Quick Take: When virtually every major forecaster predicted continued rent growth heading into 2023, Origin’s proprietary Multilytics® platform made a call no one else was willing to make — that Class A multifamily rents would go negative. That prediction proved correct. But in the spirit of the same transparency that drove us to publish an uncomfortable forecast, this article confronts where Multilytics fell short on the pace of recovery in key Sunbelt markets, explains the construction-timing blind spot that caused it, and details the three structural enhancements we have made to the platform — including satellite-based construction monitoring, enhanced parcel-level data integration, and probabilistic delivery windows — to ensure our models are as precise on the “when” as they are on the “what.”
The Call Nobody Else Made
In December 2022, when double-digit rent increases were fresh in the industry’s memory and virtually every major forecaster projected continued positive growth, Multilytics® told us something nobody wanted to hear: Class A multifamily rents were going negative in 2023.
We published that forecast anyway. To our knowledge, we were the only market observer to do so.
We did it not because we enjoyed being contrarian, but because that is the entire point of Multilytics — to cut through consensus narrative and tell us what the data actually says, even when the answer is uncomfortable. We published it because we believe our investors deserve the complete picture, not a curated one.
That prediction proved correct. Rents went negative across nearly all of our target markets in 2023, exactly as Multilytics projected. Our mid-year accuracy report confirmed that ten of our fifteen target markets came within 2% of our point estimates. The other five markets landed within 0.8% of our forecast on an annualized basis. We did not retroactively adjust a single number.
That track record — built during one of the most volatile periods in modern multifamily history — is something we remain proud of.
So let us be equally straightforward about where Multilytics was less precise, why it happened, and what we have done to make sure it does not happen again.
What We Predicted for 2025–2026
In our January 2025–January 2026 Rent Forecast, Multilytics projected that six of Origin’s target markets would return to positive year-over-year rent growth by mid-2025, with many markets in positive territory by January 2026. The direction was right. The magnitude and pace were not — at least not in several key Sunbelt markets where negative or flat growth persisted deeper into 2025 than we anticipated. Markets like Austin, Phoenix, and Denver experienced more prolonged softness than our models projected.
We owe our investors an honest explanation of why.
Two Futures: The Scenario We Saw but Underweighted
Internally, well before we even published our 2023 forecast, our data science team modeled what we called “Two Futures” — two divergent paths for rent recovery that hinged on a single variable: the pace at which new supply actually hit the market.
Future One
This future assumed that the unprecedented construction pipeline — roughly one million units permitted in 2021 and 2022 when debt was historically cheap — would deliver on a relatively predictable schedule. Under this scenario, the bulk of new supply would land in 2023 and early 2024, get absorbed by still-robust demand, and rent growth would normalize by mid-to-late 2024. This was our base case, and it is the trajectory we published.
Future Two
This future assumed something more disruptive: that construction delays — driven by labor shortages, permitting backlogs, material cost spikes, and financing disruptions — would push a significant portion of that supply wave later into 2024 and even into 2025. In this scenario, the pain of oversupply would not resolve on schedule. Instead of a sharp dip and recovery, markets would experience a prolonged trough as delayed projects continued trickling onto the market quarter after quarter, suppressing rents for longer than the headline pipeline numbers suggested.
We discussed Future Two extensively. Our internal presentations included both curves side by side. We recognized it as a legitimate risk.
But we assigned it a lower probability. And that is where we were wrong — not about the economics, but about the construction timeline.
The data now tells us what happened. According to the National Multifamily Housing Council’s construction survey, 84% of builders reported construction delay, with most citing permitting requirements and financing uncertainty as primary causes. Yardi Matrix’s Q4 2024 supply forecast revised 2025 completions upward by 8.1% to over 508,000 units — meaning supply that should have cleared by 2024 was still arriving well into 2025. Average apartment construction timelines remained elevated at nearly 20 months nationally, with projects of 20 or more units averaging over 22 months from permit to completion.
In practical terms, the supply wave did not crest and recede. It plateaued. And that plateau dragged on the recovery timeline in exactly the way Future Two described.
Why Standard Supply Models Get This Wrong
To understand why this happened — and why even sophisticated forecasters missed it — you have to understand the fundamental limitation of how the industry tracks supply.
Most supply-demand models, including earlier versions of Multilytics, rely on permit data and pipeline reports from sources like the Census Bureau, CoStar, and RealPage. These sources tell you how many units are permitted and how many are under construction. What they do not tell you with any precision is when those units will actually deliver.
This distinction matters enormously. A market with 10,000 units under construction looks the same in a pipeline report whether those units are 6 months from completion or 18 months out. But the impact on rents is entirely different. If 8,000 of those units hit in a single quarter, you get a sharp oversupply shock that absorbs within two or three quarters. If they trickle in at 2,000 per quarter over a year, you get sustained downward pressure that grinds on far longer.
The post-pandemic construction environment turned what should have been a concentrated delivery into a slow bleed. Projects that broke ground in 2021 expecting 18-month timelines were finishing in 24 or even 30 months. The total number of units was not dramatically different from what the pipeline predicted — but the distribution across time was fundamentally altered.
