Runbook is one of the better-built pages we have scanned, and it is held back by the one thing a page cannot afford to get wrong: a stranger cannot tell who it is for. The hero says AI agents that run your physical operations, but physical operations covers manufacturing, warehousing, home services, construction and freight, and the page does not narrow it. Scrolling makes it worse before it makes it better, because the photography moves through truckers, field and construction workers and warehouse staff without telling you which one you are, and the two industries Runbook actually serves, field services and trucking, live only in the navigation. By the time the work we automate section lists scheduling and AP or AR, the reader has spent real effort guessing. On top of the positioning gap, no customer is named anywhere and every number is self-asserted, so the strongest claims have nothing behind them. The craft is genuinely high. The job now is to say who this is for and prove it works.
Scored against 26 checks in three weighted categories,
plus 7 signals measured directly from the page. Section 01 shows how the score is built.
Section 01
Verdict
The bottom line
A beautifully built page that never answers the first question a visitor asks: run my physical operations, but which operation, and is this for me? A stranger stays confused about who it is for, and no customer is named to prove it works.
01
Who is it for
Run your physical operations never says which operation or which buyer. A visitor cannot place themselves in the first five seconds, and scrolling past truckers, construction crews and warehouse staff widens the question rather than answering it.
02
Show the product
The product is never really shown. It is lifestyle photography with simulated chat bubbles on top, no real interface, no case study, and every number is self-asserted with nothing behind it.
03
The weight
13.7 MB over 55 requests, against a 2.5 MB budget, for a product whose field services and trucking users are frequently on a phone on a job site.
04
What is strong
High craft, one clear call to action, and SOC2, SSO and encryption answered on the page. The making is not the problem, the message is.
The first five seconds
What a stranger takes from the hero
Lands
That it is AI agents doing operational work, and that it is fast to go live. The mechanism and the speed come through.
Does not land
Who it is for, which operation, and what it actually is. Physical operations could mean manufacturing, warehousing or freight, the imagery spans several unrelated job sites, field services and trucking sit only in the nav, and the product is never shown, so a stranger leaves unsure this is for them.
Where it stands
Three categories, 16 findings
The summary. Every finding is detailed below
1573
1 critical5 high7 medium3 low
UX and conversion
80/100 · Solid
This is where the page is losing the most. The hero cannot tell a stranger who it is for or what the product actually is, and physical operations is broad enough to fit a dozen industries. On top of that there is no proof: no named customer, no case study, and no source behind any number. The call to action and metadata are handled well, but a visitor has to understand and believe the product before the call to action matters, and right now they do neither in the seconds they give it.
Visual and behavioral
90/100 · Strong
High craft pointed in the wrong direction. The type, palette and photography are all well made, but the imagery sets a mood instead of showing the product, and there is no real screenshot of the interface anywhere. The prime post-hero slot is spent on integration logos dressed as customer proof, and eleven mobile tap targets fall under the minimum for an audience that lives on a phone. The skill is there, it needs to be turned on the product.
Compliance and trust
95/100 · Strong
In good shape overall. Privacy and terms exist and resolve, robots and sitemap are present, and no trackers run without consent. The open items are a missing cookie policy, a missing accessibility statement, an absent HSTS header, and a primary button that is not keyboard reachable. All are quick, and none are the live regulatory exposure we often find here.
Detail
The longer read
What this site is. Runbook sells AI agents that run physical operations end to end, inside the systems a company already uses, including ones without APIs. It targets operations teams in field services and trucking, automating appointment scheduling, accounts payable and receivable, order management, status updates and vendor coordination. It is pitched on speed to live (days, not a quarter) and on being an agent that does the work rather than a copilot that assists. The single conversion goal on the page is Book a Demo.
