Lumen — what's new

LUMEN

What's new

v1.153 2026-09-19

Pharmacies now come from the state license list

The 2027 draft QAP counts a licensed community pharmacy, and the Texas State Board of Pharmacy publishes every license. The map now shows exactly those: 5,291 Class A and A-S community pharmacies statewide, one dot per license, with the license number in the popup. Dallas County goes from 243 OpenStreetMap pharmacies to 482 licensed ones.
The assumed in-store pharmacies are gone. A Kroger or H-E-B with a pharmacy holds its own license and appears under its own name; a grocery without one no longer gets a guessed pharmacy at 0 ft. The old OpenStreetMap and assumed rows are kept as superseded under check features; verdicts you had already entered stand.
Every pharmacy carries its license number, status and expiry in the flag. Licenses that are revoked, expired or delinquent (260) are screened out with the reason. Closed-door pharmacies (346: long-term-care, clinic-only, mail-order, which the Board says are not open to the public) and sterile-compounding specialty pharmacies (357) still count but are flagged first, so you can rule them out.
The license list has addresses, not coordinates. Addresses were placed by the Census geocoder, then by matching house number and ZIP against Overture Places, then OpenStreetMap; anything placed by the looser methods says "location approximate" in its flag. 115 licenses, mostly on highway addresses, could not be placed and are not shown yet.
Nothing here changes the 2026 rules. Re-verify the pharmacy language when the 2027 QAP is adopted.

v1.152 2026-09-18

Urgent care clinics join the 2027 draft tie-breaker

The 2027 draft QAP out for public comment adds a sixth proximity feature under §11.7(2)(A)(vi): an urgent care clinic. Under "2027 draft rules" the score readout, the parcel search, the heat grid, the deal sheet and the printed report now measure urgent care alongside park, school, grocery, library and pharmacy. Still the sum of the three closest.
The draft's test is strict, and none of it is on any state list: 10 operating hours on weekdays, open both Saturday and Sunday, walk-ins accepted, all ages seen, and radiology on site. Freestanding ERs (Health & Safety Code ch. 254, including "ER & Urgent Care" hybrids) and licensed hospitals are excluded. Every urgent care therefore carries a verify flag spelling those tests out; where OpenStreetMap has posted hours the readout says whether they pass the 10-hour / 7-day test.
Pre-screened out by name so they never score: ERs and emergency centers, hospitals, pediatric-only clinics (Urgent Care for Kids, Little Spurs, Children's), and retail clinics with no X-ray (MinuteClinic, Walgreens, the Little Clinic). Occupational-medicine chains such as Concentra stay in but are flagged to confirm weekend hours and that children are seen. All of these show under check features with the reason.
Clinics come from two open sources: OpenStreetMap (thin, but with hours) and Overture Maps Places (much fuller). Listings are matched to their parcel like every other feature, so the distance is parcel edge to parcel edge.
The 2026 rules are unchanged: four features. Nothing here is adopted yet. Re-verify the final language when the 2027 QAP is approved.

v1.148 2026-09-17

Ask Lumen: a chat that can see your screen (bring your own key)

New button in the left panel: ask Lumen about this. It opens a chat on the right side of the map. Every question goes out with what is on your screen: the parcel you clicked and its full readout, the search results on the current page, the QAP year and target population you have set, and which layers are on. So you can ask “why is this score what it is?”, “which of these results are vacant and under 3,000 ft?”, or “what does item F mean?” and get an answer about that site.
It answers from the readout only. If nothing is selected it says so and tells you what to click; it will not guess a distance that is not on screen. Eligibility calls (is that park public, is that grocery full-service) it flags as the consultant's call.
The line at the top of the chat says what it can see right now. The ↺ button starts a fresh conversation; the conversation otherwise stays for the browser tab.
Runs on your own Anthropic account: the first time you open the panel it asks you to paste an API key from console.anthropic.com. Lumen checks it, keeps it with your account, and never shows it again (a line under the chat says which four characters it ends in, with a remove link). Your questions bill to you; nobody's key is shared with anyone else.