Our models captured the “what.” They did not adequately capture the “when.”
What We Have Fixed
Forecasting is not about being right every time. It is about building systems that learn, adapt, and get more accurate over time. The gap between Future One and Future Two taught us something important: our supply-timing inputs were not granular enough.
We have since made three significant enhancements to Multilytics.
1. Satellite-Based Construction Monitoring
We have integrated satellite imagery and geospatial data to directly observe construction activity at the site level across our target markets. Rather than relying solely on permit filings and self-reported pipeline data — which tell us what is planned but not what is actually happening on the ground — we can now monitor the physical progression of projects from ground-breaking through vertical construction to exterior completion.
This is not a conceptual improvement. It is a structural one. Satellite observation allows us to independently verify whether a project reported as “under construction” is actively progressing, stalled, or delayed. It allows us to estimate completion timelines based on observed construction velocity rather than developer-reported schedules, which are notoriously optimistic. And it allows us to detect the early signs of construction slowdowns — reduced site activity, paused crane operations, stalled vertical progress — months before those delays appear in any pipeline database.
The practical impact is that Multilytics can now distinguish between 10,000 units under construction that will deliver in 6 months and 10,000 units that will deliver over 18 months. That distinction is precisely the one that separated Future One from Future Two — and precisely the one we failed to make in 2023.
2. Enhanced Parcel-Level Data Integration
We have deepened our parcel data inputs to track not just permitted units but zoning changes, lot assemblages, demolition permits, and entitlement activity at the individual parcel level. This gives us earlier visibility into where new supply is forming — often 12 to 18 months before it appears in traditional pipeline reports — and more precise geographic targeting of where competitive pressure will emerge.
Multilytics already operated at a sub-ZIP code resolution, carving markets into 100-meter squares and analyzing supply-demand dynamics at the street level. The enhanced parcel data makes this granularity actionable earlier in the development cycle. We can now identify not just where supply exists, but where it is forming — and adjust our forecasts accordingly.
3. Probabilistic Delivery Windows
Rather than modeling supply delivery as a single point estimate, we now assign probabilistic delivery windows to projects based on their observed construction stage, local permitting timelines, and historical completion data for similar project types in the same market. This produces a distribution of likely delivery dates rather than a single forecast, which in turn generates a range of rent growth outcomes that more honestly reflects the uncertainty inherent in supply timing.
In practical terms, this means our published forecasts will increasingly present scenarios rather than single numbers — not because we lack confidence, but because intellectual honesty demands it when the largest variable in the equation is construction timing, a factor subject to labor markets, weather, financing conditions, and municipal bureaucracies that no model can perfectly predict.
How Concession Burn-Off Quickly Drives Effective Rent Growth
There is another dimension to the 2026 rent story that headline numbers obscure, and it is one that matters enormously to operators and investors: the mathematics of concession burn-off.
During the peak of the oversupply cycle, many Class A properties in competitive Sunbelt markets were offering one to three months of free rent to attract tenants. These concessions are a rational response to oversupply — they fill units and generate cash flow — but they create a mechanical dynamic that almost guarantees meaningful effective rent growth as markets stabilize, even before asking rents move.
Here is the math, because it matters.
Consider a Class A unit with a face rent of $2,000 per month on a 12-month lease:
At Peak Concession (2 months free): The tenant pays for 10 out of 12 months. Total annual revenue is $20,000, making the effective monthly rent $1,667. That represents a 16.7% discount to face rent.
As the Market Tightens (1 month free): The tenant now pays for 11 out of 12 months. Total annual revenue rises to $22,000, and effective monthly rent climbs to $1,833. This single shift — reducing concessions from two months to one — produces a 10.0% increase in effective rent without the asking rent changing by a single dollar.
At Concession Elimination (0 months free): The tenant pays full freight: $24,000 annually, $2,000 per month. Moving from one month free to zero months free delivers another 9.1% increase in effective rent.
The full cycle from two months free to zero concessions represents a 20.0% increase in effective rent per unit — again, with no change to the posted rent.
Now apply this across a 300-unit property where the average unit rents for $2,000:
| Concession Level | Monthly Effective Rent | Annual Revenue Impact |
|---|---|---|
| 2 months free | $1,667 | Baseline |
| 1 month free | $1,833 | +$598,800 |
| 0 months free | $2,000 | +$1,197,600 vs. baseline |
As supply deliveries decelerate into 2026, properties that leased during peak concession period renew tenants at reduced or eliminated concessions. Every renewal cycle that replaces a concessioned lease with a market-rate lease generates real, measurable NOI growth.
For investors tracking only headlines of asking rent figures, this recovery is invisible. For operators and funds measuring actual effective revenue per unit, it is already underway.
The timeline is also faster than most people assume. Unlike asking rent increases — which require broad market tightening and typically lag supply-demand rebalancing by two to three quarters — concession burn-off happens lease by lease, month by month, as each concessioned tenant reaches renewal. A property that offered two months free on a 12-month lease beginning in January 2024 is already renewing those tenants at today’s lower concession levels. By mid-2026, the vast majority of pandemic-era and oversupply-era concessions will have cycled through.
