The box is not the intervention: what monitoring your new driver actually changes

The Central Bank of Ireland's seventh National Claims Information Database report, published on 31 October 2025, puts the average written private motor premium at 623 euro in 2024, up 9% on the year before, against an average claims cost of 397 euro per policy. A nineteen year old with no no-claims bonus pays a large multiple of that. How large, no authoritative Irish source will tell you: the NCID report carries no breakdown of premium by driver age, and every eye-catching young-driver figure on the comparison sites is unsourced marketing. What is fair to say is that telematics is often the difference between insured and not insured.

What is on the Irish market: An Post Insurance BoxClever, underwritten by AIG Europe S.A. and aimed at drivers with zero years of no-claims bonus, records speed, how often you drive, time of day, smoothness of acceleration and braking, sharpness of turns, journey length and total mileage, with quotes by phone only and data shown in a ratemydriving portal inside the customer's MyAIG account. Aviva Ireland fits a smart black box recording speed, braking, acceleration and cornering, with a score in MyAviva that can feed into the renewal price. its4women.ie targets drivers aged 25 and under. Irish consumer guidance from bonkers.ie notes these policies commonly carry annual kilometre limits with top-up charges, that some insurers restrict late-night driving, and that a trained fitter installs the box, typically within a fortnight.

The logic is sound: measure the behaviour, attach a price to it, and the behaviour improves. Insurers are betting on feedback, which is one of the better-supported ideas in skill learning. The question this guide asks is narrower: does the measurement, on its own, change how your child drives? Because that is the version most families are running. The box goes on, the score arrives, and nothing else about the household changes.

ProjectDRIVE, published by Jingzhen Yang and colleagues in JAMA Network Open in 2026, was a three-arm parallel randomised clinical trial of 240 parent-teenager pairs, 80 per arm, enrolled between 28 September 2020 and 30 June 2024, with six months of follow-up made up of three months of active intervention and three months of observation afterwards. One thing to hold in mind from the start, because it limits everything that follows: every teenager in it was 16 or 17, already licensed, and recruited through six Ohio juvenile traffic courts after a moving violation. The trial's own protocol paper calls them high-risk teen drivers. Section 4 deals with what that does to the numbers. Both treatment arms got the same thing: an Azuga telematics system giving real-time audio alerts when the driving crossed a threshold, plus reports every two weeks. The thresholds were kinematic, not judgemental: hard braking at or beyond minus 0.45g, sudden acceleration above 0.35g, and speeding more than 10 mph over the posted limit or above 75 mph. The analysis covered 160,095 trips.

The only difference between the two treatment arms was the parent. In one of them, two weeks after enrolment, the parent received a 50-minute virtual training session from a traffic safety specialist on motivational interviewing technique: open-ended questions, affirmations, reflective listening and summarising, the four skills known as OARS. There was a booster two months later. That is the entire additional intervention. Same box, same beeps, same fortnightly report, same cars.

On the primary outcome, the rate of risky driving events per 1,000 miles, the two arms came apart completely. Feedback plus parent communication training produced an adjusted incidence rate ratio of 0.68, with a 97.5% confidence interval of 0.51 to 0.90, a 32% reduction against control. Feedback alone came in at 0.98, interval 0.74 to 1.31: a flat line, wide enough to include a modest improvement or a modest worsening. The technology was identical. What moved the number was whether anybody in the house had been taught how to have the conversation.

One more finding from the same trial, familiar to any parent of a teenage son. Male teenagers recorded higher rates of hard braking than female teenagers, at an adjusted rate ratio of 1.39 (95% CI 1.11 to 1.73), and much higher rates of sudden acceleration, at 2.13 (95% CI 1.42 to 3.19). Roughly twice the rate of sudden acceleration is not a rounding error. It does not mean your son is destined to crash, and it was measured in one specific population, but if you are deciding where to spend your five minutes of attention, that is where the signal is.

Look at the speeding outcome in the same trial. The proportion of miles driven while speeding was 46% lower with feedback plus parent training (exponentiated beta 0.54, 97.5% CI 0.47 to 0.68) and 36% lower with feedback alone (0.64, 97.5% CI 0.54 to 0.79). So the hardware, with no parental coaching at all, moved speeding a long way. That makes mechanical sense: an alert that fires when you cross a mapped limit is very direct information about a very simple thing. No judgement required.

