Case Study
The Rules Changed Before the Numbers Did.
Why I advised a Canadian immigration consultancy to cut its Meta spend by roughly 85% in the month its dashboard looked the best it ever had.
- Role
- Marketing Consultant (external): built the account in 2024, advisory from 2025
- Period
- April 2024 – 2025
- Market
- Canada, young temporary residents already in the country
Jump to:The Short AnswerThe SituationWhat ChangedWhat the Numbers SaidThe DecisionWhy It WorkedWhat We Don't KnowHow to Apply ThisEvidence & Limits
The short answer
In 2024 I built a Meta click-to-WhatsApp account for Manni Immigration, a Brampton-based consultancy whose clients were young temporary residents already in Canada. Across April to December it spent roughly CAD 101,000 and started about 4,300 WhatsApp conversations.
By December 2024 the dashboard looked better than it ever had. Cost per conversation had roughly halved since August, from about CAD 32 to about CAD 15, and December was the best month on record. Everything Meta was showing said scale.
I advised the opposite. Between September 2024 and January 2025, Canada tightened the rules that decided who could stay and work in the country, and the people our ads reached were exactly the people those rules hit. So in January 2025 spend was cut by roughly 85%, executed by the owner on my recommendation. The account kept running lean, and in my observation the leads that came through afterwards were fewer but far more likely to be eligible.
By December 2024 the dashboard looked better than it ever had. Cost per conversation had roughly halved since August, from about CAD 32 to about CAD 15, and December was the best month on record. Everything Meta was showing said scale.
I advised the opposite. Between September 2024 and January 2025, Canada tightened the rules that decided who could stay and work in the country, and the people our ads reached were exactly the people those rules hit. So in January 2025 spend was cut by roughly 85%, executed by the owner on my recommendation. The account kept running lean, and in my observation the leads that came through afterwards were fewer but far more likely to be eligible.
The transferable principle
In a regulated market, your dashboard measures how many people want a conversation, not how many are allowed to become clients. When the rules change, the dashboard is the last to know.
The situation
Manni Immigration Services is an RCIC-led immigration consultancy in Brampton, Ontario. Its core audience was not people abroad dreaming about Canada. It was people already here: mostly 18 to 28, on study permits or post-graduation work permits, weighing a college admission, a master's, or what to do before a permit expired.
I came in as an external marketing consultant in April 2024 and built the account: click-to-WhatsApp campaigns routed into the client's Zoko inbox, ad sets optimised for conversations rather than clicks or engagement, lead forms, and the tracking around them. The unit that mattered was a WhatsApp conversation with a real person, because that is where an immigration consult actually starts.
We tested the usual levers hard: province, audience, and pitch. Ontario, British Columbia and Alberta college audiences; a low-IELTS angle; the Rural and Northern Immigration Pilot; and, tellingly, one of the very first ad sets was simply called "New Changes". Policy was never background noise in this niche. It was the product.
I came in as an external marketing consultant in April 2024 and built the account: click-to-WhatsApp campaigns routed into the client's Zoko inbox, ad sets optimised for conversations rather than clicks or engagement, lead forms, and the tracking around them. The unit that mattered was a WhatsApp conversation with a real person, because that is where an immigration consult actually starts.
We tested the usual levers hard: province, audience, and pitch. Ontario, British Columbia and Alberta college audiences; a low-IELTS angle; the Rural and Northern Immigration Pilot; and, tellingly, one of the very first ad sets was simply called "New Changes". Policy was never background noise in this niche. It was the product.
Ontario colleges, 18-28
BC colleges
Alberta colleges
Low IELTS angle
RNIP pilot
"New Changes"
What none of them could varyWho is eligible to apply, which is set by IRCC, not by the ad account
Every ad set sat on the same constant. When IRCC moved it, all of them moved together, and no amount of testing inside the account could have hedged that.
What changed
Through the second half of 2024, Canada moved to reduce the number of temporary residents in the country, and the changes landed squarely on our audience:
September 2024: new limits on low-wage LMIA work permits, narrowing one route young workers used to stay.
November 2024: a field-of-study requirement for post-graduation work permits for many college programs, changing which admissions still led anywhere.
