Dashboard · Research

Energy Plus NY · Stage 01

Best-practice base for the post system

Everything the template system has to obey before a single artboard is drawn: what canvas to build on, which post formats actually earn distribution, how often to publish, how many message types the system needs, and how to keep a repeatable template from going invisible in the feed.

Scope · desk research, Sept 2026 Status · complete, pending external review Next · roadmap

01 What this settles

Six findings carry real weight for a client who has to get maximum visibility out of minimum spend. Each is stated as a decision, with the evidence that backs it in the section that follows.

QuestionDecisionConfidence
One canvas or many? One: 1080 × 1350 (4:5 portrait). Native on LinkedIn, Facebook and Instagram feeds. Zero re-layout per network. High — consistent across every 2026 spec source
Which format wins? Multi-slide document posts. ~4.4–6.6% engagement vs ~1.2–4% for single images; 3.4× the reach. High — million-post datasets, two independent benchmarks
How often? 3–5 posts a week, never two in 24h. LinkedIn surfaces roughly one post per account per day. High — repeated across 2026 algorithm guides
Company page or people? People carry it. Personal profiles out-reach company pages ~5.6×; employee shares multiply page reach ~8×. High — LinkedIn's own advocacy data + Sprout Q1 2026
How many templates? Five message types × nine chapters, with the chapter colour confined to a badge. Eleven things built, forty combinations out. Settled — chapters taken from the client's own light master document
How to stay visible? Two invariants, three variables. Fixed lockup and rule; alternating surface tone, image presence and type scale. Medium — design principle, to be validated on prototypes

02 The canvas

The instinct to build one artboard that never needs adapting is correct, and the format that does it is 1080 × 1350 px, 4:5 portrait. It is the recommended LinkedIn feed image size, the native Instagram feed size, and a supported Facebook portrait size — the same file posts to all three without a crop. It also fills more of a phone screen than a square, which is the whole reason it out-performs 1:1.

1080 × 1350
4:5 — the canvas
everything is built on
1080 × 1080
1:1 — fallback only,
loses vertical screen
1200 × 627
1.91:1 — link previews
and nothing else

Rules the canvas imposes

Decision

Build the entire system on a single 1080 × 1350 artboard. Produce 1200 × 627 only when a post's job is to carry a link preview, and treat it as a derived crop, not a designed variant.

03 Format hierarchy

Format choice moves distribution more than any other single lever, because LinkedIn's 2026 ranking leans hardest on dwell time — how long a post holds a thumb. Formats that take longer to consume get distributed further, and the gap is not marginal.

6.6%
Carousel engagement ratevs 3–4% single image
3.4×
Reach, multi-slide vs single image1M+ posts, 9,000 pages
15–20s
Carousel dwell timevs 8–10s for one image
5.72%
Median document-post engagement1,000+ B2B pages, Mar 2026
FormatWhy it ranks hereCost to produce
1 Document carousel
PDF, 5–15 slides
Highest engagement of any native format. Doubles dwell time. Swipes count as interaction. Highest — but it is the format templates help with most
2 Vertical video
site footage, 30–60s
~84% higher engagement; explicitly prioritised in the 2026 update. High if produced; near-zero if it is phone footage from a job site
3 Text-only
with a strong first line
Cheapest post that still performs. Carries commentary, opinion, quick program news. Lowest — no design step at all
4 Single image Weakest of the four, except where one frame is the whole message — an announcement, a finished building, a number. Low

There is a second-order effect worth designing around: accounts that post carousels predominantly get a novelty lift when they switch to a single image. Format rotation is itself a distribution tactic, which is convenient — it is also the answer to feed fatigue.

Decision

The template system's centre of gravity is the document carousel, not the single image. A carousel is where a template pays for itself: one built skeleton produces 5–15 slides.

Single-image templates still ship, but as the rotation format and the announcement format — roughly one in four posts, not the default.