The composite outcome, risky driving events, is closer to judgement: braking hard because you read the queue late, accelerating hard because you tried to make a gap that was not there. That is what the beeper failed to shift. The honest reading is outcome-specific rather than global. Feedback devices teach rules with clean thresholds well, and teach the anticipation that stops you needing to brake hard badly.

The largest recent trial points the same way. The AAA Foundation for Traffic Safety ran a randomised field trial of smartphone-based feedback with 1,449 drivers recruited nationally in the United States through social media advertisements, across 24 weeks: a six-week baseline, a twelve-week intervention and a six-week follow-up. Three designs were tested against a control: standard usage-based-insurance style weekly messaging, an assigned goal focused on the driver's worst behaviour, and a self-chosen goal. All three treatment groups improved against control on speeding by 11% to 13%, hard braking by 16% to 21% and rapid acceleration by 16% to 25%, and the gains persisted through the six weeks after the feedback stopped. On one behaviour there was nothing. In the report's own words, all three groups improved "but not in handheld phone use".

Start with the population, because it is the biggest problem. Those six Ohio juvenile traffic courts were in Franklin, Delaware, Cuyahoga, Summit, Greene and Warren counties, and every teenager arrived in the trial by way of a moving violation. That is a group with an unusually high baseline of risky driving, so it had unusual room to improve, and it was under court-adjacent pressure to comply. The trial's own limitations section says: "The sample, drawn from 6 Ohio counties, was predominantly White, English-speaking, and well educated, limiting generalizability". A clean-record nineteen year old in Kildare who has just passed is not that person, on those roads, under that pressure.

Then the outcomes. Neither headline trial measured a crash or an injury. ProjectDRIVE's outcome is a telematics-derived event rate over six months; the AAA trial's is app-derived behaviour over 24 weeks. A hard-braking event is not a collision, and the step from "fewer flagged events" to "fewer crashes" is an assumption, not a finding. The same discipline applies here that this site applies to driver training itself: our review of whether driver training actually makes you safer finds no crash advantage for trained drivers, and the Cochrane review of post-licence driver education (Ker and colleagues, CD003734, 24 trials) put the relative risk of a crash at 0.98, 95% CI 0.96 to 1.01, which we set out in our piece on hazard perception training. Plausible mechanism plus improved intermediate measure does not equal fewer collisions. That test has been failed before in this field.

And the smaller studies are smaller than they sound. The famous 61% reduction in flagged events from event-triggered video feedback comes from McGehee and colleagues in the American Journal of Public Health in 2010, measured in 18 newly licensed 16 year olds with no control group, a one-group pretest-posttest design over 52 weeks. Coachable events fell from about 21 to about 8 per 1,000 miles, with a 64% decline among the highest-event drivers. A striking result, and also eighteen teenagers in Iowa who knew they were being filmed. Anyone quoting the 61% without the 18 is selling something.

Charles Farmer, Bevan Kirley and Anne McCartt fitted monitoring devices to the cars of 85 teenage drivers and published the results in the Journal of Safety Research in 2010. The devices detected sudden braking and acceleration, speeding and seat belt non-use, and the teenagers were randomly assigned to four groups differing in whether an alert sounded in the car and whether parents had access to websites recording the violations. On seat belts: "Seat belt use improved when violations were reported to the parent websites, and improved even more when in-vehicle alerts were activated." Note the order. The parent seeing it was the necessary part; the beep was the amplifier.

Speeding was far harder, and their finding is oddly specific: consistent reductions came only when the teenager got the alert, believed a corrected event would not be reported, and knew the parent was getting report cards. A narrow, almost fussy combination, and it says something real: they needed room to correct without being reported, while still knowing the pattern would eventually be seen. The authors' conclusion sums up the whole literature: "Electronic monitoring of teenage drivers can reduce the incidence of risky behavior, especially seat belt nonuse. More complicated behavior is more difficult to change, however."

The Iowa video study shows the same mechanism from its weak side. McGehee's team could not control parental engagement and list that as a limitation. Only 39% of the teenagers, seven of eighteen, reported reviewing every safety event with at least one parent, while 67% discussed the reports at least monthly and 22% weekly. Even in a study built entirely around review, most families were not reviewing most events. That is the gap ProjectDRIVE later closed with fifty minutes of coaching, and it is almost certainly the gap in your house.