This is a high-craft page failing the test that matters most: a stranger cannot tell who it is for or what the product actually is in the few seconds they give it. Run your physical operations sounds specific but is not, the photography moves through truckers, construction crews and warehouse staff without naming a buyer, and the two industries Runbook serves live only in the navigation. The product itself is never shown. What stands in for it is lifestyle imagery with simulated chat bubbles, and there is no real interface, no case study, and no source behind any of the numbers. For a buyer being asked to hand a live workflow to an AI agent, who is this for, what does it look like, and who else runs it are the first three questions, and the page answers none of them. The design, the single call to action and the security section are genuinely good, which is what makes this fixable: the making is done, the message and the proof are not. Name the buyer, show the product, and add one real customer, and this page competes with anyone.
The scoring framework
Built over years of B2B and B2C product work
This is our own framework, not an off-the-shelf checklist. Every check in it earned its place
by turning up repeatedly in real reviews, across products sold to businesses and products
sold to consumers, where the same failure kept costing the same thing. Checks that stopped
predicting anything were removed.
The scoring is deliberately mechanical, so the number does not move with who ran the review
or what kind of day they were having. What stays human is the judgement: what counts as a
finding, how severe it is, and what to do about it. We use AI for the measurement pass now,
which catches the small things a person skims past on the fortieth screen, and that buys back
the time for the part that actually needs a designer.
How the score is built
Every check in the framework
Each check starts at 100 and is reduced by its most severe finding. A category is the average of
its checks. The overall score weights the three categories at 40, 35, and 25 percent, then takes
five points for every unresolved critical finding, because a critical is a business consequence
that strong scores elsewhere should not wash out.
Rewrite the hero to name the buyer and the category, and put field services and trucking on the page
Lets a stranger place themselves in the first few seconds instead of guessing. The moment the reader knows this is for their operation and what it is, every other strength on the page starts working. This is the single highest-return change.
02
Show the actual product, and add one real customer and case study
Replaces mood photography and self-asserted numbers with the two things a software buyer needs, seeing the tool and seeing that someone real runs it. A real screen, one genuine message thread, and a before and after number do more than any hero photo.
03
Fix the mechanics: cut the page toward 3 MB and make Book a Demo a real, keyboard-reachable button
Speeds the page for the mobile, in-the-field audience the product is built for, and recovers a primary call to action that a keyboard or screen reader user currently cannot trigger. Both are quick and mostly build-side.
This is where it starts, not where it ends
Before you act on these
These three are the order, not the plan
Sequencing three changes is straightforward. Knowing which one to spend a sprint on, what the replacement copy should actually say, and how to tell within a fortnight whether it worked is a different question, and it depends on numbers this report cannot see.
That is the part we would want a conversation for. Thirty minutes, your funnel data on screen, no pitch.
Book a Demo appears in the nav, the hero, and the footer, in the same treatment each time. There is no second primary button pulling attention in a different direction.
Keep it because: A single, consistent conversion goal is the cleanest way to move a visitor. Nothing here dilutes the intended next step, which is rarer than it should be.
The enterprise objections are answered on the page
A dedicated section, Your data. Your rules. Our guardrails, carries SOC2 Type I, SOC2 Type II, SSO Enforced, third-party audit and encryption at rest, and the no rip and replace, works even without APIs message addresses the integration fear directly.
Keep it because: Security and integration risk are the two objections that stall this exact buyer. Handling them before they are asked is what keeps a serious evaluation moving.
The visual craft is high and reads as serious
Deep navy, a disciplined mint accent used only on the key phrases, serif display type, and well-shot photography. Nothing on the page looks cheap or improvised.
Keep it because: The register tells an operations leader the vendor is serious before they read a word, which is the right instinct. Keep the craft and point it at the product, so the same skill is showing what the tool does rather than setting a mood.
Metadata and page speed are handled
A real title, Runbook: Autonomous Operations Platform, a written meta description, an Open Graph image, and a first contentful paint at 456ms with zero console errors.
Keep it because: The link previews properly when shared and the page paints fast. These are the quiet basics that many otherwise polished sites still miss.
Section 04
What is not working
16 findings, most severe first. Each is grounded in the captured page.
