DEV Community: Cole · Tracking AI Search

A Sanity Check for HVAC Flat-Rate Pricing: Labor, Overhead, and Margin

Cole · Tracking AI Search — Tue, 01 Sep 2026 02:37:53 +0000

Flat-rate pricing looks simple from the customer side: one task, one quoted price. Behind that number, however, is a compact cost model. If any part of that model is missing, a price book can look polished while quietly losing money.

I have found that the most useful way to review a flat-rate service price is not to ask whether it looks competitive. The better question is whether every assumption can be traced back to the shop that will perform the work.

This is a practical sanity check for small residential HVAC service teams. It is not a national price guide, and it does not tell a contractor what the market will bear. The shop supplies its own numbers.

1. Start with a real billable labor rate

The hourly wage paid to a technician is not the same as the labor cost that must be recovered through customer work. Payroll taxes, benefits, training, non-billable time, vehicles, supervision, software, insurance, and other operating expenses still have to be funded.

A useful labor-rate calculation starts with annual labor and overhead costs, then divides them by realistic billable hours. The word realistic matters. A technician may be paid for roughly 2,000 hours in a year, but holidays, training, meetings, travel, callbacks, and gaps in the schedule reduce the number of hours that can actually be sold.

If the denominator assumes perfect utilization, every task price inherits that optimism.

2. Estimate task time honestly

For each service task, record the time normally required from arrival through completion. Include the work that belongs to the job but is easy to forget: diagnosis, setup, cleanup, documentation, and a reasonable share of travel or dispatch time when that is part of the shop's model.

Do not use the fastest technician's best day as the standard. A price book has to work across ordinary jobs and ordinary operating conditions.

For a first draft, the labor component can be expressed as:

labor component = billable labor rate × task hours

The formula is simple; the quality comes from the assumptions.

3. Use replacement cost for parts

Parts should be entered at the cost the shop expects to pay to replace them, not an old invoice price that may no longer be available. Freight, consumables, and small materials also need a consistent policy.

Some shops apply a parts multiplier. Others calculate a separate markup or use price bands. Any of those approaches can be workable if the rule is documented and the final result is checked against the intended gross margin.

4. Separate markup from margin

Markup and gross margin are related, but they are not interchangeable.

If a task costs $100 and the shop adds a 40% markup, the selling price is $140. The gross margin is about 28.6%, not 40%.

To solve for a target gross margin, use:

selling price = cost / (1 - target margin)

So a $100 cost at a 40% target gross margin produces a selling price of about $166.67.

This is one of the easiest places for a spreadsheet to create a confident-looking but incorrect result. Label the input clearly: is it a markup percentage or a target margin percentage?

5. Add an explicit service-call policy

Decide how the diagnostic or service-call fee interacts with approved work. Is it always charged, credited when a repair is accepted, or built into selected tasks? There is no universal answer, but the price book and the customer-facing explanation must agree.

The same applies to after-hours work, difficult access, permits, special-order parts, taxes, and other exceptions. A flat-rate book should reduce improvisation, not hide situations where a separate policy is needed.

6. Stress-test the result

Before releasing a task price, run a few checks:

What happens if the task takes 25% longer than estimated?
Does the price still recover overhead when the schedule is not full?
Is the target margin calculated from revenue rather than mistaken for markup?
Can a technician explain what the quoted price includes?
Is the number based on the shop's costs rather than a competitor's price book?

If the result fails one of these checks, it is cheaper to fix the assumption now than to discover it through a month of underpriced work.

7. Keep the first price book small

A small team does not need hundreds of perfect tasks on day one. Start with the jobs performed most often, document the assumptions, and review actual labor and parts after the work has been completed. Expand the book as the process becomes reliable.

FlatRateKit provides a free HVAC flat-rate pricing calculator and a downloadable HVAC price-book template for organizing those assumptions. The tools are deliberately not a market-price database: the contractor supplies the labor, parts, overhead, and margin inputs.

The point of a good flat-rate price is not that it matches somebody else's number. It is that the shop can explain where it came from, update it when costs change, and deliver the work consistently.