04 Distribution

Cadence

Three to five posts a week outperforms daily posting. Moving from one a week to two-plus triggers noticeably broader distribution, but a second post inside 24 hours cannibalises the first — LinkedIn generally surfaces one post per account per day. The prime window is Tuesday and Wednesday, 8–9 AM, with Tuesday running about 20% above Friday.

Who publishes

This is the finding with the largest gap between effort and result, and it deserves to be stated plainly: company pages have collapsed as a distribution channel. Page reach fell 60–66% between 2024 and 2026. Personal profiles out-reach pages by roughly 5.6×, with median engagement around 4.7% against 1–2% for pages.

5.6×
Personal profile reach vs company page
8×
Reach multiplier when staff share a page postLinkedIn advocacy data
3%
Of employees ever share — and they generate ~30% of engagement

The brief is explicit that there is no split between the company voice and individual voices, and that is the right call for a small team with turnover — a system keyed to named people breaks the moment someone leaves. But the reach data still has to be honoured, and it can be, without introducing a second visual register: the artwork is company-issue and unsigned; the distribution is human. The same post, in the same system, is published from a person's profile and amplified by the rest of the team. Nothing on the artboard says who is speaking. Everything about where it appears does.

Decision

One visual register, no personal-versus-corporate split on the artboard. Publish primarily from personal profiles with the company page as mirror and archive, and build a share step into the weekly routine rather than a separate design track.

05 Message taxonomy

Published pillar mixes cluster tightly: roughly 40–60% educational, 20–25% perspective and storytelling, 10–20% proof, 10% soft call-to-action. Mapped onto what Energy Plus actually has to talk about — nine utility and state programs, eight qualifying measures, 500+ completed buildings, fifteen years of field work — that produces five message types. Five is the whole set. It is the "minimum gentleman's kit" the brief asks for, and it is enough to fill a year.

TypeThe jobNative formatShare
ExplainerHow a program works, what qualifies, what the paperwork actually involves. The category-education engine.Carousel40%
ProofA completed building: address, scope, measures, outcome. The 500+ portfolio is the moat and it is barely visible online.Carousel or image20%
NumberOne figure that stops a thumb — an incentive tier, a deadline, a program cap. Typographic, no photo.Single image15%
SignalProgram changes, funding rounds, rule updates. Timely, cheap, and the reason to follow the account at all.Text-only15%
FieldThe work itself — a roof, a heat pump going in, a crew. Phone footage, unpolished on purpose.Video or image10%

The mix is deliberately weighted toward Explainer, because that is where the client's structural advantage lies. Nine direct program relationships is a genuinely rare position, and nobody in the target audience — building owners, GCs, engineers, mechanical and trade contractors — understands the incentive landscape well enough not to need it explained. Explainers are also the cheapest thing to produce at volume once a carousel template exists.

06 Differentiation

The brief asks for differentiation by service line and by message type, and the trap is obvious: eight measures × five message types is forty template variants, which no small team will ever maintain. The system needs to differentiate on three independent axes that combine, rather than on one flat list that multiplies.

AXIS 01Service chapter — carried by a badge

The client already has this axis and has never written it down. The light master document divides the business into nine chapters, each with its own hue and its own icon, presented as a rotated square — a diamond that echoes the geometry of the Energy Plus mark. Nothing needs inventing; it needs formalising.

01 Rebates & Incentives 02 Financing & Tax 03 Efficiency Upgrades 04 Building Envelope 05 Clean Heat 06 Firestopping 07 Plumbing Insulation 08 EV Charging 09 Local Law Compliance

Nine colours would normally be far too many for a coherent feed. They survive here because of a hard restriction: the chapter hue appears only inside the badge and in one hairline. It never becomes a ground, never tints a photograph, never colours a headline. The post reads as Energy Plus first and as a service second, which is the correct order.

The chapters also arrive with their own sub-items already written — 2.2 NYC Programs → J-51, ICAP, PACE, 4.x seven envelope measures — which is a content calendar the client has effectively already paid for.