UX and conversion
5 issue(s) · scores 80/100
A stranger cannot tell who this is for, and stays confused after scrolling
hero and the first several sections, desktop and mobile · value prop clarity
criticalobservedmedium
Run your physical operations sounds specific but is not. Physical operations spans manufacturing, warehousing, home services, construction and freight, and the hero never narrows it, so a first-time visitor cannot place themselves. Scrolling makes it harder before it makes it easier: the photography moves through truckers, field and construction workers and warehouse staff without signalling which of them is the buyer, and the two industries Runbook actually serves, field services and trucking, appear only in the navigation and footer. Even the category is fuzzy. Is this workflow automation, ops management, or an AI agent platform? Only when the work we automate section lists appointment scheduling and AP or AR does the reader start to reconstruct it, by which point they have done the work the page should have done for them. Compounding the confusion, the product is almost never shown, so there is nothing concrete to anchor the abstract claim to.
Evidence
Headline: "AI Agents that run your physical operations" with no industry, role, or category named
Imagery spans trucking, construction and field work, and warehouse without naming the buyer, which widens rather than narrows the question
Industries named only in nav and footer: "Field Services", "Trucking"
Category label "Autonomous Operations Platform" lives in the title tag, not on the page
The specific pitch only surfaces well down the page: "The work we automate" (scheduling, AP, AR, order management)
Why it matters
This is the core job of the page and it is failing the five second test. If a field services or trucking operations lead cannot recognise themselves and cannot tell what the product actually is, the strong design, the security section and the speed claim never get read, because the reader has already decided they are not sure this is for them. Everything else in this report is secondary to fixing this.
Fix
Name the buyer and the category in the hero. Something close to: For field services and trucking operations, AI agents that run scheduling, dispatch and billing end to end, inside the systems you already use. Put field services and trucking on the page, not just in the nav, and lead each industry with the specific jobs it automates. Show one concrete workflow in the hero so the abstract claim has something real behind it.
No customer, no case study, and no real product shown, while every number is self-asserted
whole page · trust signals
highobservedmedium
There is no customer name, no testimonial, no case study, no named deployment, and no example of the product actually being used. Every figure is stated by Runbook about itself with no source: 70%, 99%, ~2 weeks, 90% automation, and 95%+ accuracy. The one logo row is labelled Integrates With Your Existing Tools and shows Oracle, SAP, Salesforce, ServiceMax, Corrigo, ServiceChannel and Microsoft, which are systems it connects to, not customers, though the placement reads as proof at a glance. The only glimpse of the product is stylised chat bubbles composited over lifestyle photos, so a buyer cannot see the real interface, where the agent's messages actually live, or a single account of someone who has run a workflow through it.
Evidence
No testimonial, customer quote, named company, or case study anywhere in the page copy
Logo row heading: "INTEGRATES WITH YOUR EXISTING TOOLS"
Self-asserted stats with no source: "70%", "99%", "~2 wks", "90% automation", "It reaches 95%+ accuracy"
Product shown only as simulated message bubbles over photography, never as a real screen or a real usage example
Why it matters
For a product asking an operations leader to hand over a live workflow, who else runs this and what does it actually look like are the first two questions, ahead of price or features. A page that makes strong numerical claims, shows no real product, and offers no external validation invites the reader to discount all of it at once.
Fix
Add proof next to the claim. A row of real customer logos under the hero, one named testimonial with a title and company, a real product screen or short recording of the agent working, and at least one case study with a before and after number. If a logo cannot be named yet, an anonymised but specific line such as a national HVAC operator, 40,000 jobs a month is stronger than a rounded percentage with no owner.
The page is heavy for an audience that is often on mobile in the field
whole page, measured on load · conversion friction
highobservedquick win
The page pulls 13.7 MB over 55 requests against a 2.5 MB budget. Seven images are shipped far larger than they display, the worst a 1440px map served into a 424px slot, and logo SVGs served at up to 690px for a 72px render. First paint is fast at 456ms, so the hero feels quick, but the full payload is heavy for a product whose field services and trucking users are frequently on a phone on a job site with a weak connection.