This creates a compounding dynamic: effective rents rise from concession burn-off while asking rents begin to rise from tightening supply, producing a period of accelerated total rent growth that will likely appear in the data as a sharper recovery than most models currently project.
The Bigger Picture: What Multilytics Still Gets Right
It is worth stepping back from the specifics of the 2025 timing miss to consider what Multilytics has consistently gotten right — and why that matters more than any single forecast.
First, the directional call on 2023 negative rent growth was not a lucky guess. It was the output of a system that processes over 4 billion data points monthly, including demographic shifts, leasing velocity, rent concessions, and construction activity across every major multifamily market in the country. No other commercially available model made that call. The industry consensus was wrong. We were right. That matters.
Second, our accuracy at the national level has been consistently strong. Our 2024 accuracy report showed Multilytics forecasts within 96% accuracy of rent growth for our target markets and 99% accuracy at the national level during the first half of 2024. The timing miss on the recovery does not erase that track record — it contextualizes it.
Third, the fundamental thesis underpinning our longer-term outlook remains intact and is now playing out in real time. New multifamily starts fell roughly 50% from their 2022 peak. Completions are expected to drop 15% in 2025 and over 50% by 2026. The pipeline is emptying. Demand remains robust, with Q4 2024 absorption reaching approximately 230,000 units nationally — one of the strongest quarters in 25 years. The supply-demand rebalancing we projected is happening. It is just happening on a lagged timeline.
Fourth, and this is the piece that often gets lost in discussions of forecast accuracy: Multilytics kept us out of trouble. Because we saw the oversupply coming in 2021 — long before it was consensus — we adjusted our underwriting, our deal selection, and our market exposure accordingly. We did not chase Austin at its peak. We did not overweight markets where the pipeline was most bloated. That risk management, powered by Multilytics, protected capital.
A Commitment to Accountability
We publish our forecasts publicly. We publish our accuracy reports publicly. We do not retroactively adjust our predictions to make them look better. When we are right, we say so. When we are wrong, we say that too.
The 2025 timing miss was a miss. We own it.
Every forecasting system has a moment that reveals its blind spots. The best systems are the ones that find those blind spots and fix them. That is what we have done.
The supply wave is cresting. Concessions are burning off. Demand is accelerating. And Multilytics, now equipped with the tools to track not just what is in the pipeline but how fast it is actually moving through it, is positioned to guide our investment decisions with greater precision than ever.
We staked our credibility on a public forecast that nobody else was willing to make. We were right when it was hardest to be right. We were late on the recovery. And we have fixed the reason why.
That is the record. We stand behind it.
There is a private real estate investment strategy for virtually every investor.
FAQ
Where did the Multilytics forecast fall short, and why?
Our 2025–2026 Rent Forecast correctly identified the direction of recovery, but underestimated how long softness would persist in key Sunbelt markets like Austin, Phoenix, and Denver. The cause: construction delays pushed supply that should have cleared by 2024 well into 2025.
What did Multilytics® predict for 2023, and was it correct?
In December 2022, Multilytics projected that Class A multifamily rents would go negative in 2023 — a call that ran counter to virtually every major forecaster at the time. It proved correct. Ten of our fifteen target markets came within 2% of our point estimates; the remaining five landed within 0.8% on an annualized basis.
What is “concession burn-off” and why does it matter?
When oversupply eases and concessioned leases reach renewal, effective rents rise — even before asking rents move. The full cycle from two months free to zero concessions represents a 20% increase in effective rent per unit, with no change to posted rent.
What is Origin’s outlook for multifamily in 2025 and 2026?
New starts are down roughly 50% from their 2022 peak. Completions are expected to drop 15% in 2025 and over 50% by 2026. Q4 2024 absorption hit approximately 230,000 units — one of the strongest quarters in 25 years. The recovery is underway.
What has Origin done to improve Multilytics?
Three structural upgrades: satellite imagery to monitor real-time construction progress; deeper parcel-level data to detect new supply 12–18 months earlier; and probabilistic delivery windows that produce a range of outcomes rather than a single point estimate.
How accurate has Multilytics been overall?
Our 2024 accuracy report showed 96% accuracy for target markets and 99% accuracy at the national level during the first half of 2024.
Sources
- Markerr. 2023 Rent Growth Forecast: Still Above Long-Term Average Rent Growth. Markerr Research.
- CBRE. 2023 U.S. Real Estate Market Outlook: Midyear Review. CBRE Research, 2023.
- JLL. Multifamily Research and Outlook, 2023. Jones Lang LaSalle. (Original report no longer publicly available. Data cited as of publication date.)
- Green Street Advisors, as cited in MBA NewsLink. “Commercial Real Estate Fundamentals Healthy.” Mortgage Bankers Association, February 2023.
- Freddie Mac Multifamily. 2023 Multifamily Outlook. Freddie Mac Research, 2023.
- RealPage Analytics. “Apartment Demand Returns in the 1st Quarter.” RealPage, 2023.
- National Multifamily Housing Council (NMHC), as cited by Berkadia. “NMHC 2023: National Trends to Watch for the Year Ahead.” Berkadia, 2023.
