You submitted through Connect Direct because it was meant to be the quick route. Weeks later the job is still with the DNO, the customer has rung twice, and the scaffold is booked.
That is not you filling the form in badly. On the only published figures anyone has, it is the normal outcome.
The two numbers
Renbee, who sell DNO application software, published a response to Ofgem's ED3 consultation in November 2025. Drawing on their own platform's sample data over a 12-month period, they report:
Over 60% of applications submitted via Connect Direct are escalated for manual review, while only 0.6% are automatically approved.
Take the interest into account. Renbee sell the thing that sits between you and this process, and their sample is their own platform, not the whole market. It is still the only number of its kind on the public record. Connect Direct is the Energy Networks Association's single front door for connection applications. Fewer than one application in a hundred gets through that door without a person picking it up.
Notice what those two figures do not do: they do not add up. Over 60% escalated, 0.6% automatically approved, and Renbee say nothing about the balance. That silence is worth holding onto, because "not escalated" is not the same as "approved". Their own subheading for the section is blunt enough: "Connect Direct: a good idea, poorly executed".
Escalation means a queue, and the queue depends on your DNO
Once a human has it, the clock is theirs. Renbee again:
Response times also vary significantly from 2 days (UKPN) to over 50 days (Last Mile) with some 'fast-track' processes taking as long as full applications elsewhere.
Two days at one operator. Over fifty at another. Same form, same standard, same installer. And a fast-track lane that in some cases runs no faster than another DNO's ordinary one. Renbee report that "installers routinely face delays exceeding 100 days, causing customer frustration and project backlogs".
Installers have been saying the same thing to each other. The Only Panels podcast ran "Rejected G99s, DNO Delays & the Real Cost to Installers" in August 2025. By June 2026 the episode was called "Are DNOs Forcing Installers to Break the Law?", citing "the crippling 45-working-day wait times holding the industry back", "specific struggles with Northern Power Grid", and "why Connect Direct is failing installers".
That same episode names the part that really stings — "the DNO double standard: why heat pump approvals are highly automated and practically instant, while solar PV applications are stuck in the dark ages". The automation exists. It is pointed at somebody else's technology.
The cut-out reader that cannot read cut-outs
The ENA's answer to manual review was an AI-enabled cut-out reader. Renbee's assessment:
Tools such as the 'AI-enabled cut-out reader' often add time and cost without delivering meaningful benefits. Despite significant investment by the ENA, the tool fails to read cut-out ratings, and fails applications especially when the cut-out is partially obscured, located in tight spaces (garages, under-stairs cupboards), or poorly lit. This leads to installers having to revisit the site for new photos…
Read the failure conditions again, because the failure conditions are simply the conditions. Cut-outs live under stairs. They live in garages, behind a chest freezer. They are rarely well lit and almost never unobscured. A tool that fails on obscured, cramped and badly lit cut-outs fails on cut-outs.
And the cost of that failure lands on you: a second visit, in your van, on your day, to rephotograph something you already photographed.
What the process is producing
Here are two installers, quoted anonymously in Renbee's submission.
"We'll apply for 3.68kW even though we're installing 5kW — the customer wants maximum export."
"We just use the cut-out rating as max demand — it always goes through."
Renbee's reading of them is the fair one:
These comments reflect the confusion and procedural fatigue faced by installers, more than malice. They are being asked for new data points, such as estimated maximum demand post-installation or the cut out rating, without sufficient guidance or training to work this out.
Fatigue, not malice. That is right, and it does not make the first quote survivable. Applying for less capacity than you are installing is a false declaration on a connection application, and the installer who signs it carries it — not the DNO, not the customer. Do not adopt it. The second quote is a softer failure: using the cut-out rating as maximum demand is a guess rather than a falsehood, but it is a guess wearing the clothes of a figure, and it is being made because nobody has told the installer how to work the number out properly.
Renbee, from their work with hundreds of installers across the UK, put a scale on this. Compliance rates, they say, "may be as low as 50%". Some installers "are completing thousands of installations without notifying the relevant DNO; others apply creative 'hacks' to make the process workable".
A process that roughly half its participants have quietly stopped using is not suffering from a compliance problem. It is suffering from a design problem, and the design is not yours.
What you can control
You cannot fix Connect Direct, and you cannot make Northern Power Grid answer faster. Four things sit inside your control.
Photograph the cut-out as though a machine will read it and a human will overrule it. Straight on, not at an angle. The full faceplate and the rating in one frame. Move the freezer. Carry a torch. Take three shots from three positions rather than one and hope. You cannot fix the reader; you can remove its excuses. Every re-visit is a day you do not get paid for.
Treat estimated maximum demand as a calculation, not a field. The cut-out rating is not the maximum demand. Renbee's point stands — installers are being asked for post-installation maximum demand without guidance on how to derive it — but the answer to bad guidance is a method you can show your working on, not a number that has historically gone through.
Get the cut-out rating right the first time. It is the single data point the automated route most often fails on, and it is the one most likely to trigger a manual review or a rejection you then queue behind.
Keep the escalation in mind while you prepare. On these numbers, escalation is the base case rather than the exception. The person who eventually opens your application is a person. Give them something clean, internally consistent, and legible, and you are not waiting on a clarification email that adds a fortnight.
The longer argument — the one behind Solar Engine — is that estimated maximum demand and cut-out rating should fall out of the design rather than be typed into a box. A system that already holds the array, the inverter and the property holds most of what the application is asking for. Nobody should be entering 3.68 because it goes through.