What we've shipped.

Shipped

The log

• Change

  New pack prices: more credits per dollar on the bigger packs.

  Credit packs are now Starter $49 for 49 credits, Crew $99 for 125, Hunter $199 for 300 and Pro $399 for 800. That is $1.00 a credit on Starter, $0.79 on Crew, $0.66 on Hunter and $0.50 on Pro. They are still one-off payments with nothing that renews, and credits you buy still never expire. What a credit buys has not changed, the free trial is still 10 credits with no card, and drawing an area and ranking every roof in it is still free.

• Change

  No more subscriptions. Buy credits when you need them.

  Hunter ($199 a month) and Hunter Pro ($499 a month) are gone, and so is anything that renews. After the free trial you buy a one-off credit pack when you need one: Starter $25 for 25 credits, Crew $79 for 100, Hunter $199 for 350, Pro $499 for 1,200. Credits you buy never expire. Every pack gets the whole product, the trial included can buy one, and a bigger pack only costs less per credit. What a credit buys has not changed, and drawing an area and ranking every roof in it is still free.

• Change

  Scanning new ground is free. Any size, any number of roofs.

  Reading ground nobody has read yet used to cost 1 credit per 150 roofs. It now costs nothing at all, however big the polygon. The reason was not generosity: the charge only ever applied on one of the three paths into a scan, so most ranking was already free and the price was effectively a lottery over which button you happened to press — and it was stopping roofers from reading their own neighbourhoods. Credits now buy exactly three things: reading roofs properly, the owner’s contact details, and opening a house. Draw as much ground as you like and rank every roof in it for nothing.

• Feature

  Tell it once. It remembers how you work.

  Say “I only do asphalt”, “there’s two of us so eight doors is a full day”, “nothing under 2,000 square feet is worth my morning” — and it holds on to it, so every search after that is already shaped the way you work. It keeps what stays true and forgets what was only this morning’s request: “show me Katy” is a question, “I work Katy on Tuesdays” is a fact about you. Everything it believes about you sits on one page in your own words, each line marked with where it came from — what you said outright versus what it worked out from how you work. Agree with a line, edit it, switch it off without deleting it, or forget it entirely. Nothing it worked out on its own is treated as settled until you have seen it.

• Change

  One read buys the picture and the record. The owner is its own step.

  Five things you could pay for became three. Reading roofs now returns BOTH photographs and the full property record — year built, size, beds, baths, lot, value, last sale, whether it is off market — in one purchase at 1 credit per 8 roofs. There is no separate records buy and no separate street-photo buy, because paying twice for the same house never made sense. The owner’s name, phone and email is its own step at 1 credit per 2, and that is deliberate: the phone lookup is roughly three quarters of what knowing everything about a house costs us, so you should not pay for it on houses you have not decided about yet. Read, pick the doors worth knocking, then buy the owners for those. Bulk reading used to be 1 credit per 4 for the record ALONE — it is now twice the roofs with the photographs included. Opening one house completely is unchanged at 1 credit, and searching, sorting, filtering and saving lists are still free and unlimited.

• Improvement

  The street photo now overrules the aerial, and the roof gets measured when you look at it

  A contractor drove nine doors off his own list and eight had already been re-roofed. The roofs had not changed since we read them — the picture had. His area was ranked from a July 2016 aerial, and the street-level photo was newer on 397 of those 500 houses by an average of 7.9 years. Two things changed. Every shortlist is now checked against the newer street photo before you see it, and anything already done is moved down and said out loud. And the ranker’s own sentence now counts: a roof described as recently replaced is treated as finished everywhere — on the chip, in the filters, on the map — instead of in one filter that happened to check. A roof’s row also shows both dates when they disagree, so you can see which picture the verdict came from.