How We Built an HVAC Quote Checker That Treats Scope as Data

Cole · Tracking AI Search — Sat, 29 Aug 2026 12:27:46 +0000

Most HVAC quote tools begin with equipment price. We found that the harder homeowner problem starts one step earlier: two contractor bids can show similar equipment and still describe very different jobs.

A low bid may exclude permits, electrical corrections, duct repairs, old-equipment removal, startup testing, or labor warranty. A higher bid may include those items, but bury them in paragraphs of proposal language. A useful quote checker therefore has to model scope, not just price.

Start with a normalized project model

The calculator and quote checker share a small set of planning inputs: location, home size, system type, efficiency tier, project type, and known installation conditions. We keep those inputs separate from the contractor's written scope so the interface can explain two different things:

A planning range based on the home and system.
Missing or ambiguous line items in the contractor bid.

A simplified TypeScript shape looks like this:

interface QuoteReview {
  project: {
    systemType: string;
    homeSize: number;
    replacementType: "full" | "partial";
  };
  scope: {
    permit: "included" | "excluded" | "unclear";
    ductwork: "included" | "excluded" | "unclear";
    electrical: "included" | "excluded" | "unclear";
    startupTesting: "included" | "excluded" | "unclear";
    laborWarranty: "included" | "excluded" | "unclear";
  };
}

The important design choice is the third state: unclear. In a real proposal, absence is not always the same as exclusion. Treating every missing phrase as a definite problem would create false confidence. The checker instead turns ambiguity into a question the homeowner can take back to the contractor.

Separate the benchmark from the verdict

A planning range is useful, but it should not declare a bid fair or unfair by itself. Local labor, access, code requirements, climate, equipment availability, and home condition can move the installed price substantially.

Our workflow uses the range as context, then reviews the written scope. The output highlights what may explain the difference: difficult attic access, return-air work, panel capacity, refrigerant-line changes, permit responsibility, removal, controls, or warranty coverage.

That produces a more defensible message: "Here are the items that could explain this quote" instead of "This contractor is overcharging you."

Make missing scope actionable

Each warning needs a next step. A generic red flag is less useful than a question such as:

Who pulls the permit and pays for failed-inspection corrections?
Is duct repair included, excluded, or priced after installation starts?
Are the indoor and outdoor units an approved equipment match?
Does startup include measured airflow and refrigerant checks?
How long is labor covered separately from the manufacturer warranty?

The goal is not to replace a licensed contractor or professional load calculation. It is to help the homeowner request an itemized proposal and compare at least three qualified local bids on equivalent scope.

Build transparency into the product

The public tool is built with Next.js, React, TypeScript, and Tailwind CSS. We keep the methodology separate from the interface so assumptions and limitations can be reviewed without reading component code.

The public methodology repository documents the estimate inputs, the quote-comparison workflow, and the safety boundaries:

https://github.com/fishcoco-code/clear-hvac-cost-methodology

You can try the homeowner-facing quote review here:

https://www.clearhvaccosts.com/hvac-quote-checker

The broader installed-cost calculator is here:

https://www.clearhvaccosts.com/hvac-cost-calculator

What we learned

The main lesson was that price transparency is mostly a data-modeling problem. If the product stores only a total price, it cannot explain why bids differ. Once scope becomes structured data, the interface can compare inclusions, surface uncertainty, and generate questions without pretending to know facts that are not in the proposal.

That pattern extends beyond HVAC. Any high-cost local service quote becomes easier to evaluate when the product separates the benchmark, the written scope, and the unknowns.

How to Measure Organic AI Visibility Without Turning It Into Rank Tracking

Cole · Tracking AI Search — Sat, 15 Aug 2026 06:04:42 +0000

AI visibility is often discussed as if it were a traditional rank-tracking problem: choose a keyword, record a position, repeat every day.

That mental model is convenient, but it misses the most useful question for many founders and marketing teams:

When a buyer asks an AI assistant for solutions in my category without naming my brand, does the assistant recommend me, recommend competitors, or skip me entirely?

That is organic AI visibility. It is closer to a structured recommendation audit than a search-engine position check.

This article lays out a practical way to measure it without pretending that probabilistic AI answers behave like a stable list of blue links.

