Whitepaper · v0.1 · May 2026

Binly

A behaviour-change platform for the next era of municipal recycling.

Author: Mathaius · Birmingham, UK
Pilot proposal · Birmingham City Council · Lozells ward · Q3 2026
Contact: mathaius3333@gmail.com

CONFIDENTIAL. This document contains proprietary business information, unit economics, and product strategy. Recipient agrees not to distribute, reproduce, or share with third parties without written permission from Binly. © Binly 2026.

1. Executive summary

2. The Birmingham recycling problem

Birmingham is the second-largest local authority in the UK by population, and one of the lowest-performing for household recycling. The most recent Defra figures put household recycling rates in the city well below the English average and at less than half of the national 50% statutory target. The city's existing waste contract with Veolia has, over the past five years, focused on infrastructure: bin rollouts, MRF (materials recovery facility) capacity, and procurement of replacement vehicles. None of this addresses the human bottleneck: residents don't know what goes in which bin, and when in doubt they default to the wrong one.

The contamination cost

Contamination - the wrong items ending up in the recycling stream - is the single most expensive failure mode in modern municipal waste. Once a load is contaminated above a threshold (typically 5–10% by weight), the MRF cannot economically separate it, and the entire lorry-load is rerouted to general waste or incineration. The council pays a gate fee for this. Since China's National Sword policy in 2018 closed the export market for low-grade mixed recyclables, this gate fee has roughly doubled. Industry figures for English councils put the cost of rejected contaminated recycling at approximately £80–£140 per tonne - a meaningful line item in any council's waste budget.

The reform window

Birmingham is rolling out the largest reform of its waste service in over a decade, starting June 2026. The reform has three components: (a) a separate weekly food-waste collection using kerbside caddies; (b) a second recycling bin to separate paper/cardboard from mixed containers; and (c) a target of 40% household recycling rate by year-end. The reform is driven partly by Defra's Simpler Recycling regulations, which mandate standardised collection streams across English councils by 2027.

The reform itself does not change behaviour. New bins and new caddies will sit on doorsteps with their contents determined by the same residents who have, in aggregate, produced a sub-30% recycling rate. Without an explicit behaviour-change intervention, the council is relying on leaflets and goodwill to hit a 12-point recycling-rate increase in seven months.

3. Why reward schemes failed before - and why 2026 is different

Recycling reward schemes are not a new idea. Between 2009 and 2015, multiple UK and US councils ran pilots, some at city scale: Royal Borough of Windsor & Maidenhead partnered with Recyclebank from 2009 (later scrapped); Birmingham City Council itself ran a Nectar-points-for-recycling pilot in the early 2010s; Recyclebank ran in many US cities. By 2016, most had been quietly wound down. Any serious investor or council partner will know this history - and will reasonably ask why we expect a different outcome. This section addresses that directly.

Why prior schemes failed

4. The Binly solution

Binly is a free consumer app that residents open before binning an item. They scan a barcode or take a photo, the app identifies the item and tells them which bin it belongs in, and - if they confirm they binned it correctly - they earn points. Points are redeemed for small, immediate rewards from local Birmingham businesses: coffee, swim sessions, bus credit, market vouchers. The app integrates with Birmingham City Council's existing waste service rules and, in pilot ward areas, with Veolia's MRF contamination samples to close the verification loop at ward level.

The mechanism is the teaching moment, not the reward

Almost every recycling reward scheme that has been built — including the schemes named in §3 — has framed itself as a rewards scheme. That framing is wrong both substantively and politically. Substantively, behaviour-change literature (Fogg 2009, Duhigg 2012, Wood & Rünger 2016) consistently shows that habit forms around cue + immediate feedback, not around the reward. The reward matters far less than the dopamine hit of getting it right in the moment. Politically, leading with rewards invites the "rewards schemes have failed before" objection (true, see §3) and makes the integrity pipeline the primary defence rather than the secondary one.

Binly is best understood as a teaching app with a rewards layer attached as a hook:

• The mechanism of behaviour change is the five-second voice announcement that fires on every scan: "Coke can, mixed recycling, plus twelve. Rinse if sticky." Per-item, per-second feedback at the point of decision. This is what the leaflet, the bin lid colour, and the council awareness campaign cannot do — they can describe the right answer in general; Binly tells you the right answer for the specific thing in your hand.
• The hook is the reward — the points that open the app, the local-rewards economy that makes the habit visible to peers, the streaks that make absence feel like loss. Rewards drive months 1-3; teaching is what compounds across months 4-24.
• The integrity pipeline (§5) exists to keep the rewards economy honest enough for a council to put its name on it. It is defence-in-depth for the rewards layer, not the primary mechanism of behaviour change.