21 January 2025: spousal open work permits restricted for most workers, removing an option many of our enquirers had been planning around.
None of these made Manni's services less wanted. People were more anxious than ever and had more questions than ever. What they changed was how many of those people could still be helped.
September 2024: new limits on low-wage LMIA work permits, narrowing one route young workers used to stay.
November 2024: a field-of-study requirement for post-graduation work permits for many college programs, changing which admissions still led anywhere.
21 January 2025: spousal open work permits restricted for most workers, removing an option many of our enquirers had been planning around.
None of these made Manni's services less wanted. People were more anxious than ever and had more questions than ever. What they changed was how many of those people could still be helped.
What the numbers said
Here is the part that makes this worth writing up. While the rules tightened, the account's cost per WhatsApp conversation did not get worse. It got better, month after month:
- May to August 2024: roughly CAD 29 to 32 per conversation, on CAD 9,500 to 13,800 a month.
- September 2024: about CAD 24, with 578 conversations.
- October and November 2024: about CAD 23, then about CAD 19.
- December 2024: about CAD 15, and 651 conversations, the best month the account ever had.
- January 2025: about CAD 13 on the ad sets still running, as spend came down.
The decision
A dashboard like that is an invitation to scale. Cheaper conversations, more of them, a clean downward trend. If I had only been looking at Meta, I would have recommended more budget, not less.
Instead I was watching three signals that sit upstream of the dashboard. The policy announcements themselves, as they were published. What the client's front line was hearing in real consultations. And what people were actually telling us in chats and lead forms, where more and more of them described situations the new rules had closed off. No single one of those was proof. All three pointing the same way, at the same time, was enough.
So in January 2025 I advised cutting spend by roughly 85%, from around CAD 10,000 a month to a lean CAD 1,000 to 2,000, and keeping the account alive rather than switching it off. By then I had moved from running the account day to day into an advisory role, so the owner made the change himself in Ads Manager. There was no pushback and no drama. When the people closest to the clients and the person reading the market see the same thing, the decision is quick.
Instead I was watching three signals that sit upstream of the dashboard. The policy announcements themselves, as they were published. What the client's front line was hearing in real consultations. And what people were actually telling us in chats and lead forms, where more and more of them described situations the new rules had closed off. No single one of those was proof. All three pointing the same way, at the same time, was enough.
So in January 2025 I advised cutting spend by roughly 85%, from around CAD 10,000 a month to a lean CAD 1,000 to 2,000, and keeping the account alive rather than switching it off. By then I had moved from running the account day to day into an advisory role, so the owner made the change himself in Ads Manager. There was no pushback and no drama. When the people closest to the clients and the person reading the market see the same thing, the decision is quick.
Early signals, upstream of the dashboard
Policy announcementsIRCC changes, Sep 2024 to Jan 2025
Client front lineWhat real consultations were revealing
Chat and lead-form answersMore people the rules had closed off
What the dashboard said
Cost per conversation fallingAbout CAD 32 in Aug to about CAD 15 in Dec 2024
Best month on record651 WhatsApp conversations in Dec 2024
The decision
Cut spend roughly 85%January 2025My recommendation
Executed in Ads ManagerBy the ownerClient
What followed
Lean always-on accountCAD 1,000 to 2,000 a month
Fewer, better-qualified leadsMy observation, see limits
Why it worked
Meta goes out and buys whatever result you ask it for. We asked for WhatsApp conversations, and it got very good at finding people who would start one. What it cannot see is whether that person is still allowed to use the service. Eligibility lives in immigration law, not in the ad auction.
That is why the numbers improved as the market shrank. My read, which the ad data alone cannot prove, is that the policy news itself drove the drop: more worried people were messaging with questions, conversations got cheaper to start, and a growing share of them came from people the new rules had already ruled out. The dashboard was counting anxiety as demand.
Spending more into that would not have bought more clients. It would have bought more conversations with people who could not become clients, at a cost the dashboard would have reported as efficient. Cutting back kept the budget pointed at the narrower group who could still act, and kept the account warm for whatever came next instead of abandoning a channel that had worked.
