Prior to me joining, the team had been building Echo, a documentation tool already piloting in a few homes. But through early conversations with care providers, the referral problem kept surfacing. Unlike documentation, where solutions already existed, nobody was tackling referrals with dedicated technology. The need felt more urgent. We didn't abandon Echo, but we chose to focus on referrals.

Before touching any design tool, I needed to understand a domain I had no experience in. My first question wasn't "what should we build?" It was "what actually happens when a referral arrives, and who's involved?"

Desk research: mapping the ecosystem

I started by mapping the referral process: which agencies are involved, who does what, and how they relate to each other.

Local authorities hold the duty to find placements. Their social workers compile referral documents. Placement teams distribute these to providers, usually mass-emailing the same referral to dozens of homes at once ("spray and pray"). Care providers (from single homes to 200+ home groups) assess each referral against capacity, training, regulation, and the needs of children already living there. Ofsted oversees everything, scrutinising matching decisions and holding managers accountable.

The relationships are strained. Local authorities are under pressure to find placements quickly. Providers feel overwhelmed and frustrated by scattered information. Communication is difficult. Managers submit expressions of interest and most times hear nothing back.

Once I had a picture of who was involved, I mapped the dynamics between providers and local authorities: what happens when things go wrong, where pressure builds, and why the system stays stuck.

Three things stood out.

The system is caught in a pressure loop. When urgency increases, local authorities rush referrals out with poor information. Poor information leads to poor matching. Poor matching leads to breakdowns. Breakdowns create more urgency. There's no point where it self-corrects. This told us any intervention needed to target speed and information clarity first, because those are the points where the loop could be interrupted.

Providers absorb the burden to keep everything functioning. The admin overhead (reading, logging, tracking, responding, re-reading the same referrals) isn't a side effect. It's how the system compensates for poor upstream processes. Providers have to work very hard to sustain the system.

There is no shared data visibility across the system. Local authorities and providers operate with almost no insight into each other's patterns or decisions over time. Providers can't see what the needs are across local authorities, while local authorities can't see why providers decline. Neither side has the data to recognise trends, learn from outcomes, or improve their decision-making over time. The institutional knowledge that does exist is held by experienced individuals, built through years of pattern recognition, and lost when they leave.

User Interviews

Over June and July 2025, I ran 8 semi-structured interviews across 3 care providers with people covering roles from referral managers, home managers, directors and clinical psychologists.

Together, they painted a picture of a system under enormous strain:

The information problem is structural. Over 150 local authorities each use a different format. Documents are routinely outdated and sometimes deliberately incomplete. There is no national standard.

Admin crowds out the real work. Managers aren't slow because they lack skill. They're buried in admin that leaves no room for assessing compatibility, building relationships, doing qualitative work.

Nobody wanted full automation. This came up in every conversation, unprompted. The direction shifted from "AI matches children to homes" to "AI clears the admin so humans can match better."

"It's the worst thing about referrals, and that is that there isn't anything centralised. So you've got however many placing authorities, 40, 50 placing authorities, and every single one of them uses a different form." - Registered Manager

Two Personas, One Structural Insight

The most involved roles in the referral process were those of the referral managers and of the home managers.

Both personas described the same frustrations: volume and the need for standardisation stood above everything else, raised in every single interview. That made it our starting point and established Linda as our primary persona.

• Volume overwhelms everything. Most of the day goes to reading referrals that will never be suitable.

• Every referral is a puzzle. The same information, in different places across every document, makes it time consuming to read through.

• Workflow is scattered across platforms. Outlook, Egress, placement portals, Excel, notebooks. Time lost switching, referrals missed between channels.

• Repeat referrals eat up time. Same referrals are sent 2 to 4 times a week, sometimes with different subject lines, so they appear as new referrals.

• The response cycle is broken both ways. Officers can't get managers to respond in time. Managers send interest and hear nothing back.

• Manual logging is universally required and universally hated. Every organisation needs to track referrals. Everyone described it as painful, but leadership requires it.

• Information quality is unreliable, sometimes deliberately. Experienced managers develop a skill for reading between the lines that newer ones don't have yet.

• The "positive information" gap. Referral documents focus almost entirely on risk. Both personas wanted to see strengths, hobbies, interests. One manager described looking for "the voice of the child" and "glimmers of things they might be genuinely interested in."

  "Don't get me started on the Egress or the Microsoft security access or the number of ways that they want to send it and the formats that are all different." Referral Officer

The structural insight: not every organisation has a referral officer. Some have a dedicated coordinator, others don't. The interface had to support both a coordinator-led workflow and a self-serve workflow from day one.

The Competitive Landscape

Almost nothing purpose built existed for children's residential care referrals. There have been attempts to solve this by local authorities grouping to create portals, but this is still on a local level, and it ended up in maximixing effort for providers having to access several portals. “Found by Lottie” provided great inspiration, but it serves adult care (different context). “Leaf” and some broader CRMs have basic features buried in larger systems. We even came across one organisations that started building their own internal solution but eventually stopped due to other priorities.