This reframe matters for two practical reasons. First, the teaching layer has almost no fraud surface — you cannot meaningfully "fake" the moment a resident learns that pizza-box-bottoms are recyclable but pizza-box-tops are not. The integrity overhead is therefore proportionate to the rewards layer only. Second, the long-term retention story (§4 below) is much stronger under the teaching frame: habits persist when rewards are removed; reward-only schemes collapse the moment the points dry up.

The friction problem, and how Bin Session solves it

A scan-per-item flow that requires "open app, point at item, tap confirm, pick bin, tap submit" takes ~30 seconds. The average household bins ~30 items a day. Asking 30 × 30s = 15 minutes a day is not realistic — and any model that requires near-100% scan coverage to be useful is brittle.

Binly does not target 100% per-item coverage. It targets a high-frequency behavioural fingerprint: 5–15 scans per active household per day, enough to characterise their material mix, accuracy rate, and behaviour over time. Combined with Veolia's existing route-level MRF sampling, the two signals triangulate. Where the household-level pattern and the MRF route data disagree, the system flags it.

To make 5–15 scans/day realistic, Binly ships Bin Session: a hands-free flow where the resident taps once to open the camera, holds each item to the lens for ~1 second, and the app auto-logs + voice-announces ("Coke can, mixed recycling, plus 12") with no per-item taps. Lock-on takes 700ms, a 3.5s per-category cool-down prevents the same item being logged twice while the resident walks to the bin, and a centre+size filter ignores background clutter. Five items takes ~15 seconds.

User journey

What the council gets

The council-facing dashboard (see prototype/admin.html) provides anonymised, ward-level data on what's being scanned, what people are getting wrong (contamination "top offenders"), and how recycling rates are tracking against the 40% target. This is information the council does not currently have. It enables ward-targeted education campaigns and gives Veolia early signal on contamination trends - months before the MRF audit data lands.

5. Verification methodology - the integrity pipeline

5.1 Why this is the critical problem

Without trustworthy verification, every other claim in this whitepaper collapses. Councils cannot defend awarding public-money rewards to residents who haven't earned them. Grant funders won't pay for an honour-system. Partner businesses won't fulfil rewards they don't trust. And residents who do recycle correctly will resent freeloaders. The previous wave of schemes (see §3) attempted to verify behaviour at the lorry - weighing bins, RFID tags, contamination samples averaged across whole streets. We move verification to the point of decision (the scan) and close the loop at the point of collection (the bin). The result is a system where each component is doing modest work, and the chain as a whole is rigorous.

Coverage philosophy. We do not require every binned item to be scanned. Average UK households generate ~30 binnable items per day; demanding 100% scan coverage is unrealistic and any model that relies on it will fail under real-world adherence (40–60% in even the best-engaged habit apps). Instead we target a behavioural fingerprint: 5–15 scans per active household per day, ranging across material categories. Combined with Veolia's existing MRF route samples (ground truth), the household-level signal and the route-level signal triangulate. Where they disagree, the integrity score and audit pipeline kick in. This is statistically equivalent to how air quality networks, election polls, and credit scoring systems operate — sampling plus anomaly detection, not exhaustive measurement.

5.2 The four-layer pipeline

Layers 1 and 2 are live in the prototype as of June 2026. Layers 3 and 4 require the council pilot to activate.

Layer 1 - Verified resident

At signup, a user provides their address (postcode + house number) and is cross-checked against Birmingham City Council's records (electoral roll and Council Tax) under a data-sharing agreement. The user then takes a photo of their passport, UK driving licence, or BRP, followed by a selfie with a basic liveness check (random head-movement instructions). The document and face are matched by a third-party identity provider - Yoti, Onfido, or Veriff are all viable; per-check cost is typically £1.20–£2.80 in 2026. Successful verification produces a binding between a real person, a real address, and the Binly account. The ID image itself is deleted within 24 hours of verification; only the binding is retained.

Equality safety net. A hard digital-ID requirement excludes rough sleepers, undocumented residents, people without smartphones, and people without photo ID. To remain compliant with the Equality Act 2010 and the council's own inclusion duties, we operate a manual vouching path: a council officer (or trusted community partner - e.g., a Lozells ward champion) can vouch for a resident in person, creating the same binding without digital ID. We expect 5–10% of pilot accounts to use this path.

