Stop the investment when the business case fails—not when patience does
An owner rarely asks this question after a good month. Spend has continued, the dashboard reports activity and the sales or ecommerce result does not justify the cash leaving the business. One adviser says the system needs more data. Another says to change the creative. A third proposes a larger budget.
My verdict is: paid advertising deserves neither infinite patience nor a panic shutdown. Continue only while there is a credible, bounded route from the current evidence to suitable customers and contribution. Pause or stop when the commercial case, customer promise or risk boundary fails.
This intent is distinct from asking how long paid advertising takes to work, which concerns evidence windows, and from deciding whether to cut marketing when sales slow, which concerns a broader budget response. Here, the owner is deciding whether paid acquisition itself still deserves capital.
The Paid Advertising Exit Test
I use six gates before recommending continue, narrow, repair, pause or exit. A failed campaign metric is a symptom; the gates identify whether the acquisition model or another part of the system has failed.
What business result was paid meant to create?
Name the suitable customer, qualified opportunity, purchase or contribution target. Traffic and form volume are not complete outcomes.
Can the measurement be trusted?
Reconcile platform conversions with unique leads, sales records, refunds, customer status and finance before judging the channel.
Does the market accept the proposition?
Inspect customer objections, price, proof, differentiation and genuine buying behaviour—not only advertising engagement.
Where does the customer journey break?
Separate unsuitable reach, message failure, website friction, weak follow-up and delivery constraints. Each needs a different decision.
Can a suitable customer repay acquisition?
Use contribution after variable costs, sales effort, cancellations and the real time to cash. Do not borrow a universal ratio.
What is the next-best use of capital?
Compare continued learning with fixing the product, offer, sales process, retention, organic discovery or cash position.
The first two gates protect against false exits. A campaign optimised for easy form submissions may look successful while sales rejects most leads. The same campaign may look unsuccessful when qualified outcomes happen offline but never return to the reporting system. Google distinguishes primary conversion actions used for bidding from secondary actions used mainly for observation; the wrong primary signal can teach the system to pursue the wrong behaviour.
The next three gates protect against false repairs. Paid media cannot make an unwanted offer profitable, repair a sales team that never follows up or create margin that the product economics do not contain. The final gate forces an opportunity-cost decision: continuing can be rational when the next bounded test is more valuable than any alternative, even before the channel reaches profitability.
Judge paid acquisition from customers back to spend
Start with accepted customers or qualified opportunities, then work backwards through sales, conversion and channel evidence. For lead-generation businesses, Google allows qualified-lead and converted-lead stages to be defined from the company's own sales process and connected with offline outcomes. That does not solve attribution by itself, but it is stronger than asking a platform to optimise every form submission equally.
Use finance, ecommerce, CRM, refunds, cancellations and sales acceptance as the commercial record.
Inspect audience fit, message, page, sales response, fulfilment and the normal buying delay.
Separate captured demand, assisted journeys and genuinely new suitable customers where the evidence allows.
Do not use a universal lifetime-value-to-acquisition-cost ratio as an automatic stop rule. Subscription, ecommerce, professional-service and high-value B2B models have different gross margins, cash timing, service costs and retention uncertainty. Build the threshold from the company's own contribution and cash exposure.
Set an illustrative loss boundary before the next test—for example, a maximum additional investment and decision date. The amount must come from the business's cash position and value of the unanswered question, not this article. Label forecasts and scenario assumptions clearly; do not present them as proven channel performance.
Choose continue, narrow, repair, pause or exit
| Evidence pattern | Decision | Owner action |
|---|---|---|
| Suitable customers and contribution remain credible; volatility is understood; the next test has a specific question | Continue | Fund the bounded learning cycle and protect the economic ceiling |
| One segment, offer, market or campaign produces value while broader activity destroys it | Narrow | Protect the viable pocket and stop subsidising the weak expansion |
| Demand exists, but measurement, message, page or follow-up creates an identifiable break | Repair | Hold scale, fix the earliest material failure and retest one change |
| Cash, capacity, stock, product availability, claims or compliance make continued delivery unsafe | Pause | Stop exposure, preserve evidence and define explicit restart conditions |
| Trustworthy evidence rejects the economics or customer fit after the important controllable tests | Exit | Reallocate capital and document what would need to change before any future test |
Pause is not the same as deleting the evidence. Google states that pausing stops ads and further charges while keeping settings and history available; removing a campaign is permanent and cannot be resumed. Meta similarly provides campaign-level on/off controls. Preserve account ownership, change history, exports, landing pages, creative learning and downstream customer outcomes.