1. Start with the decision you want to understand

A useful prompt represents a real buying or evaluation decision. It should contain enough context for an assistant to make a recommendation, but it should not lead the model toward your brand.

Weak prompt:

Is Acme Analytics a good product?

This measures what the model says after you name the brand. It does not measure organic discovery.

Stronger prompt:

What are the best analytics tools for a small B2B SaaS team that needs simple activation and retention reporting?

This gives the assistant room to assemble its own shortlist. The resulting answer reveals which brands it associates with the need, how it describes them, and which criteria appear to drive the recommendation.

2. Build a small prompt set around buyer intent

You do not need hundreds of prompts for an initial snapshot. Start with 10 to 20 prompts covering a few distinct intents:

Category discovery: "What are the best tools for X?"
Use-case fit: "What should a team use to solve Y?"
Constraint-based evaluation: "Which options work for a small team, limited budget, regulated market, or technical buyer?"
Comparison intent: "What are the strongest alternatives for this type of workflow?"
Implementation intent: "How should a company set up or improve this process?"

Keep the market, language, and audience consistent. Record the exact prompts. If you later repeat the exercise, methodological consistency matters more than adding a large number of loosely related questions.

3. Classify the answer before interpreting it

For each answer, capture a small set of observable facts:

Was your brand recommended?
Which competitors were recommended?
Where did each brand appear in the answer?
What reasons, use cases, or differentiators were attached to each brand?
Were sources or citations provided?
Did the answer express uncertainty or qualify the recommendation?

A simple classification works well:

Recommended: your brand appears as a genuine solution to the unnamed need.
Mentioned: your brand appears, but not as a clear recommendation.
Competitors only: the model recommends alternatives and excludes your brand.
No usable recommendation: the answer is generic or avoids naming products.

This avoids a common measurement error: treating every textual mention as equivalent visibility.

4. Separate presence from positioning

A mention alone is not necessarily a win.

Suppose an assistant names your product but describes it as suitable only for enterprises, while your actual target customer is a five-person startup. That is visibility, but the positioning is inaccurate.

For every meaningful mention, record:

the category the model places the product in;
the audience it thinks the product serves;
the strengths and limitations it repeats;
the competitors it groups with the product;
the evidence or sources behind those claims.

This turns a vague "AI mentioned us" observation into a concrete content and positioning backlog.

5. Treat the result as a snapshot, not a permanent rank

Generative answers can vary because of model updates, sampling, prompt phrasing, geography, and changing source material. A single run should therefore be presented as a directional sample.

Avoid statements such as "we rank number three in ChatGPT." There may be no stable third position to own.

A more defensible statement is:

In this prompt set and sample, the brand was organically recommended in 4 of 12 buyer-intent questions, while two competitors appeared in 8 and 7 answers.

That description preserves the evidence and the limits of the method.

6. Convert gaps into specific improvements

The most valuable output is not a score. It is the explanation of what to improve next.

Common gaps include:

the category is unclear on the product site;
comparison pages do not answer real evaluation questions;
use-case pages lack concrete examples;
third-party reviews describe an outdated product;
the product has weak evidence for important claims;
documentation and pricing information are difficult to find;
competitors have stronger coverage on trusted, crawlable sources.

Map each gap to an asset or action. For example, if assistants repeatedly associate a competitor with agencies but never connect your product to that audience, publish a genuinely useful agency workflow page with examples, limits, and proof.

7. Repeat only when the comparison is meaningful

A later snapshot can be useful after you ship material changes: clearer positioning, new comparison content, better documentation, updated third-party coverage, or customer evidence.

Use the same prompt set and sampling method when comparing snapshots. Otherwise, changes in the test itself can look like changes in visibility.

This is also why AI visibility should not automatically be packaged as continuous daily rank tracking. Many teams first need a reliable diagnosis and a prioritized improvement plan, not another dashboard to watch.

A lightweight workflow

A practical first audit can fit in a spreadsheet:

prompt | intent | brand outcome | competitors | positioning | citations | next action

Run the prompts, save the exact answers, classify them consistently, and summarize:

recommendation coverage;
competitor frequency;
recurring reasons for inclusion or exclusion;
positioning errors;
the highest-impact next actions.