Layer 2 - Bin QR binding

Every household's bins (recycling, food, general) receive a unique, weatherproof QR sticker. The format is short (e.g. BR-K42R) and human-readable, so a Veolia operative can also key it in manually if a sticker is damaged. Stickers are applied during Birmingham's June 2026 bin-replacement rollout - Veolia is already swapping containers across the city, so the marginal logistics cost is near zero. Each scan a user makes in the app is logged against the bin the user is intending to use, creating a per-bin scan ledger.

The user can optionally scan the bin's QR before binning (we call this "opening a session") for a +1 bonus, which both nudges good behaviour and gives us higher-confidence binding for that disposal event.

This is now live in the prototype. In Bin Session mode, the resident taps "Verify bin" in the camera UI, points at the bin's sticker, and the session enters attributed mode. Every subsequent auto-logged item is tagged with the bin QR, gets a 15% point boost when it matches the recommended bin, and recovers the resident's integrity score by +2 (capped at 100). If the scanned item does NOT match the active session's bin — e.g., the AR detector identifies a Coke can while the user is binning into a general-waste-attributed session — the item is logged as a wrong-bin event, points drop to 0, and the integrity score drops by 3. This is Layer 2 doing real work: per-event bin attribution + a feedback signal the resident sees the same second they bin.

Layer 3 - Collection-time scan

When the Veolia truck collects a household's bin, the operative - or, in newer trucks, an onboard camera + lift-mechanism reader - scans the bin's QR. This timestamps the collection event and, importantly, identifies which scanned-but-not-yet-collected items are now "out of system." A rotating 5–10% of collected bins are routed to the MRF audit lane, where contamination is measured by mass. The QR is re-scanned at the audit station, tying the contamination measurement back to the originating bin and (through Layer 2) the originating resident's scan ledger.

This is the layer that catches the dominant fraud mode: scanning items without binning them. If the audit ledger shows a user scanned 30 items in the recycling bin over the week but the bin arrives at the MRF nearly empty, that's a flag. If the audit shows the bin is full of contamination items the user never scanned, that's also a flag. Honest residents experience this layer as invisible; their audits come back clean.

What happens at the MRF — and what we don't try to do

A common question: "won't the depot sort everything anyway?" Partially. The MRF runs optical sorters, magnets, air classifiers, and manual pickers to separate the recycling stream into commodity bales (PET, HDPE, aluminium, paper, glass). But three things constrain what depot-side sorting can do:

1. The MRF only processes the recycling stream. Anything that went into the general-waste bin goes directly to landfill or incineration; it never reaches the MRF. A Coke can in the wrong bin is gone forever, regardless of depot capability.
2. Contamination above a threshold rejects the entire load. UK MRFs commonly reject loads above ~15% contamination at a cost of £80–140/tonne. The depot can pick contamination OUT, but it can't tell you which household put the nappy in the paper bale — and one bad household can spoil a tonne of paper for everyone else on the route.
3. Once tipped, individual items are unattributable. When the truck tips onto the MRF floor, contents from dozens of bins mix together. No optical sorter or manual picker can reverse-engineer "this Coke can came from house 14." This is a physical limit, not a software one.

Binly's response to these constraints is operationally modest. We do not try to identify every item back to its source after tipping — that's physically impossible without per-item RFID, which we explicitly do not require. Instead the audit lane keeps a sample of intact bins isolated long enough to compare their actual contents against the household's scan ledger:

• 5–10% of collected bins are diverted to the audit lane before being tipped into the truck (or, in some operational models, kept in a separate compartment of the truck and tipped onto a marked area of the MRF floor)
• The bin's QR is re-scanned at the audit station, anchoring the contents to a specific household's scan ledger from Layers 1–2
• The bin is then audited the same way any contamination sample would be — visual identification, weighing of contamination classes, photo log
• Post-audit, contents are tipped onto the MRF floor like any other bin and processed normally — the audit doesn't waste recyclable material

The household-level signal this produces is sparse (each bin is audited only a few times per year) but it's high-confidence at the moments it fires. Combined with the per-scan behavioural fingerprint from Layer 2, the system catches fraud through pattern-matching across many cheap signals (scan velocity, time-of-day, bin-mix consistency) and a few expensive ones (the audit results). This is the same defence-in-depth pattern as credit-card fraud detection: most checks are pattern-based; the expensive checks fire only when patterns look wrong.

Layer 4 - Escalation ladder

Audit mismatches feed back into the resident's integrity score, which determines what they're able to earn. Flags accumulate; consequences escalate. At every tier, residents retain the right to appeal in writing to the council, and the educational features of the app (scanning, learning, bin guidance) remain available even when reward-earning is paused. This is the "rewards earned, not entitled" framing.

5.3 Data protection, framing, and political risk

Linking verified identity to bin-level recycling data is sensitive personal data under UK GDPR. Three commitments are non-negotiable:

• Council is the data controller; Binly is the data processor under a DPA. Bin-level data shown to Veolia operatives is pseudonymised - they see "bin BR-K42R: 14% contamination" not "Mathaius: 14% contamination."
• Explicit, informed consent at signup, with full data-export and deletion rights honoured within 30 days. Withdrawal of consent ends the account but does not penalise the resident.
• Framing matters. Councils are sensitive to accusations of "policing residents." The escalation ladder is explicitly framed as gating rewards, not punishing recycling behaviour. A flagged resident does not have their bin emptied differently, does not face any council action, and is not reported to authorities. They simply do not earn rewards until the issue is resolved.

6. Pilot design

Why Lozells

Lozells is selected as the pilot ward based on four criteria: (a) the 2022–2026 Ward Action Plan already prioritises community participation and gives us natural channels for door-to-door engagement; (b) the current recycling rate is below the city average, providing meaningful upside to measure; (c) the demographic mix is broadly representative of Birmingham overall, improving the generalisability of the result to a city-wide rollout; and (d) Veolia operates the existing waste collection in this ward, simplifying MRF data integration.

Timeline

Primary KPIs

1. Contamination rate delta. Pre vs post-pilot, measured by Veolia MRF samples. Target: ≥5 percentage point reduction.
2. Active users. Defined as ≥1 scan in the last 14 days. Target: 500.
3. Retention. Week-4 retention of week-1 users. Target: ≥40%.
4. Cost per active user. All-in (rewards + ops). Target: ≤£2.

7. Impact projections

Three scenarios - conservative, base, optimistic - modelling the city-wide cost saving if the Lozells pilot's contamination reduction is replicated across Birmingham's roughly 450,000 households. All figures assume the £80–£140/tonne contamination rejection cost band and Birmingham's reported annual household waste tonnage.

Long-term retention — what survives when the rewards economy isn't growing

Behaviour-change apps reliably show three phases of user adherence: (1) enthusiastic period (months 1–3) when novelty and the rewards economy carry users; (2) plateau (months 4–6) as the points cease to feel novel; (3) equilibrium (months 7+) where only habit, identity, and community signals retain users. Reward-only schemes collapse during phase 2 — see the historical schemes in §3.

Binly's retention strategy explicitly invests in the phase-3 mechanisms before the phase-1 traction has been fully captured:

• Habit cues — collection-day notifications, the iOS Live Activity that surfaces the truck near you, the Control Centre quick-scan button that means opening the app costs one tap from any state of the phone. Each of these makes the behaviour cue-driven rather than reward-driven.
• Identity signals — the trust score (§5) is visible on the home tab and earned over time. Households become "the kind of household that recycles right." Identity outlasts motivation.
• Community signals — street-level and postcode-level visibility (planned for the pilot's community layer) creates social proof. Behaviour-change literature is consistent that peer-visible behaviour outperforms private incentive in the long run by a wide margin.
• The teaching loop itself — the per-item voice feedback (§4 above) is a habit. People keep scanning because they want to know if they got it right, not (only) because of the points.

This is the part of the model that most prior schemes never reached, because they collapsed during phase 2. Binly's pilot KPIs (§6.3) include a phase-3 retention metric (week-24 active users / week-1 active users) precisely because that's the number that decides whether the model is durable.

Scope limits — what Binly does not solve

8. Business & funding model

Revenue streams

Two distinct streams across two product lines. Year-1: council-paid SaaS on the free-to-resident app. Year-2: Binly Premium consumer subscriptions for speciality waste streams the council does not collect. Critically the two lines reinforce each other - Premium leverages the verified-resident base built by the council pilot, and the council relationship gives Binly Premium a credible local brand without expensive paid acquisition.

Binly Premium - the Year-2 consumer subscription

The council kerbside service handles the bulk streams: general waste, mixed dry recycling, food (single brown caddy from June 2026), paper/card in SERCO areas. There is a long tail of speciality streams that the council either doesn't take or takes only at recycling centres - and households consistently report guilt about throwing these items away. Binly Premium is a low-capex private collection product for that tail. We're not competing with the council's lorries on general kerbside (uneconomical at startup scale); we're operating small electric vans on monthly speciality routes, collecting what they don't.

Per-route unit economics

One small electric van + one driver serving ~1,500 paying subscribers across mixed tiers (one Birmingham postcode district):

Add £5,000–£15,000/route/year from selling sorted streams to specialist recyclers (textiles to resale partners, soft plastics to TerraCycle / Re-Mix, batteries to ERP UK).

Year-1 unit economics (free council-funded app)

• Cost per active user, all-in: £1.50–£2.00/year (rewards budget + hosting + support).
• Revenue per active user (council-funded): £3.00–£5.00/year (council SaaS / user).
• Gross margin per active user: ~£2.00/year. At 50k users (city-wide Birmingham), £100k contribution before Premium.

Grant stack

9. Operations - how Binly Premium delivers

9.1 The four-party chain

The operational model is deliberately lean. Most "specialty waste" startups fall over because they try to build sorting infrastructure on a startup balance sheet. Binly Premium uses existing infrastructure as a paid customer and earns rebates from existing buyers. We own only the collection layer.

9.2 Why we do not need our own depot at pilot scale

Three reasons the standard "build a depot, hire sorters" model is wrong for Year-1 and Year-2:

1. Sort-at-source removes the most expensive step. A typical materials recovery facility (MRF) spends 60-70% of its operating cost on labour and machinery to separate mixed streams. Binly's app teaches residents to do that separation themselves, in their own kitchen, with a colour-coded bag per stream. By the time our van arrives, the work is done.
2. Transfer station infrastructure already exists. Birmingham has at least six commercial transfer stations within 20 minutes of Lozells (Biffa Tyseley, Veolia Smethwick, FCC Hams Hall, and others). They have surplus capacity, charge competitive tipping fees (£20-80/tonne depending on the stream), and welcome consistent specialist suppliers because high-quality pre-sorted material is more profitable for them than mixed loads.
3. Capital efficiency matters at startup stage. An owned mini-MRF runs £500k-£2m capex plus £100k+/year opex (planning permission, environmental permits, fire safety, business rates, sorting equipment maintenance). Outsourcing tipping fees at £20-80/tonne is fractional in comparison: 50 tonnes/year per route × £40/tonne = £2,000/year. The break-even tipping spend that would justify an owned facility is around £80k/year - five thousand subscribers, multiple routes, consistent volume. That is a Series A problem, not a Smart Grant pilot one.

9.3 Regulatory footprint

Operating Binly Premium legally in England and Wales requires the following, all of which we will register before the first pickup:

9.4 Per-route operational economics

One Binly van + one driver serving 1,500 mixed-tier subscribers in a single Birmingham postcode district:

Revenue at realistic tier mix (30% Soft Plastic Saver, 50% Eco Plus, 20% Full Stream) is approximately £142,000. Add £5,000-£15,000 in commodity rebates from selling clean sorted streams to specialty recyclers. Net profit per route per year: £60,000-£70,000 at the realistic mix, scaling linearly with subscriber count.

9.5 When the calculus changes (and we revisit the depot question)

The outsourced-logistics model holds while three conditions are true: tipping fees stay below ~£80/tonne, transfer stations have surplus capacity within 20 minutes of the routes, and our volume per stream stays below the threshold where specialty recyclers prefer direct delivery. All three are true at Year-1 (Lozells pilot, 500 subscribers) and Year-2 (Birmingham city-wide, 10,000-22,500 subscribers).

At Year-3 to Year-4 scale - 5+ councils, ~50,000 subscribers, 30+ routes - it becomes worth modelling a mini-sorting hub. Capex would be £500k-£2m, opex £100k-£200k/year, and the trade-off is against the tipping fees we would save. We do not commit to that path in this pilot proposal because the answer depends on data we will only have at the end of Year-1: stream tonnages, contamination rates at our own collection, and the going rate for commercial transfer station capacity in the West Midlands.

10. Implementation roadmap

11. Risks & mitigations

12. References

End of whitepaper · Binly v0.1 · May 2026
© Binly 2026 · Confidential - do not distribute without permission