lesson |
| 6:39 | Why higher ROAS missed the stock and cash problem |
| 10:35 | Profit is not the whole commercial picture |
| 14:10 | Why channels need different success measures |
| 18:28 | Changing onboarding: question the objective |
| 21:25 | Talk to finance, merchandising, and operations |
| 24:46 | Targets must move with ecommerce conditions |
| 27:50 | Give Google Ads the right business context |
| 31:21 | How to respond when a client leaves |
| 33:29 | PPC horror stories and the product-feed blind spot |
| 40:36 | AI recommendations need human scrutiny |
| 48:27 | Moneyball and the numbers that help you win |
Key facts
- Episode: We Smashed Our PPC Targets... So Why Did the Client Fire Us? ft Chris Avery
- Series: PPC Live The Podcast - The F*ck ups that led to Triumphs
- Channel: PPC Live (YouTube)
- Published: November 19, 2026
- URL: https://www.youtube.com/watch?v=dZdxzZZ-9dw
- Length: about 53 minutes, 13 chapters
- Host: Anu Adegbola, founder of PPC Live and paid media editor at Search Engine Land
- Guest: Chris Avery, director and co-founder of JudeLuxe (Google Ads and Microsoft Ads for ecommerce brands)
- Clients named in the episode: Thermos, Triumph, UK Soccer Shop, REFY, Closure LDN
- JudeLuxe framework: BOI (bid on intent)
- The lost client: unnamed retailer, heavy autumn and winter trading, taken over in March, fired after two months, "four or five years" before the episode
- Audit example: recommended budget increase of about 500% for a campaign of Notts County products (UK Soccer Shop)
- Next PPC Live in-person event: February 4, 2027 (scheduled)
Timeline
- December 2021 - Chris Avery becomes a director and co-founder of JudeLuxe.
- October 15, 2024 - A Google Ads representative changes a UK client's account without authorization; the changes do not appear in change history.
- July 10, 2025 - WordStream finds about 20% of AI answers to 45 PPC questions inaccurate.
- October 3, 2025 - Google limits brand names in the short title attribute to manufacturers.
- December 29, 2025 - Kirk Williams wins Best Shopping Follow at the Marketing O'Clock Clockscars.
- January 14, 2026 - Reporting finds the ad industry keeps LLMs away from autonomous bidding and budgets.
- February 9, 2026 - OpenAI's Asad Awan describes businesses prompting ChatGPT for ads instead of hiring performance marketers.
- March 6, 2026 - A former agency's account closure wipes a client's Merchant Center account.
- May 16, 2026 - Fospha and Google report ROAS gains for 25 retail brands adding Google channels.
- May 24, 2026 - A guide sets out eight Merchant Center attributes tied to AI Mode.
- June 15, 2026 - Google announces promotion mode and the August 17 bidding change.
- August 5, 2026 - Google tells advertisers that campaign-level broad match will be upgraded to AI Max.
- August 17, 2026 - Google's bidding change pulls budget-limited campaigns that beat their targets back towards them.
- September 1, 2026 - The AI Max upgrade for broad match campaigns begins.
- October 5, 2026 - An advertiser reports lost Shopping sales at a 300% target ROAS after the August 17 change.
- October 22, 2026 - Google's 1.1x to 5x new customer value multipliers leave beta.
- October 22, 2026 - Smart Bidding Exploration's 5% to 30% lower ROAS targets are first reported.
- November 9, 2026 - A Funnel guide shows a 3.0 ROAS hiding a $0.40 marginal return.
- November 19, 2026 - PPC Live publishes the Chris Avery episode.
- February 4, 2027 - Next PPC Live in-person event (scheduled).
Related PPC Land coverage
- Funnel guide shows how a 3.0 ROAS can hide $0.40 returns on extra spend - A 96-page measurement guide on why average returns conceal marginal ones.
- Google lets Shopping and PMax campaigns lower ROAS targets by up to 30% - Smart Bidding Exploration's 5% to 30% target reduction for Search, Performance Max and Shopping.
- Promotion mode is here - Google's Ginny Marvin explains what actually changed - Google's June 2026 package, including 3 to 14 day promotion windows.
- Google optimises its own revenue. Advertisers have to optimise theirs - An analysis of the August 17 bidding change and its effect on campaigns beating their targets.
- Advertiser says Google's August 17 change cut Shopping sales at 300% tROAS - One advertiser's account of lost Shopping sales after the change.
- Google Ads lets retailers multiply new customer values up to 5x - New customer value multipliers leave beta across Search, Shopping, PMax and Demand Gen.
- The Google channel mix secret that doubled ROAS for 25 ecommerce brands - Fospha's vendor data on adding Google channels.
- 8 Google Merchant Center attributes your feed needs for AI Mode - Feed attributes Google ties to AI Mode and AI Overviews, with their limits.
- Google clarifies brand usage in Shopping product titles - The October 2025 product title policy update.
- Old agency closes client's Google Merchant Center - what it wipes and how to recover - A handover that deleted a client's feeds and account links.
- Google Ads broad match campaigns face AI Max auto-upgrade on September 1 - The automatic move of broad match campaigns to AI Max for Search.
- Google Ads Rep's unauthorized account changes spark industry concern - Changes to a client account that did not show in change history.
- One in five AI responses for PPC strategy contain inaccuracies, study finds - WordStream's test of five AI tools on 45 PPC questions.
- The ad industry won't hand LLMs 'the keys' to spend budgets - Where the industry draws the line on AI control of spend.
- OpenAI exec wants advertisers to ditch agencies and just prompt ChatGPT for ads - OpenAI's view of ads without performance marketers, and the pushback.
- Marketing O'Clock crowns industry leaders in 2025 Clockscars ceremony - The 2025 awards, including Kirk Williams for Best Shopping Follow.
Summary
Who: Chris Avery, director and co-founder of the UK ecommerce PPC agency JudeLuxe, interviewed by Anu Adegbola, founder of PPC Live and paid media editor at Search Engine Land.
What: Avery described being fired by an ecommerce client two months into the relationship despite higher ROAS, higher contribution and lower spend, because the client needed to clear seasonal stock and raise cash, and explained how JudeLuxe changed its onboarding, its view of ROAS, its focus on product feeds and its stance on AI recommendations.
When: The episode was published on November 19, 2026; the client loss happened "four or five years" earlier, in Avery's estimate, after a March takeover.
Where: On PPC Live The Podcast, published on PPC Live's YouTube channel.
Why: The story shows how a platform target set without commercial context can mislead an agency and an automated bidding system alike, at a time when Google is adding temporary target controls, automating more of the account and leaving the feed and the brief as the main inputs advertisers still control.
Discussion