A qualitative study published alongside ProjectDRIVE, by Kaur and colleagues in Frontiers in Public Health on 23 March 2026, interviewed 38 parents from the same trial, 82% of them mothers and 84% college educated. One explained the mechanism better than the statistics do: "The feedback was useful. Once again, if we were not involved, she never would've had any feedback other than coming from me. And that's not always what they desire to hear from parents."

The training in ProjectDRIVE was fifty minutes on motivational interviewing, whose four core skills SAMHSA's TIP 35 describes as the basic ones used early and often: open questions, affirmations, reflective listening and summaries, or OARS. Motivational interviewing was developed for counselling, not driving, so importing it into a kitchen conversation is a translation rather than a protocol. The structure survives the translation, and it is short enough to actually do.

Two things make it stick. Pick a fixed time, because a conversation that happens when you are annoyed is a row, and one that happens every Sunday is a routine. And read two data points, not the score. The composite score is designed to be glanceable; the underlying events have times and places attached, and a specific event at a specific junction is something you can actually talk about.

American trials of parent-managed limits, most notably the Checkpoints programme built around a written parent-teenager driving agreement plus persuasive newsletters, sit on top of a legal scaffold Ireland does not have: graduated driver licensing there restricts how many passengers a new driver may carry and when they may drive at night. Checkpoints has been tested in randomised trials. The largest was a state-wide Connecticut trial of 3,743 newly licensed teenagers, which produced significantly greater parent-imposed limits on high-risk driving at licensing and at three and six months afterwards, and no treatment effect on crashes at all. A smaller Connecticut trial of 469 parent-teenager pairs found the stricter limits still in place at twelve months. So the honest summary is that the written agreement reliably changes what parents allow, and has never been shown to change the crash rate. What transfers cleanly is the format: agree the limits in writing, revisit them on a schedule, loosen them as evidence accumulates.

Ireland gives you plenty to build the agreement on, because the legal consequences for a new driver are harsher than most parents realise.

• N plates for two years. Required for the two years after passing, and the plate is a public statement that everyone else on the road can read. Our guide to N plates and novice driver rules covers the detail.
• Seven penalty points, not twelve. A learner permit holder, and a driver during the first two years of a first full licence, is disqualified for six months at 7 points. Everyone else gets to 12. Our penalty points guide sets out which offences carry what.
• Two phone tickets is nearly the whole allowance. A handheld phone offence is 120 euro and 3 points. Two of them puts a novice driver at 6 of 7.
• A 20mg blood alcohol limit. Learner drivers, novice drivers in their first two years, and professional drivers sit under 20mg per 100ml, against 50mg for everyone else. The lower limit follows the driver, not the vehicle, so borrowing your car does not raise it.
• No passenger limit and no curfew in law. Ireland has neither. Three friends in the back at eleven at night is legal for a driver who passed last month. If you want a limit there, you write it.

One more thing to put in the agreement, and it is the item most families skip: what you will look at, and what you will not raise. Agree that you will look at braking and speeding events and will not comment on where the car was parked. That single concession is what converts the system from surveillance into a shared tool, and Section 9 explains why it is worth more than the information you are giving up.

Start with the invisible list. A phone face-up in the lap, read at the lights and glanced at on the move, which the AAA trial's feedback failed to improve even though the app was scoring it. Three friends in the back at eleven at night, the single most reliable risk multiplier for a young driver, and a factor no box records. Four hours of sleep before a seven o'clock start, which degrades exactly the anticipation that keeps braking smooth. One pint on a 20mg limit. And the broadest one: driving that is technically smooth and completely inappropriate for the conditions, a steady 78 km/h on a wet, unlit country road with no verge, generating no flags at all because nothing crossed a threshold.

Then the flags that fire when nothing was wrong. Hard braking is a proxy for the situation as much as for the driver. A segment-level analysis of connected-vehicle telematics published in Vehicles in 2026, matching more than 8.5 million telemetry records and around 45,000 police-reported crashes on New Jersey highways to one-mile segments, found each additional harsh braking event associated with roughly a 1% increase in expected crash frequency for that segment, a positive and statistically significant association. Read that carefully, because it is the sentence people misuse: it is a statement about roads, not about which drivers are dangerous. A run out the Naas Road at half five, or the N81 out of Tallaght in evening traffic, or the M50 approaching the Red Cow, will produce more braking events than the same driver on the same road at ten in the morning. The flag is partly a measurement of Dublin.