v1.145 2026-09-15

A walk-through list of the SPARK parks

New button in the left panel: SPARK school parks (Houston). It opens a list of all 158 on the right side of the map, like check-features, but with nothing to decide: click a park and the map jumps to its campus with the parcel outlined, and the usual score readout fills in on the left.
Each one you open gets a tick so you can see how far through the list you are. Press n for the next park and b to go back. The ticks are kept in your browser; the ↺ button clears them.
Search by school name, or show only the ones not yet looked at. Parks that could not be placed by their TEA campus record say so under their name, and every entry links to SPARK's own map pin for a quick comparison.
Also fixes check-features: opening a feature there tried to score its spot and silently failed. It scores now.

v1.144 2026-09-14

SPARK school parks added as public parks (Houston)

All 158 SPARK School Parks on sparkpark.org are now park features: community parks built on public school grounds across Harris County and the Fort Bend edge, open to the public after school, on weekends and on holidays at no charge.
Each park is placed on its school campus and measured to the campus parcel boundary, the same way every other feature is measured. The school's appraisal-district parcel is used wherever one exists (all but three).
They load as owner-verified parks and show a flag saying what they are. The 2026 QAP text excludes school campus facilities from the park feature, so a reader can see the basis at a glance.
SPARK's own map links are wrong for a few schools (Duryea and Edison point at each other). The school name was trusted over the link and the location taken from the TEA campus record.

v1.142 2026-09-11

Site search now finds every qualifying parcel

The parcel search was missing real sites. It ranked parcels by a rough guess and only measured the best 900 exactly — but the guess overstates distance badly next to a large parcel, so qualifying sites could rank thousands deep and never be measured. In one county it listed 895 sites when 3,753 qualified, and missed one at 140 ft.
Now every parcel that could possibly qualify is measured exactly. Checked on Kaufman County: the new search found everything the old one did plus 38 sites it had never listed, the closest at 937 ft, and ran faster.
Some distances came down. Where the old search said 2,860 ft the click readout says 2,476, and the new search now agrees with the readout — the two paths measure the same way at last.
Your feature corrections from the check-features panel now apply to search results as well as the click readout.
The deep-search toggle is gone; there is nothing left for it to do.

v1.137 2026-09-04

Opportunity Zones 2.0 on the map

New layer: the 605 census tracts Texas nominated for Opportunity Zones 2.0 on 4 September 2026, across 105 counties.
These are nominations, not designations. Treasury certification is expected 28 November 2026 and the zones would run 1 January 2027 to 31 December 2036. The layer and the report both say so. OZ 1.0 is unaffected and runs through 2028.
Every scored site says whether it is in one, and it appears on the parcel report beside the basis boost. It is a financing consideration, not a QAP scoring item, so nothing about the tie-breaker changes.
Source is the Governor's Economic Development & Tourism office. Re-runnable with tiebreaker_oz.py when Treasury certifies.

v1.134 2026-09-02

Filter the feature list by chain

The check-features panel now filters by brand. Pick H-E-B, Kroger, Tom Thumb, Costco, Sam's Club, Brookshire or any other chain from the list, with a count beside each.
“The big chains only” covers every recognised brand at once — 740 of Region 3's 1,208 groceries. “Independents / unbranded” is the other 468.
A search box filters by name as you type, and it composes with everything else, so “Kroger” plus “not yet reviewed” plus a search all narrow together.
Each row shows its chain, so a mixed list stays readable.

v1.129 2026-08-24

Underserved item (F): the population gate is now worked out for you

Item (F) only applies where the site sits in a Place, or its ETJ, of 50,000+ in an Urban subregion or 10,000+ in a Rural one. The report used to just tell you to go and check that. It now looks the population up and says whether the threshold is met.
What it still cannot tell you is whether the site sits *entirely* inside that Place or its ETJ — that is a boundary question, and ETJ lines are set by each city and are not in any map layer we have. The report says so plainly rather than implying a clean answer.
The tract half of (F) was already right: your tract and every tract touching it, including ones touching at a single point, checked for awards in the last 10 years serving the same target population.
Also fixed: appraisal districts write the city as “MESQUITE (DALLAS CO)”, which stopped the place lookup finding a match. Sites in those counties now resolve, which fixes their urban/rural designation too.

v1.128 2026-08-24

Sign-in outage fixed

The site locked everyone out from Sunday evening until Monday morning with “this copy has been offline too long”. Nothing was wrong with anyone’s account.
The cause: a licence check meant for installed desktop copies, which have to check in periodically or lock. The credential it used to check in expired, and the website counted itself as an unlicensed stale copy.
That check no longer applies to the website. It is the system that issues access, so it has nothing to check in with. Installed copies are unaffected — the check still works there.
Removing or changing a user still takes effect within five minutes, and now without depending on any credential that can expire.