• Change

  New credit rates

  Reading new ground is 1 credit per 150 roofs, looking at the photographs from above and from the road is 1 per 20, and property data in bulk is 1 per 4. Opening a house completely is unchanged at 1 credit, and it still buys the full roof read, the address, the owner’s name, phone and email, and the whole property record. The trial is unchanged at 10 credits, and everything that was free stays free and unlimited on every plan: asking questions, searching roofs already read, sorting, filtering, saving and naming lists, and however many service areas you want at whatever size. The exact roof measurement — size in squares, pitch, planes and the capture date — is now bought for the roofs a search actually puts in front of you and for any house you open, at our cost rather than yours, and once bought it is never charged again to anyone.

• Feature

  Ask anything about the roofs in your area

  There is now a box you type a sentence into. “Find me the worst roofs in 75216.” “Houses where most of the neighbours have already re-roofed.” “Six doors for Tuesday morning, in walking order.” It searches the roofs already read in your areas and answers instantly — free, unlimited, spending nothing and opening nothing. When you ask it something the data cannot settle, it opens the actual photographs and looks: tarps, patches, moss, rusted metal, trees hanging over the roof, solar panels, a second storey, a detached garage. That looking is free too, and takes about half a minute. Ask it to do the work — rank a new block, pick the worst, put them in walking order, open them and pull the owners — and it tells you what that will cost before it starts, and waits for you to say yes. It has never spent a credit nobody agreed to.

• Feature

  Every search saves itself as a list

  The first search in a conversation is already on a list by the time you have read it, so nothing you found is lost by asking a second question. Tell it “call that one Tuesday” and the list is renamed; tell it “add those to the Chariton walk” and they go somewhere else. Lists are free, and there is no limit on them.

• Change

  The caps are gone. Credits are the only meter.

  Service areas were 1 on the trial and 10 or 25 on a plan; coverage had a house ceiling; there was a daily read cap. All of it is removed. Draw as many areas as you want, as big as you want, in as many towns as you want. Searching, sorting, filtering and saving what has already been read are free and unlimited on every plan including the trial. The only thing that costs anything is acting: reading ground nobody has read yet is 1 credit per 300 roofs, looking at the photographs from above and from the road is 1 per 40, property data in bulk is 1 per 8, and opening a house completely is 1 credit. Hunter is 350 credits a month and Hunter Pro is 1,000. Nothing is gated by plan — the only difference between them is how many doors a month you can open.

• Feature

  The date of the photograph is on the roof

  A roof read off a 2016 aerial and a roof read off a 2024 aerial are not the same evidence, and until now nothing on the screen said which you had. Every ranked roof now carries the capture date of the imagery it was judged from, along with its size in squares, how many planes it has and how steep it is. It matters more than it sounds: coverage is a patchwork rather than one flight, so roughly one roof in nine is older than its neighbourhood’s date suggests, and the Street View shot is newer than the satellite one on the large majority of houses. Where we have a roof’s own date, it wins.

• Feature

  Property records are part of the free read

  Year built, last sale date and price, whether the owner actually lives there, assessed and estimated value, beds, baths, living area, lot size, taxes and the mortgage. These come free inside a house you open, and can be pulled in bulk for a list before you open anything at 1 credit per 8 roofs. Imagery can tell you what a roof looks like; it cannot tell you when the house was built or who bought it last year, and a tier is not an age proxy — we measured houses built after 2016 ranking ‘worn or worse’ just as often as pre-1980 stock.

• Change

  The free trial is 10 credits

  Down from 25. What the credit buys is unchanged — one credit is one house: the satellite and street-level read together, the property record, the storm-claim clock, a door pitch, and the owner’s name, phone and email. Paid plans are untouched at 500 and 2,000 a month. The free part of the trial is also untouched, and it is the part worth having: every house in your service area ranked against its neighbours and put on the map, up to 1,000 of them, at no cost and with no card. Ten is what it takes to judge whether the order is any good — you knock the top of a list, not twenty-five houses — and spending less on people who never come back is what keeps the ranking free for the ones who do.