690px served, 72px displayed: Service-Channel logo
7 images shipped far larger than their displayed size
First contentful paint 456ms, so the weight is below the fold, not blocking paint
Why it matters
Weight is felt most by exactly the operator this product is for, on cellular in the field. A page that is slow or expensive to load on a phone undercuts a pitch built on reliability and speed to live.
Fix
Serve images at the size they render with responsive srcset, and cap the logo SVGs to their display box. That alone should remove most of the payload. Aim the full page under 3 MB and re-measure on a throttled mobile profile.
Heavy pages lose visitors before they render, and the loss is worst on the mobile connections most visitors use.
Fix
Compress and correctly size images, serve modern formats, and defer non-critical scripts.
Visual and behavioral
4 issue(s) · scores 90/100
The only logo wall is integrations, positioned where customer proof belongs
below the hero, desktop and mobile · social proof density
highobservedmedium
Directly under the hero, in the slot the eye expects a customer logo row, sits a strip of eight brand logos under the heading Integrates With Your Existing Tools. Visually it is indistinguishable from an as used by wall, and a fast scroller will read Oracle, SAP, Salesforce and Microsoft as customers. There is no actual customer logo row anywhere on the page to correct the impression.
Evidence
Integration strip in the prime post-hero position: Oracle, SAP, Salesforce, ServiceMax, MAS, Corrigo, ServiceChannel, Microsoft
Heading present but small: "INTEGRATES WITH YOUR EXISTING TOOLS"
No separate customer or as-trusted-by logo row exists on the page
Why it matters
Borrowing the visual grammar of social proof without the substance is a short-term win that a careful buyer sees through, and it wastes the single best position on the page. Used honestly, that slot is where a real customer row would do the most work.
Fix
Keep the integration strip, it is a genuine strength, but move a real customer logo row above it and let the integrations sit lower as a capability signal. If there are no logos to show yet, replace the strip's prime position with a named testimonial rather than leaving borrowed proof there.
Lifestyle photography crowds out the product, so you never see what you would get
hero and most sections, desktop and mobile · product demonstration
highobservedmedium
The page is carried by large, well-shot lifestyle photographs of drivers, warehouse and field workers, and the product appears only as stylised chat bubbles composited on top of them. Nowhere does the real interface appear: no dashboard, no inbox, no view of where the agent's messages actually arrive or how an operator reviews and corrects them. The craft is high, but the imagery is doing mood work where it should be doing proof work, and it spreads across so many different job sites that it widens the who is this for question rather than answering it.
Evidence
Hero and section backgrounds are lifestyle photography, not product
The only product depiction is simulated message threads overlaid on photos
No screenshot, recording, or real view of the interface, the agent inbox, or the audit log the copy describes
Photography spans trucking, construction and field, and warehouse, with no single buyer anchored
Why it matters
For software, seeing the product is a large part of believing it works, and this buyer is being asked to trust an agent with a live workflow. Mood photography without a real screen reads as a company that is not ready to show the thing, which is the opposite of the reliability the copy promises.
Fix
Trade some of the lifestyle photography for the product itself. Show the real interface where the agent works, one genuine message thread from a real job, and the audit trail the copy references. Keep the photography as accent, and let a real screen carry the hero.
Eleven tap targets fall under the minimum on mobile
mobile, 390px viewport · responsive integrity
mediumobservedquick win
Eleven tappable elements measure under the 44px minimum on a 390px viewport, including the primary navigation entries (Home, Platform, Field Services, Trucking, Careers, Contact) at roughly 19px tall and two controls around 32px. The audience for this product is disproportionately on a phone, which makes small targets more costly here than on a typical desktop-first B2B site.
Evidence
11 tappable elements under 44px on a 390px viewport
Nav items at ~19px tall: Home, Platform, Field Services, Trucking, Careers, Contact
Two controls at 140x32px and 224x32px
Why it matters
Mis-taps on a phone read to the visitor as the site being slightly broken, and that impression transfers to a product whose whole pitch is that it runs operations reliably.
Fix
Give every tappable element a 44 by 44px minimum hit area using padding rather than font size, and verify the mobile nav in particular on a real device.