v1.126 2026-08-21

Print a whole region, or the whole board

“Print all” in the dashboard header runs a deal sheet for every deal on the board — one document, one page per deal, straight to the printer.
Each column in the Regions view has its own printer button, so you can run just Region 3.
Sheets come back sorted by county, so a stack of them is in a sensible order to work through.
The satellite pictures are cached after the first build, so re-printing a board is quick.

v1.125 2026-08-21

A one-page deal sheet you can write on

New button on every deal in the dashboard: a single printed page with the deal name, address, the numbers that matter, and a satellite picture of the site with the parcel outlined — about half the page.
The bottom half carries whatever notes are already in the portal, then ruled lines to write on. Hand it to someone who works on paper, and type their marks back in afterwards.
It is not the screening report cut down. The report is for deciding; this is for carrying around, so it stays to one page however much is on file — a chatty deal spends its writing room on notes rather than spilling onto a second sheet.
The deal number is printed in the footer, so a scanned sheet can be matched back to its deal.

v1.124 2026-08-19

Seventeen stores were sitting in the wrong place

Checked every mapped store against the street address on file, then settled each disagreement against the county appraisal district — a parcel the chain owns is what scoring measures to, and it is independent of both address records.
Seventeen stores moved onto their real site, including a Randalls in Lakeway that was 3 miles out, a Kroger in New Caney sitting on a homebuilder’s land, and Walmarts in Greenville, Manor and Woodville.
Two Houston Aldis turned out to be swapped rather than misplaced — each was sitting on the other’s site, both correctly, so they were left alone. Moving one would have stacked two stores on one site and emptied a real one.
Confirmed positions are now recorded, so the check stops re-reporting stores that have already been settled. Thirteen remain unresolved, mostly addresses that will not geocode.

v1.122 2026-08-19

Nine stores were on the map in the wrong place

A Walmart Supercenter in Greenville was mapped 6.7 miles from its own address, so sites beside it read as having no grocery nearby. Checking every store against its street address on file found 49 more than 1,500 ft out of place.
Nine were confirmed wrong and have been moved onto their real address and parcel: Walmart 427 Greenville, Aldi Fulshear, H-E-B Elgin, Brookshire Paris, H-E-B Palmhurst, Sam’s Club Round Rock, Target and Walmart in San Antonio, and Brookshire Brothers Teague.
Six others turned out to be mapped correctly, with the address record at fault instead — those were left alone.
Added a standing check that compares every mapped store against the street address on file. That is what would have caught Greenville automatically — a store mapped in the wrong place is invisible to a missing-store check, because as far as that is concerned the store is on the map.

v1.121 2026-08-19

Grocery pipeline report, on the map

New button in the layers panel: every new grocery store being built in Texas, from the state construction-permit register — 6,002 permits swept, 188 of them a real store.
Shows what is already open, what finishes later in 2026, and what is coming in 2027 and beyond, with cost and square footage on each.
Each open store says whether the tool has it, and where it does not, whether that is a real gap or just a permit address that geocoded badly. Only 24 of the 138 open stores are genuinely worth a look, and most of those are 2026 openings that no source has caught up with yet.
Like the OI flip report, the link needs no login, so it can be forwarded to partners.

v1.114 2026-08-19

Grocery data now checks itself against state permits

Swept the state construction-permit register properly this time — 6,002 projects, of which 190 are genuinely a new grocery store. The first pass had looked at 82.
Checked all 102 of those stores that should already be open against the map, one at a time, using the appraisal district to tell a genuinely missing store apart from a permit address that simply geocoded badly. Most of the apparent gaps were bad addresses.
Added the Mesquite Walmart Neighborhood Market on Cartwright Rd, which no other source had yet — it opened in June 2026 and is too new for the USDA register, the 2025 appraisal roll, and OpenStreetMap alike. Pinned to the Walmart-owned parcel, so a site next door now measures 0 ft instead of 4,150.
Moved the Irving H-E-B onto its own parcel; its listed address sat 1,709 ft off the store.
The whole refresh now ships as a script that can be re-run, so the grocery list can be brought current on demand instead of going stale again.