• Improvement

  One credit per house, and a number on the free coverage

  Two allowances became one, and the one that was invisible got a number. Unlocking a house used to spend a roof read AND a separate contact unlock, which meant you could be told “15 reads left” and “0 contacts left” at the same time about the same house — two numbers that both mean you cannot work this door. Now one credit buys the whole house: satellite and street-level imagery read together, the property record, the storm-claim clock, a door pitch, and the owner’s name, phone and email. Trial 25 credits, Hunter 500 a month, Hunter Pro 2,000. Prices unchanged. Separately, ranking houses onto the coverage map still costs no credits, but it now carries its own figure rather than being described as unlimited: 1,000 houses on the trial, 10,000 on Hunter, 25,000 on Hunter Pro, counted across your active areas. It is a standing footprint rather than a monthly budget — a rank is cached for a year, so a house ranked once stays ranked and is never re-charged, and archiving an area you have finished with releases the allowance for a new one. Service areas are 1 / 10 / 25 rather than unlimited, which changes nothing in practice: because coverage is counted across areas and not per area, drawing more of them was never going to get you more of the map.

• Feature

  See every roof we read, not just the ones we picked

  A scan has always compared every roof in your service area against its neighbours — that is how it decides which ones to look at closely. Until now you never saw that work. One first scan in Cumming, Georgia ranked 778 roofs, found 99 that looked worse than the houses either side of them, and handed over nine leads. The other 766 existed only in our database. They are now a map. Every roof in your area, tiered, with its satellite tile next to it, on every plan including the free trial — ranking is free and has no cap, however many roofs there are. What your plan meters is the full read: satellite and street-level imagery taken together, the property record, the storm-claim clock and a door pitch. Each scan spends a few automatically. The rest you spend yourself, on whichever roof you want, one read at a time. Nothing about your price or your monthly numbers changed — we just stopped hiding the part you were already paying for. A comparative rank is a starting order rather than an inspection, and the map says so: it tells you a roof looks worse than its neighbours, which is a thing you can check from the street, and the full read is what tells you whether the owner can and will buy.

• Improvement

  Leads you can't actually call now sort lower

  When you unlock a homeowner's contact we learn things that change how good the lead is, and we used to throw all of it away. Half the phone numbers in the country are on the federal Do Not Call registry, some addresses come back with no contact at all, and sometimes the contact vendor names a different person than the deed does — a recent sale returns the seller, an entity-owned parcel returns a tenant who cannot authorise a re-roof. A lead with none of those problems and a lead with all of them used to sit in exactly the same place in your list. Now the ones you cannot legally or usefully work sort below the ones you can, and the reason is written on the card. Applied to every contact you have already unlocked, so your existing list re-sorts too.

• Feature

  Connect your CRM, and score the leads already in it

  Roofbird now connects to JobNimbus, AccuLynx and SalesRabbit, and it works both ways. Leads from a scan go straight into your CRM — a contact in JobNimbus tagged Roofbird, a contact and a job in AccuLynx with priority set from our rank so your worst roofs sort to the top of a list your team already reads, or a lead pinned on the SalesRabbit map with real coordinates. Every one arrives carrying the full assessment in your CRM's own notes field, so it works on day one with nothing configured; mapping a custom field is optional and only buys you a sortable column. The other direction is the more interesting half: point Roofbird at the leads already sitting in your CRM and it reads the roof on each one and tells you which to follow up with. You choose the segment — a city, a pipeline stage, everything created before a date, anything nobody has touched in a year — and it tells you how many records match before you spend anything. It also flags the ones somebody else has already re-roofed, which no CRM field can tell you. Your addresses, your phone numbers, no contact unlocks, and it draws on the same scored-roof allowance your plan already includes.

• Feature

  You choose exactly what transfers, field by field

  Ninety fields are available to send into your CRM, and you decide where each one goes: into the notes, which is free and needs no setup on their side, or into its own custom field, which costs you a field but buys sorting and filtering. A live preview renders the note against one of your own real leads, so a field that is blank on your data shows as blank rather than looking populated in a sample. Two things are deliberately not configurable: the DNC warning travels on every lead that has phone numbers, and a registry number never lands in the field a rep taps to dial. Those are compliance, not preference.