AXIS 02Message type — carried by layout skeleton

The five message types get five layout skeletons, not five palettes. An Explainer is a numbered sequence; a Proof is photo-led with an address knocked out of the image, reusing the device already built for the brochure project pages; a Number is typographic with the figure at display size; a Signal is a text card; a Field post is photograph-first with almost no type. Colour stays free to say something else.

AXIS 03Surface tone — carried by light / dark

Every skeleton exists on a pine-dark surface and on a paper-light surface. This is the cheapest variety in the system: it doubles the apparent number of templates at no design cost, and it is the single most effective defence against a feed that starts looking like wallpaper.

Five skeletons × two surface tones is ten built artefacts, plus one badge component that swaps colour and glyph. Eleven things to build; every chapter and every message type addressable. That is the whole argument for a system rather than a folder of posts.

07 Banner blindness

A recognisable template is the goal and the hazard at once. Recognition is what makes a feed read as one voice; sameness is what makes an audience stop seeing it. The resolution is not to vary everything — that destroys the system — but to fix a small number of elements absolutely and let everything else move.

Fixed, always — the recognition layer

Variable, deliberately — the attention layer

This gives a simple production rule that a non-designer can follow without judgement calls: never publish two consecutive posts that share tone, image presence and format. One shared attribute is fine, two is a warning, three is wallpaper.

08 Production model

The economics here are unforgiving — the client is paying attention to every line item — so the system has to be cheap to run, not just cheap to build. The published guidance on batching is unanimous and matches what the format hierarchy implies: produce in sessions, not daily. Six to eight feed posts per session is the standard target for a small operation, which at 3–5 posts a week is one session roughly every two weeks.

Watch this one

Posts carrying recognisable AI-writing patterns are reported to lose about 47% of organic reach. Any drafting assistance in the pipeline must be followed by a genuine human rewrite — this is a constraint on the workflow we design, not an afterthought.

09 Risks

RiskWhat it means for the system
Utility and program marksCon Edison, National Grid, NYSERDA, PSEG and the rest are third-party marks. The brochures already carry the correct disclaimer line. Any template that displays program logos must carry it too, and the templates should make that automatic rather than optional.
Incentive claimsProgram terms, caps and eligibility change. Numbers on an artboard outlive the program that produced them. Every Number template needs a visible date or program-cycle reference, or the system manufactures its own liability.
Single-operator dependencyOne person will run this. If a post requires design judgement, it will not ship. The templates have to be fillable without taste — fixed slots, fixed type sizes, no free canvas.
Format driftCarousel slide ratios must not vary. A single 1:1 slide dropped into a 4:5 document crops the entire deck. Worth a hard check in whatever export tooling we build.
Budget attritionThe scope has to survive a client who cuts. Sequence the build so that the first deliverable — one carousel skeleton and one Number template — is already useful on its own if nothing else ever gets made.

10 Open questions

These are the points where desk research runs out and the external review has something to add. They are the brief for the next stage.

  1. Is 4:5 genuinely safe as a single canvas, or does LinkedIn's desktop feed punish it?Every spec source recommends it for mobile. Energy Plus sells to building owners and GCs, an audience that may well read LinkedIn on a desktop, where portrait crops differently.
  2. Does a document carousel actually beat a single image for a nine-program contractor, or is that benchmark driven by SaaS marketing pages?The million-post datasets skew heavily toward software companies. Worth challenging before we make the carousel the centre of the system.
  3. Five message types — too many or too few?Our synthesis. An outside read may collapse Number into Signal, or argue that Field deserves more than 10%.
  4. Do nine chapter colours survive a real feed, or does the badge need to drop to four families?The restriction that keeps the hue inside the badge is what makes nine workable. Worth testing against an outside read before the whole template set is built on it.
  5. What is the realistic weekly time cost for one non-designer running this?Determines whether the roadmap ships a template pack, a generator, or both.
  6. Which incentive claims are safe to put on an artboard at all?New York and New Jersey program rules, and what a contractor may state publicly about rebate values.

11 Sources