v1.113 2026-08-19

Every new grocery store in Texas, added

The grocery list came from the USDA's SNAP retailer register, but as a one-time snapshot — every store authorized after it was taken was simply missing, and a missing store makes every site near it look worse than it is. Re-pulled the live register and added 121 supermarkets and warehouse clubs across the state that weren't on the map.
Cross-checked against the state construction-permit register: 62 new grocery buildings have been permitted in Texas, 29 of which should already be open. 24 are now confirmed mapped; the rest are newer than any source that could place them.
107 stores turned up in the refresh that were second copies of stores already mapped, sitting a little off. Those were dropped rather than added — a duplicate in the wrong place hands a site credit for a store that isn't there. The USDA's coordinates scatter by up to two miles, so they were matched by chain and store number instead of by distance.
New stores measure from the appraisal district's parcel for that store, not from an address point, so a site next door reads the real distance.
252 small-format stores from the same refresh are loaded but held out of scoring until they can be checked — the USDA files corner stores and bait shops in the same category as real neighborhood groceries.

v1.104 2026-08-04

Big accuracy pass on feature matching

Audited all 81,000 mapped amenities against the appraisal district's own land-use records. A grocery can no longer be matched to a house, a vacant lot, or farmland, and a school can't be matched to an empty lot — 1,173 bad matches corrected statewide.
Fixed the matcher reaching too far when a store or school's mapped location was off. It used to search half a mile and could grab the wrong campus or a different store of the same chain; Houston-area bad placements dropped from 155 to 59.
111 matches that failed multiple independent checks now measure from the store's own mapped location instead of pointing at the wrong building.
New warnings on the score readout: '⚠ matched 1,206 ft away' when a feature's location is far from the parcel it was matched to, and '⚠ 59 ac parcel' when a store is counted from a large shopping-center parcel.

v1.98 2026-08-03

Search a distance range

Search a band instead of just a ceiling: 'parcels between 500 and 1000 feet', '500-1000 ft in Rockwall', 'over 800 ft in Collin county'. Miles work too.
The results filter bar now has a min and max distance box.
Remove individual results from a list with the ✕ button while you work through them, and type a page number to jump straight to it.

v1.92 2026-08-02

Works properly on a phone

The map now fills the whole screen on a phone, with the panel sliding in from a button instead of squeezing the map into a sliver.
Search results come up as a bottom sheet you can thumb through.
The deals board stacks into one column, and text boxes no longer make the screen zoom when you tap them.

v1.86 2026-08-01

Historic buildings and jobs data

Type 'historic' in the search to find buildings 50+ years old — the threshold for a Texas Historical Commission determination, which is worth 2 points and can matter more than the tie-breaker.
Filter those by total building square footage ('historic over 20k sq ft'), counting all buildings on a parcel together so an old multi-building complex qualifies on its total.
Every scored site now shows primary jobs within 5 miles (2023 Census data — the same figures OnTheMap reports), with a download to verify it on the Census site yourself.
Every site also shows its official TDHCA urban or rural designation, and searches can filter to rural-only or urban-only places.

v1.78 2026-07-31

Share reports without a login

Send a parcel report to anyone with a link — no account needed. Useful for lenders, partners, and sellers.
Email a report straight from the tool.
Sign in with an emailed link instead of a password. Links last an hour and survive corporate spam filters.

v1.70 2026-07-29

Deals dashboard rebuilt

Full-screen board with drag-between-stages, plus Regions and Map views.
Rename your pipeline stages to whatever your team calls them.
Comments and tags on every deal, shared across your team.
Sort by last edited or last added to see what's moving.
Open a full parcel report straight from any deal.

v1.60 2026-07-28

New look, and a dark mode

Cleaner, lighter design throughout the map, reports, and dashboard.
A real dark mode — toggle it in the sidebar; the whole tool follows.
Amenity dots are now one button that expands to pick individual types, and the map loads faster because it no longer draws 80,000 dots on startup.

v1.48 2026-07-27

2027 preliminary Opportunity Index

The 2027 income quartiles, computed from the December 2025 census release using TDHCA's exact published method — roughly eleven months before their official workbook.
1,368 tracts change quartile; 205 become Q1 for the first time. Switch datasets from the sidebar to search and score against either year.
A side report showing every tract that gained or lost eligibility, with a live map.

← back to the map