Homa · Paid Growth

Why lead retention jumped: the Meta LAB loop

TL;DR

Weekly lead retention jumped from a ~10% baseline to 27% because of one channel: Meta. Meta-acquired leads came back in week two at 46% (cohort of Jul 20) - roughly 4x every other channel - and Meta now brings in ~950 new property-viewing leads per week, up from ~134.

The mechanism is a closed loop, working as designed: prospecting and dynamic ads acquire a first property view, the pixel instantly builds a "viewed homes" audience, and dynamic retargeting pulls the same people back with the homes they looked at. 96% of the returns came through another Meta ad click - and once back, people genuinely browse (median 5 pages).

The catch: these leads browse but do not sign up yet - ~2% registration vs 15% for other traffic. The loop builds a large, warm, repeatedly-engaged audience; the next job is converting it.

Week-1 return rate, Meta leads (wk of Jul 20)
46%
vs 9-13% for all other channels
New property-viewing leads from Meta / week
950
was ~134 the week before launch
Returns driven by another Meta ad click
96%
410 of 429 returners; only 19 came back on their own
Registration rate of Meta leads
2%
vs 15% for untagged traffic - the gap to close

1The jump, isolated to Meta

Each week's cohort = people viewing their first property page on app.tryhoma.com that week. "Week-1 return" = came back to the app the following week. Attribution is event-level first touch.

Week-1 return rate by cohort week
Meta bars start Jul 6 - before that Meta acquired fewer than 20 leads/week (too small to rate). Hover for cohort sizes.
Meta first touch All other channels
View data table
Cohort weekMeta leadsMeta W1 returnOther leadsOther W1 return
Jun 86-1,26610.5%
Jun 150-1,0808.9%
Jun 226-7999.6%
Jun 298-69113.2%
Jul 61533.3%93411.7%
Jul 1313433.6%1,29610.7%
Jul 2092946.2%1,27112.5%
Jul 2798125.8%1,3598.8%
New property-viewing leads per week, by source
Meta went from a rounding error to roughly 42% of all new leads in two weeks. Other channels stayed flat - Meta's volume is additive, not cannibalizing.
Meta first touch All other channels
What changed: the Meta LAB campaign fleet first shows traffic Jul 16 (prospecting, retargeting, first-party) and Jul 19 (dynamic ads "ddcold" / "ddwarm", TX and FL intent, moneyflow and the other edge concepts). The week of Jul 20 was its first full week at scale - the timing matches the jump exactly.

2The mechanism: a retargeting loop, not organic pull

We checked how the Jul 20 Meta cohort's 429 returners actually came back.

How Meta leads returned in week 2
Share of the 429 returners, by how their return visit arrived.
Returned via another Meta ad click
410 people · 95.6%
Returned on their own (direct / organic)
19 people · 4.4%
410 came back through another Meta ad click (95.6%) 19 returned on their own

This is the important nuance: the retention jump is manufactured by retargeting - and that is fine, it is what the system was built to do. Ads acquire a first property view; the ViewContent pixel puts that person into the viewed-homes audiences the same day; dynamic ads then re-serve them the homes they browsed. What matters is that the re-pull works and people engage when they land: median 5 pages, average 10 pages in their return week.

How fast the loop pulls people back
Days from first visit to first return, Jul 20 Meta cohort (460 returners within two weeks). The re-pull peaks at days 3-7.
Which campaigns drive the return clicks
Campaign tagged on the return visit (Jul 20 cohort returners). Dynamic warm ads - the "here are the homes you looked at" layer - do the heaviest lifting.

3Campaign scorecard: acquisition and re-engagement

New leads acquired and their week-1 return rate, by acquiring campaign. Two cohort weeks shown side by side.

CampaignJul 20: leadsW1 returnJul 27: leadsW1 return
moneyflow6862%9928%
ddwarm (dynamic, warm)15057%22924%
ddcold (dynamic, cold)16252%13417%
prospecting14250%19131%
firstparty2846%1-
conquest12-3842%
tx-intent9538%4624%
retargeting7137%4920%
fl-intent9036%13526%
Reading notes

Leads = first-ever property page view attributed to the campaign that week (all devices). Campaign names shortened from "homa-lab-meta-*-v1". Cells with fewer than 20 leads are left unrated. For context, non-Meta channels return at 9-13% week over week.

Watch the decay: the Jul 20 spike (46%) cooled to 26% for the Jul 27 cohort. ddcold specifically fell from 52% to 17% week-1 return. Still 2-3x the baseline, but the first week's number was a launch-week high, not the steady state. Worth tracking weekly.

4The retention edge persists past week 1

It is not a one-week bump - Meta cohorts stay 3-4x more likely to return in weeks 2 and 3 as well.

CohortSourceWeek 1Week 2Week 3
Jul 13Meta (n=134)33.6%25.4%10.4%
Jul 13All other (n=1,296)10.7%6.6%2.8%
Jul 20Meta (n=929)46.2%15.1%-
Jul 20All other (n=1,271)12.5%4.2%-

5The gap: engagement is not converting to signups yet

Registration (completed onboarding) and tour requests by cohort and source.

Registration rate by source
Share of each cohort that has completed onboarding to date.
Untagged, Jul 20
149 of 990 registered
15.1%
Untagged, Jul 27
124 of 828 registered
15.0%
Meta, Jul 20
19 of 929 registered
2.0%
Meta, Jul 27
15 of 981 registered
1.5%

Tour requests follow the same pattern: 21 from the untagged Jul 20 cohort vs 3 from the Meta cohort.

Two honest caveats. First, ~93% of registrations sitewide are unattributed, so some "untagged" registrants are really Meta people whose later visit lost its tag - the true Meta rate is understated. Second, these cohorts are only 1-2 weeks old; home buying is a slow funnel and repeated exposure may convert later. Even so, the gap is too large to explain away: Meta leads are browsing repeatedly without creating accounts.

6What this means, and what we would do next

1

The loop is the asset - keep feeding it. Dynamic warm ads (ddwarm) are the workhorse of the re-pull, and moneyflow / prospecting are the best acquirers of people who come back. Budget shifts inside Meta should favor these until the data says otherwise.

2

Convert the browsing into accounts. ~1,900 Meta leads viewed homes in the last two weeks and 96% of their return visits land on property pages, not signup. Concrete moves: prompt account creation at the moment of repeat visits (save this home, price alerts, "see what you qualify for" with the 2% rebate math), and point a slice of the retargeting spend at a registration-focused variant for people with 2+ visits.

3

Expect the steady state to be ~25-30%, not 46%. Launch week benefited from fresh audiences. 25-30% week-1 return is still 3x every other channel - budget and expectations should be set against that number.

4

Watch cost per returning lead, not just cost per lead. Since retargeting drives the returns, every return click is paid media. The loop is only a win if the fully-loaded cost of a repeatedly-engaged lead beats what other channels pay for a one-and-done visit. We can wire this into the weekly report.

Methodology. Cohorts: first-ever property details page view on app.tryhoma.com, restricted to the Leads cohort (visited search + viewed a property). Retention: any return pageview in the following calendar week (Mon-Sun). Attribution: event-level UTM on each person's first property view - PostHog's default first-touch fields are unreliable on app landings and were not used. Bot check: the Meta catalog crawler (desktop, no campaign tag) was fingerprinted and confirmed absent from these cohorts - the Meta cohort is 97% mobile/tablet with real campaign tags. Registration = completed onboarding. Data pulled Aug 5, 2026 from PostHog project 141181; the Aug 3 cohort is excluded (week incomplete).