Touch targets under 44px on mobile
code · responsive integrity
mediumobservedquick win
11 tappable element(s) under the 44px minimum on a 390px viewport.
Evidence
"" — 140x32px
"" — 224x32px
"Home" — 45x19px
"Platform" — 64x19px
"Field Services" — 101x19px
"Trucking" — 64x19px
"Careers" — 57x19px
"Contact" — 60x19px
Why it matters
Targets under 44px cause mis-taps on mobile, which reads to the user as the site being broken.
Fix
Give every tappable element a minimum 44x44px hit area, using padding rather than font size.
Compliance and trust
7 issue(s) · scores 95/100
The primary Book a Demo button is a div with no keyboard access
navigation, desktop · accessibility risk
mediumobservedquick win
The Book a Demo control in the navigation is built from a div or span with no button role and no keyboard focus, so it cannot be reached by Tab or activated by keyboard. This is the single most important conversion element on the page. There is also no accessibility statement linked anywhere, which enterprise procurement increasingly asks for.
Evidence
1 clickable element built from a div or span with no button role and no keyboard focus: "Book a Demo" class="navbar-button hide-tablet"
No accessibility statement linked on the page or at the usual paths
Why it matters
A keyboard or screen reader user cannot trigger the main call to action from the nav, which is both a lost conversion and a concrete accessibility failure. For a company selling into enterprise and public sector operations, an accessibility gap on the primary button is the kind of thing a procurement review flags.
Fix
Make it a real button or anchor element with a visible focus state, so it is keyboard reachable and announced correctly. Add a short accessibility statement and link it in the footer.
No cookie policy, though privacy and terms are in place
footer and standard cookie paths · legal pages
mediumobservedquick win
The privacy policy and terms of service both exist, are linked in the footer, and resolve to real documents, which is more than many sites manage. What is missing is a cookie policy: it is not linked from the page and is not present at /cookies or /cookie-policy. Note that no third-party trackers were detected on the page, so a consent banner is not currently required, which keeps this at a documentation gap rather than a live consent problem.
Evidence
Privacy policy resolves and looks real: /privacy-policy
Terms of service resolves and looks real: /terms-of-service
No cookie policy linked, and none at /cookies or /cookie-policy
0 third-party trackers detected, so no consent mechanism is required today
Why it matters
A cookie policy is a standard item on the enterprise security questionnaire this buyer will run, and its absence is an easy point to lose in a review even when no trackers are live. Adding one closes the gap before it is asked about.
Fix
Publish a short cookie policy and link it in the footer alongside privacy and terms. If tracking analytics are added later, add a consent mechanism at the same time.
Clickable divs and spans instead of buttons or links
code · accessibility risk
mediumobservedquick win
1 clickable element(s) built from divs or spans with no button role and no keyboard focus.
Evidence
"Book a Demo" class="navbar-button hide-tablet"
Why it matters
Controls that are not real buttons cannot be reached by keyboard and are not announced as actionable. Some users simply cannot complete the action.
Fix
Replace clickable divs with <button> or <a>, so they are focusable and announced correctly.
No cookie policy
code · legal pages
mediumobservedquick win
No cookie policy found. Not linked from the page and not at /cookies or /cookie-policy.
Evidence
https://myrunbook.com/cookies
https://myrunbook.com/cookie-policy
Why it matters
If the site sets non-essential cookies it has to say which, why, and for how long. A privacy policy alone does not cover it.
Fix
Publish a cookie policy listing each cookie, its purpose, and its lifetime, and link it from the consent banner.
No Strict-Transport-Security header
server response headers · transparency
lowobservedquick win
The site does not send a Strict-Transport-Security header. A first visit typed as http can be intercepted in the window before the redirect to https completes.
Evidence
No Strict-Transport-Security header present on the main response
Why it matters
HSTS is a one-line hardening step that enterprise security reviews check for. Its absence is minor on its own but is exactly the sort of item a buyer selling on data trust should not concede for free.
Fix
Add a Strict-Transport-Security header with a sensible max-age and includeSubDomains, and consider preload once verified.