That is why the numbers improved as the market shrank. My read, which the ad data alone cannot prove, is that the policy news itself drove the drop: more worried people were messaging with questions, conversations got cheaper to start, and a growing share of them came from people the new rules had already ruled out. The dashboard was counting anxiety as demand.
Spending more into that would not have bought more clients. It would have bought more conversations with people who could not become clients, at a cost the dashboard would have reported as efficient. Cutting back kept the budget pointed at the narrower group who could still act, and kept the account warm for whatever came next instead of abandoning a channel that had worked.
The second principle
A falling cost per result is only good news if the thing you are counting is still the thing you are selling.
What we don't know
I would rather mark the edges of this clearly than let the story do more work than the evidence can.
I cannot prove the leads got better with a number. The lead sheet that tracked qualification is no longer accessible to me now that the engagement has ended. "Fewer but better" is my observation from the time, not a measured figure.
The months right after the cut are hard to measure. From February to June 2025 the account ran mainly on boosted posts and message campaigns that do not report conversations the same way, so there is no clean like-for-like comparison for that stretch.
I cannot separate why conversations got cheaper. Policy-driven anxiety is my best explanation, but creative fatigue, seasonality and audience learning could all have contributed, and the data does not split them.
There is no counterfactual. I cannot show what would have happened had the account kept spending CAD 10,000 a month. I believe it would have bought mostly ineligible conversations, but that is a judgement, not a result.
Later results are not all mine. In the second half of 2025 the owner's own campaigns ran at roughly CAD 4 to 6 per conversation. My involvement ended during that period, so I am not claiming them.
I cannot prove the leads got better with a number. The lead sheet that tracked qualification is no longer accessible to me now that the engagement has ended. "Fewer but better" is my observation from the time, not a measured figure.
The months right after the cut are hard to measure. From February to June 2025 the account ran mainly on boosted posts and message campaigns that do not report conversations the same way, so there is no clean like-for-like comparison for that stretch.
I cannot separate why conversations got cheaper. Policy-driven anxiety is my best explanation, but creative fatigue, seasonality and audience learning could all have contributed, and the data does not split them.
There is no counterfactual. I cannot show what would have happened had the account kept spending CAD 10,000 a month. I believe it would have bought mostly ineligible conversations, but that is a judgement, not a result.
Later results are not all mine. In the second half of 2025 the owner's own campaigns ran at roughly CAD 4 to 6 per conversation. My involvement ended during that period, so I am not claiming them.
How to apply this
What transfers out of this account into any regulated niche, immigration, lending, healthcare, education:
- Keep a regulation calendar next to your ads calendar. In regulated markets, announcements and effective dates are leading indicators. Your dashboard is a lagging one.
- Measure eligibility, not just enquiries. Put one qualifying question into the WhatsApp flow or lead form, so the metric you optimise can tell a curious person from a qualified one.
- Treat a sudden efficiency gain in a shrinking market as a question. Cheaper results while the rules tighten usually means you are reaching more people who cannot buy, not that the ads got smarter.
- Listen to the front line. The people taking consultations hear the market change weeks before it shows up in cost per result.
- Cut fast, but do not switch off. A lean, always-on account keeps the learning and the audience warm, so you can scale back up when the rules settle.
- Be clear about who decides and who executes. A consultant's job is the call and the reasoning. Saying so plainly keeps the case, and the relationship, honest.
Evidence and limits
I worked with Manni as an external marketing consultant from April 2024: building and running the account through 2024, then in an advisory role from early 2025 until the engagement ended in 2025. The account's activity log shows the January 2025 budget changes were made by the owner, which matches that split.
The spend and conversation figures are measured. They come directly from the Manni Meta ad account: monthly spend and "messaging conversations started" on the conversation-optimised ad sets, April 2024 to January 2025. Figures are rounded.
The lead-quality claim is my observation, not a measurement, for the reason given above. The policy dates are public and are linked below. Spend by client is not otherwise published.
The spend and conversation figures are measured. They come directly from the Manni Meta ad account: monthly spend and "messaging conversations started" on the conversation-optimised ad sets, April 2024 to January 2025. Figures are rounded.
The lead-quality claim is my observation, not a measurement, for the reason given above. The policy dates are public and are linked below. Spend by client is not otherwise published.
What can be independently checked:
