Research and partnerships

This page is the entry point for researchers, insurers, fleet operators, and public-sector environmental programmes who want to understand what Odoma measures, how it can plug into a study or a pilot, and what data is — and is not — available for sharing.

For the technical detail on detection and scoring, see the Methodology. For live aggregate impact metrics, see /impact.

EU policy and regulatory context

  • EU 2035 zero-emissions target for new cars and vans — European Commission, Cars and vans. The technical pathway is fleet electrification; the behavioural pathway — driving style — remains uncovered by the regulation, which is exactly the gap Odoma addresses.
  • EEA — Sustainability of Europe's mobility systems 2024 — eea.europa.eu. The reference document on whether the EU is on track for transport decarbonisation. Spoiler: not yet.
  • EEA — Average CO₂ emissions from new passenger cars and future targets — eea.europa.eu.
  • EEA — Decomposition analysis of EU-27 passenger-car CO₂ 2000–2023 — eea.europa.eu. Quantifies how much of the change came from technology vs activity vs intensity.
  • GDPR (Regulation EU 2016/679) — eur-lex.europa.eu. All Odoma data flows are designed to satisfy Articles 5, 6(1)(a), and 25. See §8 of the Methodology.

Eco-driving — peer-reviewed and EU-funded research

  • Sanguinetti, A. et al. (2020) — Average impact and important features of onboard eco-driving feedback: A meta-analysis. Transportation Research Part F 70, 80–90. The single most useful synthesis: across 17 studies, average improvement is 6.6 %. doi · open-access · local mirror
  • ecoDriver (FP7, 2011–2016) — €14.6 M, 170 drivers, 7 countries. The benchmark for GPS-only smartphone applications: 2.5 % fuel savings, OBD systems reach 6 %. CORDIS 288611
  • GamECAR (H2020, 2016–2019) — €1 M, 36 drivers, Spain. Demonstrated 0.59 L/100 km savings with eco-score, leaderboard, and challenge mechanics (p = 0.004, Cohen's d = 1.026 — large effect). CORDIS 732068
  • Baumgartner, M. et al. (2019) — Long-term effects of gamification on eco-driving behaviour. Energy Research & Social Science 57. 22-month BEV study (n = 108) showing gamification builds transferable skill. doi

Telematics, harsh-event detection, and crash-risk research

  • NJIT / MDPI Vehicles 2025 — Evaluating Harsh Braking Events as a Surrogate Measure of Crash Risk Using Connected-Vehicle Telematics. The reference paper on why harsh braking is the strongest crash-risk predictor in telematics. Industry thresholds range 0.20 g – 0.45 g. mdpi.com
  • ETSC 2019 — Telematics and Road Safety. European Transport Safety Council. The canonical EU-policy document on how telematics signals translate to measurable safety outcomes. etsc.eu · local mirror

Insurance and UBI market context

The same signals Odoma collects (acceleration, braking, mileage, time-of-day) underpin the European usage-based insurance market. Useful market references:

  • Berg Insight 2025 — Insurance Telematics in Europe and North America, 9th edition. European UBI policies: 13.8 M at end-2024, 20.1 M projected by 2029 (CAGR 7.8 %). berginsight.com
  • PTOLEMUS Consulting Group 2025 — UBI solution provider ranking. Lists IMS, Octo, The Floow, Cambridge Mobile Telematics, Targa Telematics, Munic, Geotab, Vodafone Automotive, Arity, FairConnect, Dolphin Technologies, Redtail Telematics. ptolemus.com
  • Mordor Intelligence 2025 — Europe Insurance Telematics Market Size. Market projected from USD 0.97 B (2025) to USD 2.78 B (2030). mordorintelligence.com
  • Database / DataBridge Market Research 2025 — Europe Usage-Based Insurance Market. databridgemarketresearch.com

Odoma is not an insurance product. We publish the methodology and dataset schema in the open so that insurers and fleet operators can audit it before any pilot conversation.

For external research and pilot use, Odoma offers an opt-in, two-tier crowd-sourced dataset:

  • Tier A — non-longitudinal, anonymous by default. Per-trip aggregate record (no coordinates, no exact timestamps, all numeric features bucketed into discrete bins). Default opt-in available in Settings.
  • Tier B — longitudinal, explicit consent. Adds a device-stored pseudo_id (Keychain on iOS, app-private storage on Android) so that multiple trips from the same device can be linked. Required for studies of within-driver behaviour change. Withdrawable at any time; pseudo_id rotation supported.

Architectural invariants:

  • No GPS coordinates ever leave the device. Only bucketed aggregates do.
  • No exact timestamps. Only time_of_day_bucket (4 bins) and day_of_week.
  • No raw streams. Speeds and event rates are reported as multi-threshold CDFs.
  • No direct identifiers. No BT MAC, no BT name, no phone model, no VIN, no email.
  • K-anonymity guard at backend with auto-promoted k = 20 for small countries (EE / LV / LT / LU / CY / MT etc.) to prevent re-identification.
  • Region-aware data residency. EU users land in an EU Firestore project; non-EU users land in a US project. ISO 3166-1 country code drives the routing.
  • Schema versioning is additive. New fields are nullable adds within upload_schema_v: 1; any breaking semantic change bumps to v2.

The full schema and the rules governing it are public. (Documents are in the codebase repository under docs/dataset_schema.md and memory/crowd_dataset_schema_rules.md; we will publish the rendered HTML on this page once the uplink ships in v1.4.)

Driver-attribution and passenger mode

A practical issue for any phone-based driving telematics is what happens when the phone is in a taxi, on public transport, or in the passenger seat of a friend's car. The same accelerometer signal arrives, but it is no longer the holder's behaviour.

Odoma classifies non-self trips with three confidence tiers:

  • High confidence — public transport pattern. Eight-feature heuristic (stops-per-km, duration variance, speed P95 ≤ 80, route linearity, etc.). Silently classified as .publicTransport.
  • Medium confidence — style mismatch. Z-score against the user's per-driver baseline, off-hours, BT mismatch. Classified as .passenger with a non-modal hint.
  • Low confidence — single signal. Defaults to .self and prompts the user to confirm in Trip Detail.

For all non-.self trips, eco / safety / driver-loss aggregates are not computed and not displayed, and the trip is excluded from the crowd-sourced uplink. This is enforced as a hard guard, not a UI suggestion.

A Travel Mode auto-activates after multiple signals (foreign country / multiple no-BT trips / multiple public-transport matches / no-saved-geofence starts) and biases classification toward .passenger for 7 days, suspending eco / safety aggregates for the period.

The full design — eight ethical guardrails, classifier signals, override actions, and the off-roadmap Tier C (taxi/transit research with separate consent) — is documented internally and will be summarised here on uplink launch.

Working with us

We're open to:

  • Pilot partnerships — schools, fleet operators, environmental NGOs, universities, public-sector environmental programmes, and insurers interested in field-validating the methodology.
  • Co-funding — to scale beyond Estonia and from one product to a portfolio aligned with our six statutory aims.
  • Volunteers — translators, designers, environmental specialists, lawyers, communicators.

For partnership enquiries:

See also: About Odoma · Methodology