Best Practices vs. Contextual Fit in Social Services
A housing program works in Vancouver. Word travels. A mid-size city in Ontario adopts the same model, staffs it the same way, reports on it the same way, and waits for the same results. Two years later, the numbers are flat, the funder is asking questions, and everyone in the room is surprised.
The surprise is the tell. It means we expected the practice to carry its results with it, like a recipe that turns out the same in any kitchen. Social programs don't behave that way, and the language of "best practice" quietly encourages us to forget it.
When a model succeeds somewhere, the success belongs to the practice and the conditions around it: the population it served, the housing market it operated in, the partner agencies it could lean on, the funding that held it together. Borrow the practice without the conditions, and you've borrowed half the thing.
Why "Best Practice" Is a Weaker Standard Than It Sounds
The phrase does useful work as shorthand. It signals that an approach has been tried, studied, and found to help. The trouble starts when "best" gets heard as "best everywhere," because that's almost never what the evidence actually showed.
Most evidence behind a best practice comes from specific places, populations, and moments. A program evaluated in a large coastal city with a tight rental market and a dense network of mental health providers was tested under those conditions, not yours. The label travels well. The conditions that produced the result do not.
So "best practice" answers a question that's narrower than it appears. It tells you something worked, for some people, somewhere, under conditions that may or may not resemble your own. That's worth knowing. It just isn't the same as knowing it will work for you.
A Best Practice Is Really a Practice-Plus-Context Bundle
Think of any program as two things bound together: the active ingredients and the environment that lets those ingredients act. A peer-support model depends on a workforce you can actually recruit. A rapid-rehousing approach depends on landlords willing to rent. A coordinated intake process depends on agencies willing to share a client.
When the environment matches, the practice looks powerful. When it doesn't, the same practice underperforms, and we tend to blame the practice (or the team running it) rather than the mismatch. The model didn't fail. The transplant did.
This is why two faithful copies of the same program can produce different results a few hundred kilometres apart. Nothing about the practice changed. The landlord landscape, the service ecosystem, and the population did.
What the Evidence Actually Says About Transfer
This isn't a hunch. It's one of the more durable findings in program evaluation.
Canada's largest test of Housing First was built around the same logic. The $110-millionAt Home/Chez Soi project, run by the Mental Health Commission of Canada from 2009 to 2013, didn't only ask whether Housing First works. It ran the model across five cities (Vancouver, Winnipeg, Toronto, Montréal, and Moncton) specifically to learn how it performed in different environments and for different groups. Toronto went further and ran a dedicated arm, adapting the model for people from ethno-racial communities. Thenational final report gives real attention to local variation rather than averaging the five sites into one tidy headline.
That's the difference between citing a best practice and understanding one. The study's strength wasn't that it crowned a winner. It was that it mapped where, for whom, and under what conditions the approach was delivered.
Contextual Fit: A More Useful Question Than "Best"
If "is this a best practice?" is the weak question, "does this fit our context?" is the stronger one. Contextual fit asks whether the conditions that made a model work somewhere else exist, or can be built, where you are.
It's a more demanding standard, and a more honest one. It forces a program team to name the active ingredients, identify the conditions those ingredients need, and check those conditions against the local reality before committing budget and staff. It replaces a yes/no verdict with a diagnosis.
Fit also reframes failure. When a well-fitted program struggles, you can look at which condition was missing rather than concluding the model is broken or the staff fell short. That makes the next decision sharper instead of just disappointing.
None of this means starting from scratch every time. Evidence still matters enormously. The point is to import the model and the question that comes with it: what did this approach actually depend on, and do we have it?
What Rigorous Program Adaptation Actually Involves
Adapting a proven model isn't the same as diluting it, and it isn't improvisation. There's a real discipline to it, much of it developed in implementation science. Four moves matter most.
Separate the core components from the adaptable periphery
Every effective program has core components: the elements that carry the effect and can't be dropped without breaking it. Around that core sits an adaptable periphery: the features you can change to fit local conditions without losing the mechanism.
The work is telling them apart. In Housing First, immediate access to housing without preconditions is core; the exact staffing ratio, the type of housing stock, and how outreach is organized are more peripheral. Get this distinction wrong and you'll either protect the wrong things or quietly gut the thing that worked.
Map the local conditions before you import the model
Before adopting a model, profile the conditions it will land in: the population and its needs, the rental or service market it depends on, the partner agencies it assumes, the workforce you can realistically hire, and the funding structure that has to sustain it.
Lay that profile beside the conditions in the original site. Where they match, you can adopt with confidence. Where they diverge, you've found exactly where adaptation (or honest caution) is required. That mid-size Ontario city wasn't wrong to look at Vancouver. It was wrong to skip this step.
Build measurement in from the start
Adaptation only works if you can tell it apart from drift. Adaptation is a deliberate change to the periphery to improve fit; drift is the slow, unnoticed erosion of the core. They can look identical in a quarterly report and produce opposite results.
The way you separate them is by measurement. Implementation research consistently finds that how well a program is delivered shapes whether it works at all, which means you need to track both fidelity to the core and the outcomes you actually care about. Without that, you're flying blind and calling it judgment.
Treat adaptation as ongoing, not a one-time setup
Contextual fit isn't a launch task you complete and file away. Populations shift, housing markets tighten, partners come and go, and funding changes shape. A program that fits well in year one can fall out of shape by year three without anyone touching it.
The strongest teams revisit fit on a schedule, using their own outcome data to ask whether the conditions still hold and whether the adaptations are still earning their place. Adaptation becomes a habit rather than an event.
What This Means for Decision-Makers and Internal Champions
For people who approve budgets and answer to boards and funders, the practical shift is in the questions you ask before adopting something.
Instead of "is this a recognized best practice?", ask "what conditions did this depend on, and which of them do we have?" Instead of "did it work elsewhere?", ask "what would we need to change for it to work here, and would those changes leave the core intact?" Those questions turn a procurement decision into a fit assessment.
There's an infrastructure point underneath all of this. You can't distinguish adaptation from drift, or fit from mismatch, if your systems only count activities. That depends on case management and data tools that track outcomes and local conditions over time, and that bend to your workflows rather than forcing a single template on every site. Picking systems for that kind of flexibility, rather than feature lists, is part of making contextual fit a standard you can actually uphold. It's also what lets a network of programs compare results against shared outcomes instead of incompatible local metrics.
For internal champions, this reframing is quietly empowering. It gives you a credible way to push back on "just copy what they did in the bigger city," and a structured method for doing the adaptation well when a model genuinely deserves to travel.
The Takeaway
"Best practice" is a useful starting point and a poor finishing line. It tells you something worked somewhere. It doesn't tell you it will work for your population, in your market, with your partners, and your funding.
Contextual fit is the better standard because it asks the question that actually predicts results: do the conditions that made this work exist here, and can we tell whether they still do? Answering it well means separating the core from the periphery, mapping local conditions honestly, measuring delivery and outcomes, and treating fit as something you maintain.
If your organization is weighing whether to adopt or scale a proven model, that's the conversation worth having before the launch, not after the disappointing report.