• Feature

  Roofbird tells you where to scan, before you draw anything

  Search the area you work — a town, a suburb, a ZIP or a street — and Roofbird reads the neighbourhoods in view and ranks them. Each block reports its housing era, what share of roofs read as weathered from satellite, the real home count from building footprints, hail in the last 12 months, and how walkable it is — scattered acreage on long driveways is called out as bad canvassing even under a perfect storm, because a morning is measured in doors per hour. Click a block and it draws the area for you, still draggable, or draw your own polygon instead. It only suggests blocks worth the trip: when nothing in view clears the bar it says so rather than offering three neighbourhoods you should skip. Block quality is the same fact for every roofer, so every scan anyone runs sharpens the map for everyone.

• Feature

  Two scores per lead: how bad the roof is, and whether anything is forcing a decision

  Need is condition and roof age. Now is whether something specific to that house is forcing a decision — visible active damage, hail on that roof, a purchase in the last 18 months, patching already attempted, or neighbours who have visibly re-roofed. Alongside it is a one-line 'why now' in plain words, or an honest 'no trigger yet'. Neighbourhood-wide hail deliberately does not raise the Now score: every home in the scan got the same storm, so it can't separate one door from the next. It belongs in your pitch, not the ranking.

• Feature

  Neighbour re-roof detection

  Roofbird now compares each roof against the houses immediately either side, from a wider satellite view centred on the target. When they've visibly gone new and this one hasn't, it's flagged and it raises the Now score — a street where everyone else has already replaced is the strongest close in the trade. It only counts when the contrast is obvious; similar-looking roofs either side is not treated as a trigger.

• Feature

  Filter the list down to who you can actually call today

  Card view or a side-by-side table with sticky headers, frozen actions, zoomable imagery and inline unlock. Filter on owner-occupied vs absentee, contact state (including 'callable' — at least one number confirmed off the federal DNC registry, with unconfirmed 'verify' numbers deliberately excluded), whether there's a live trigger, active damage visible, insurance claim window still open, property record present, and minimum need score, roof age or owner equity. Every term explains itself on hover.

• Feature

  Insurance claim window countdown

  Where a hail event is on record for the property, leads now show roughly how many months the homeowner has left to file, based on the typical notice deadline in that state. It's a prompt for the conversation, not a legal deadline — always verify with the carrier. Roofbird does not predict whether a claim will be approved.

• Improvement

  Owner equity estimate, and the full property record shown free

  Every scanned lead now shows the whole property record without an unlock: owner name, owner-occupied vs absentee, estimated market value, year built, last sale, beds/baths, square footage, lot, tax and lender. Roofbird also estimates how much of the home the owner actually owns, because equity is what decides whether someone CAN say yes. When the deed record can't support an estimate it says so instead of guessing — a missing mortgage record used to be read as no debt at all, which wrongly showed roughly half of all homeowners as owning outright. Only the DNC-scrubbed phone and email now sit behind a click.

• Improvement

  Scans return far more leads, and show what they're doing

  A scan's candidate pool was sized off the per-scan lead cap, but two cheap pre-filters added earlier drop most of a raw pool — so a dense area could return a single lead where it should have returned eight or more. The pool is now sized for that attrition, with a separate budget capping what a scan actually spends. Scans also narrate their real progress ('mapping every building in your area', 'inspecting roofs — 3 leads found so far') instead of showing a bar that barely moves, and repeat scans of the same area are dramatically faster now that the building map is cached.

• Fix

  A visible tarp can no longer be filtered out before assessment

  The cheap first-pass filter that rejects obviously sound roofs had no instruction about damage, so a roof with a tarp, missing shingles or exposed decking could be dropped before the full assessment ever looked at it — losing the single highest-intent lead in the area. Visible damage now overrides every rejection rule and jumps the assessment queue.

• Feature

  Zip-level slot exclusivity (5 subscribers per zip)

  Each zip is capped at 5 active subscribers. When the cap fills, new sign-ups for that zip are rejected at Stripe checkout. Slot reservation is atomic on the webhook so no two roofers can race for the last slot.