Stop immediately when the advertisement makes unsupported claims, promotes an unavailable offer or creates material customer or regulatory risk. The U.S. Federal Trade Commission states that advertising claims must be truthful, non-deceptive and evidence-based. A performance review cannot justify continuing a claim the business cannot support.
A 30-day paid-advertising exit decision
Reconcile
Match spend and platform actions to unique leads, accepted opportunities, customers, contribution and the normal outcome delay.
Locate
Apply the six gates. Identify the earliest material break and separate observed facts from hypotheses and missing data.
Test or protect
Run one bounded repair using a stable control where safe, or pause immediately when the cash, promise or compliance boundary fails.
Decide
Continue, narrow, repair, pause or exit. Record the evidence, owner, capital movement and the conditions that would reverse the decision.
Google Ads custom experiments can compare a change with the original campaign using a defined share of traffic and budget. They are useful when the question is narrow and the outcome is configured correctly. They are not a reason to keep spending when the product economics, customer promise or cash boundary has already failed.
If the account may simply be too early, use the Paid Advertising Readiness Test. If acquisition cost is rising, use the CAC Pressure Map. If one paid route has become a business-continuity risk, use the Channel Resilience Test. Review growth services, case evidence, evidence standards, Thomas's operator model and the diagnostic route before deciding.
Sources and evidence notes
Sources and current search results were checked on 28 August 2026. Search priority is qualitative; no unverified keyword volume, universal profitability ratio or minimum campaign duration is claimed. The Paid Advertising Exit Test, three-layer evidence model, decision matrix and 30-day process are original ThomPerformance practitioner analysis. The loss-boundary example is illustrative planning guidance, not a benchmark or client result.
Frequently asked questions
How do I know when to stop paid advertising?
Stop when trustworthy evidence shows no credible path from paid demand to profitable customers, the main controllable failures have been tested, and the capital has a stronger use elsewhere. Pause sooner when claims, compliance, cash exposure or customer harm create unacceptable risk. A disappointing week or expensive click is not enough evidence by itself.
Should I pause paid ads or remove the campaigns?
Pause them when the decision may be reversible. Google states that pausing stops delivery and further charges while preserving settings and history; removing a campaign is permanent and it cannot be resumed. Preserve access, exports, change history and commercial outcomes before ending any provider relationship or restructuring the account.
How long should paid ads run before deciding they failed?
There is no universal duration. The decision window must cover enough qualified customer outcomes for the business model, including the normal sales or repeat-purchase delay. Set the loss boundary and evidence requirement before launch. A low-volume B2B sale needs a different review window from an ecommerce purchase with immediate contribution data.
Can paid ads be unprofitable at first and still be worth continuing?
Yes, when the early loss is bounded, the business is buying useful evidence, and leading indicators connect credibly to later customer value. Continue only with a named learning question and stop rule. Do not call open-ended losses an investment when the offer, measurement or commercial outcome remains undefined.
What should a business do after stopping paid ads?
Protect customer and campaign evidence, document why the decision was made and fund the constraint that now matters most. That may be the offer, product, website, sales follow-up, retention, organic discovery or cash position. Define the conditions for a future retest so stopping does not become either permanent avoidance or an automatic relaunch.
Stop buying activity when the investment case has ended
Paid advertising should continue while it creates credible commercial value or bounded learning that the business cannot buy more effectively elsewhere. It should narrow, pause or stop when trustworthy evidence rejects that case. Preserve the learning, fund the real constraint and define what would have to change before paid acquisition deserves another test.
Which fact would change the decision today: customer quality, measurement, the offer, conversion, economics or the next-best use of cash?