Speed flags have their own failure mode. A telematics speed event is your child's recorded speed compared against a mapped speed limit in the provider's database, so if that map is out of date or the limit on that stretch changed, a compliant driver can be recorded as a speeder. Ireland has been through a period of speed-limit changes, which makes this less theoretical than it sounds. It is also the one flag genuinely worth disputing with the insurer, because it is checkable: go and look at the sign.

The adolescent development literature traces a fairly consistent path from perceived privacy invasion to increased secrecy and, as a result, less parental knowledge rather than more. Hawk and colleagues followed 497 Dutch adolescents and both their parents over three years and reported it in Developmental Psychology in 2013: teenagers who felt their privacy was being invaded were the ones whose mothers and fathers reported knowing less about them a year later, with the teenager's own secrecy carrying the effect. Their conclusion is that privacy invasion is counterproductive to parents' efforts to stay informed. We are deliberately not converting that into a percentage for your house, and a later within-person reanalysis of the same pathway by Dietvorst and colleagues found the association behaves differently within a single family than it does across families, so treat this as a direction rather than a dose. The direction is well established enough to act on: covert tracking tends to buy you less information, not more. Which is convenient, because it points the same way as the trial evidence. What works is the conversation, and a conversation requires the other person to know the data exists.

On the legal side, two separate things get blurred constantly. A family app used for family reasons sits within what Ireland's Data Protection Commission calls the household exemption, reflecting Article 2(2)(c) of the GDPR: data protection law does not apply where an individual keeps personal data solely for their own personal, family or household affairs, provided it is not used for a professional or commercial activity and not published. The insurer's processing is a different animal. That is commercial and squarely inside the GDPR, and the European Data Protection Board's Guidelines 01/2020 on connected vehicles, version 2.0 adopted on 9 March 2021, treat the driving data a telematics box generates, mileage, braking pattern, rapid acceleration and the rest, as personal data about the driver in a pay-as-you-drive context. Two legal bases sit on top of each other there, and they are worth separating. Reaching the data stored in the car needs your consent, under Article 5(3) of the ePrivacy Directive. The insurer's processing of it afterwards can rest on performance of your contract, Article 6(1)(b) GDPR, which is what you signed. The EDPB also says the product must always be optional, that a non-telematics policy has to be available, and that the insurer should ideally receive scores rather than raw journey data. So the household exemption covers your family app. It is not a shield for anything the insurer does.

One structural point surprises most parents. In the Irish products we could read, the portal belongs to the policyholder, not to the parent. An Post Insurance shows the data through the customer's MyAIG ratemydriving portal; Aviva shows the score in MyAviva. its4women.ie states that only authorised personnel can access telematics data and that it is never shared with third parties without consent unless legally required, and says nothing about parental access. So if your child is the policyholder, you see the data because they show it to you. That is not a loophole to work around. It is the argument of this whole guide arriving as a practical fact: the conversation has to be consented to, because the data flows through your child. The insurance side of the decision is covered in our guide to new driver insurance in Ireland.

1. Pick the five minutes and put them in your phone. A fixed weekly slot, not "when it comes up". The tested version was under an hour of coaching plus a routine. You supply the routine.
2. Read two data points, never the score. Two specific events with times and places attached.
3. Tell them you are looking, and exactly what at. "I'll look at braking and speeding events. I won't comment on where you were." Say the second half out loud, because it is the half that buys their cooperation.
4. Write the agreement down. Passengers, night driving, phone in the glovebox before the engine starts, and a review date. Irish law supplies none of these limits, so the page on your fridge is the only place they exist.
5. Find the cancellation, mileage and night-driving clauses in the policy booklet. Twenty minutes with the document now, rather than a phone call about a cancelled policy later.
6. Book one assessment lesson at around six months. Not because the box failed, but because a flagged event is a symptom and an instructor beside the driver can find the cause.

On that last one, briefly and without a sales pitch. The box tells you a hard-braking event happened at 17:32 on a Tuesday. It cannot tell you that the driver was looking at the bumper of the car ahead instead of through the junction to the queue beyond it, and it cannot fix that. An hour or two with an ADI, with three flagged events as the agenda, turns a list of numbers into two specific habits. We make no promise about your child's score and attach no pass-rate claim to it, because that would be exactly the overselling the rest of this article argues against. For the wider version of how to support a new driver, our page for parents and sponsors and our piece on whether to buy your teenager their own car are the companions to this one.