If you prefer a structured report instead of manual collection, TrackAIMentions provides a free AI visibility check focused on ChatGPT-style answers plus a Perplexity visibility sample. It is report-first: the goal is to show whether AI organically recommends your brand or competitors and what to improve next, rather than imitate a traditional SEO rank tracker.

The key principle is simple: measure the recommendation decision, preserve the evidence, and be honest about uncertainty. That produces a smaller number of metrics, but far more useful work.

How to Check If a Roof Replacement Quote Is Too High

Cole · Tracking AI Search — Thu, 04 Jun 2026 14:24:23 +0000

A roof replacement quote can be hard to judge because most homeowners only see this purchase once or twice in their lives. A $9,000 quote and a $19,000 quote can both look plausible until you compare the scope line by line.

Here is the checklist I use when reviewing a roofing estimate.

1. Start with roof size, not house size

Many quotes are built around the roof surface area, not the interior square footage of the home. Pitch, overhangs, garages, dormers, hips, valleys, and waste factor can all change the final number.

If the quote does not clearly state the measured roof area or number of squares, ask for it. Without that, it is almost impossible to compare two bids fairly.

2. Confirm the material system

A quote should not just say "shingles" or "metal roof." It should identify the material system and any major components included. For example:

3-tab asphalt shingles
Architectural shingles
Standing seam or exposed-fastener metal
Tile
Flat or low-slope membrane

The material choice changes cost a lot. It also changes warranty expectations, underlayment needs, flashing details, and installation complexity. A cheap quote may simply be pricing a different system.

3. Check tear-off and disposal

One of the biggest quote gaps is whether the contractor is removing the old roof. Tear-off, hauling, dump fees, and cleanup should be explicit.

If one quote includes full tear-off and another assumes an overlay, they are not equivalent. Overlay can sometimes lower upfront cost, but it may hide decking issues and can affect future replacement options.

4. Look for decking repair assumptions

Decking is often the unknown item. A good quote will explain whether it includes a small allowance, a per-sheet replacement price, or no decking repair at all.

This matters because the final invoice can move quickly if rotten plywood or plank decking is found after tear-off. The red flag is not necessarily a decking charge. The red flag is a quote that is vague about how decking will be handled.

5. Read the flashing, ventilation, and underlayment lines

These details are easy to miss, but they are part of what makes a roof perform. A complete estimate should mention items like:

Drip edge
Valley treatment
Step flashing and wall flashing
Pipe boots
Ice and water shield where relevant
Synthetic or felt underlayment
Ridge vents, box vents, or other ventilation approach

A quote that only lists shingles and labor may be too thin to evaluate.

6. Ask who handles permits

Permit rules vary by city and county. Some quotes include permit handling, some pass permit fees through to the homeowner, and some leave the topic out completely.

That omission can create confusion later. If permits are required in your area, the estimate should say who pulls them and whether fees are included.

7. Compare the price against a local range

National averages are useful, but local labor, disposal, access, code requirements, and weather exposure matter. A coastal Southern California roof will not always price like a similar roof in inland Florida.

I built Roof Cost Data as a free homeowner research tool for this exact comparison problem. You can estimate a rough planning range with the roof replacement cost calculator, then sanity-check a written bid with the roof quote checker.

For local examples, the site also has city pages such as Long Beach roofing cost estimates, where coastal exposure, permits, material choices, and roof size assumptions can change the range.

8. Do not treat the lowest quote as automatically best

A low quote can be legitimate. It can also be missing tear-off, permit handling, ventilation work, flashing replacement, decking allowances, cleanup, or warranty detail.

A high quote can be padded. It can also include better materials, more difficult access, full tear-off, damaged decking allowances, or stronger warranty support.

The better question is: what exactly is included, and can another contractor price the same scope?

Quick red flags

Be careful if a roof quote:

Does not state roof size or squares
Uses vague material names
Leaves tear-off unclear
Has no decking repair terms
Ignores flashing or ventilation
Says nothing about permits
Pushes for a same-day decision
Has a very large deposit requirement
Refuses to put scope details in writing

Roof Cost Data is not a roofing contractor and does not perform roofing work. The goal is to help homeowners understand the quote before they decide who to hire.