• Feature

  Beyond-the-roof property signals

  AI vision now also reads solar panels (lowers score), multi-story (raises price), pool, detached structures (garage/shed/barn → bundle into quote), heavy tree overhang, partial repairs visible, newer adjacent roofs, recently landscaped, chimneys + penetrations. Every signal shows as a chip on the lead card.

• Feature

  Diagnostic roof report PDF (replaces door-hanger)

  Per-lead PDF now carries a full roof grade A–F with verdict, age band + estimated lifespan remaining, plain-English damage explanations for 10 signal types, cost range estimate, satellite shot, and your branding. Still one page, printable, handover-ready.

• Sample

  Denton, TX live sample dashboard

  Pre-scored polygon covering Denton + Argyle + Sanger. ~700 homes, asphalt + tile mix. Anyone can open it at /denton-may24 without signing up.

• Improvement

  AI confidence rating + cost-range estimate per lead

  Every score now ships with the AI's confidence (low/medium/high) plus a $X.Xk–$Y.Yk cost range, derived from roof material × estimated squares × complexity. Use confidence to filter the list when you only want sure-thing knocks.

• Feature

  Storm intelligence with 62-day time-decay

  NOAA SPC storm reports are now folded into the ranking via a recency × distance × severity model (62-day half-life on hail signals). Marine waterspouts auto-filtered. Daily county-level digest ranks outbound priority by population × severity.

• Sample

  Kansas City May-18 sample dashboard

  KC suburbs sample with a focus on aging asphalt in Overland Park, Lenexa, Olathe. ~900 scored homes, available at /kc-may18.

• Feature

  Vision-only scoring + signal tag chips

  Buy-probability score is now driven by direct visual reads of the roof (material, condition, age signs, damage indicators) plus business fit. Score signals show up as chips on every lead card so crews can skim the list in seconds.

• Feature

  Smart keep-filter — low-likelihood roofs don't burn quota

  Roofs the AI rates low (recently replaced, wrong material, no visible wear) are filtered out before they count against your monthly limit. Live scan summary now shows the full funnel: in-polygon → scored → kept vs filtered.

• Feature

  Branded door-hanger PDFs

  One-page printable per lead with your logo, brand color, the roof score, the AI-generated door pitch, and a satellite shot. Auto-pulled branding means zero setup — works the moment you drop your website during onboarding.

• Feature

  Website auto-pull → business profile

  Drop your website in onboarding and the AI reads it to build a profile: materials you install, sales angle, brand voice. Then it scores every roof against THAT profile. Same map, different roofers, different leads.

• Feature

  Unlimited service areas on paid plans

  Hunter and Hunter Pro now support unlimited polygon service areas. Draw single streets, side-of-highway shapes, multiple neighborhoods, whatever matches how you actually canvass.

• Improvement

  Manual address search + batch pin-drop

  Type any address inside your service area to score it on demand. Drop multiple pins on the map and batch-score them in one click. Manual adds still get the full AI workup, tags, and door-hanger PDF.

• Feature

  Hunter + Hunter Pro billing (Stripe)

  Two paid tiers live: Hunter ($199/mo, 500 leads) and Hunter Pro ($499/mo, 2,000 leads + priority support). Monthly billing, cancel-any-time from the account page, no save-offers.

• Feature

  NOAA storm overlays

  Public NOAA SPC storm history is now layered on the map. Useful for roofers who care about hail-touched neighborhoods, even though the score itself doesn't rely on damage-from-satellite claims.

• Feature

  Activity log per property

  Track which leads you've knocked, contacted, quoted, won, or skipped. The activity feed shows your 7-day pace and a per-lead history once you start working a list.

• Sample

  DFW live sample dashboard

  First public sample. Pre-scored DFW metroplex polygon (Plano + Frisco + North Dallas) at /dfw, no signup required.

• Improvement

  Score explainer modal

  Click any score to see the exact signals the AI used to build it. Roof condition, business fit, neighborhood signal each get a sub-score. No more black box.

• Feature

  First scored polygon end-to-end

  Ask in plain words → we read every home in the polygon and surfaces high & medium likelihood candidates. The product loop in its simplest shape, working.