No Strict-Transport-Security header
code · transparency
lowobservedquick win
No Strict-Transport-Security header. A first visit over HTTP can be intercepted before the redirect to HTTPS happens.
Why it matters
Without HSTS the first request of a session can be intercepted before the redirect to HTTPS completes.
Fix
Send Strict-Transport-Security with a max-age of at least one year, then consider preloading.
No accessibility statement
code · accessibility risk
lowobservedquick win
No accessibility statement linked. Required for public sector bodies in the UK and EU, and expected by enterprise procurement everywhere else.
Why it matters
An accessibility statement is required for public sector bodies in the UK and EU, and is a standard question in enterprise procurement.
Fix
Publish an accessibility statement covering your conformance target, known gaps, and a contact route for problems.
About these fixes
Every fix above is deliberately high level
Each one names what to change and why, which is enough to act on. What it leaves out is the reasoning underneath: which to do first given how people actually move through your funnel, what the replacement should say in your voice, and whether the change still holds once someone is inside the product rather than looking at the page.
Every recommendation we make is backed by a framework and a method, and tested against two things at once: whether it helps the person using the product, and whether it helps the business. A change that wins one and costs the other is not a fix, it is a trade nobody agreed to.
That is why we describe this as product design rather than visual work. If the thinking in this report is useful, the same thinking applied with your data in the room is where it compounds.
Static capture only. Anything behind a click, a login, or a payment step was not exercised and is flagged
as a risk rather than a confirmed fault. Findings marked "likely" or "possible" need a live walkthrough
to confirm. Scores are the framework's weighting of severity and confidence, not an industry benchmark.
Generated by UX Signal Studio using the UX Risk framework.
Section 07
Art direction
Design verdict
High craft, aimed at mood instead of the product
Opinion, not measurement
A deep navy ground with a mint green accent used sparingly on the words that carry the claim (doing the work for them, We ship in days, automate). Display headlines set in a serif, body in a clean sans. Large, well-shot lifestyle photography of drivers, warehouse and field staff. The product appears only as simulated agent message threads composited over that photography. There is no screenshot of the real interface anywhere.
This is a design judgement rather than a measurement. The craft is real and the register is right for the buyer, but the imagery is doing the wrong job. It sets a mood where it should be showing the product, and it spreads across so many different job sites that it widens the who is this for question instead of answering it. For a software product, a visitor wants to see the thing: the interface, where the agent's messages actually land, how an operator reviews and corrects them. Handsome photography of a trucker is not that, and after three or four seconds of not seeing the product or recognising themselves, most visitors leave. The skill on display is exactly what is needed, it is simply pointed at atmosphere rather than proof.
What we would do instead
Turn the same craft on the product. Lead with a real screen of the interface, show one genuine message thread from a real job rather than a staged one, and surface the audit trail the copy describes. Every image should advance what the product is and who it is for, so a field services or trucking lead sees themselves and sees the tool at once. Keep the palette and the type, they are working.
Everything else in this report is measured or quoted. This section is not. It is a design judgement,
offered because the gap between the imagery and the product is the kind of thing a scan will never
flag and a buyer will always feel.
Section 08
This is the surface
Your whole site next, not just this page
A scan reads one page the way a stranger does. It can tell you the category is missing from the
hero and that the closing call to action contradicts its own copy.
What it cannot do is decide what the product should promise, or design the flow that keeps that
promise once someone is inside it. Everything in this report is above the signup. The compounding
return is below it.
What the engagement adds
Every screen, not one. The full site, the onboarding flow, and the paths this scan cannot reach because they sit behind a signup.
Designed, not just diagnosed. We do the work, so the fix ships rather than joining a backlog as a note.
Your analysts in the room. We read the funnel data with the people who own it, so we are arguing from numbers rather than instinct.
Experiments, not opinions. Each change ships as something measurable, so you know within a fortnight whether it worked.
What a scan reaches
One public page, as a first-time visitor sees it
Copy, hierarchy, and the first five seconds
Accessibility, performance, and consent in the markup
