I Published the Number That Makes My Own Database Look Bad

RepWatchr is mine. I built it, I own it, and I am about to hand you the worst number on it.

There are 8,914 official profile records in the directory. Of those, 8,615 have a source check date older than ninety days, or no usable date at all.

That is not a leak. That is printed on the coverage page, in a heading, with the calculation date stamped on it.

Most companies would never put that number in public. I want to explain why I did, and what it has to do with whether your business actually owns anything.

The reason is sitting on a different page

RepWatchr has a page offering scoped public data pilots to newsrooms, researchers and agencies. The proposed tiers run from $49 a month for up to 100 records, to $249 a month for up to 500, to a figure starting at $999 a month for up to 2,500.

That page also says, in plain type, that the public API and paid data subscriptions have not launched, and that sending a request starts a scope review rather than issuing a key, reserving a dataset or collecting payment. It calls the tiers proposed pricing and states they are not active subscriptions.

Then it gives the order of operations. Name the scope. Review a sample. Set delivery terms.

Agree on the data before the price.

You cannot honestly tell a buyer to review a sample and then hide the audit of the thing they would be sampling. If the coverage page said nothing about stale dates, that three step process would be theater and the sample review would be a formality.

So the ugly number goes on the page. It is the thing that makes the process real.

What the number actually means

It does not mean 8,615 records are wrong.

It means nobody has confirmed them recently, and the system knows the difference between those two statements.

A record can be perfectly accurate and still be past due for a check. A record can also have been re imported last week and be a year out of date. That is exactly why an import date and a verification date cannot be the same field.

The page says it too. A later import or edit date does not confirm that a person still holds office.

That distinction is the product. Anybody can scrape names. Knowing which of your names you have personally looked at lately is the part that takes work.

Now the ownership question

Here is the part that matters for your business, and it took me a while to see it clearly.

Look at what is free.

The coverage page links out to Congress Legislators and Open States People, two public projects that publish reusable identity data for federal and state officials under CC0. Anybody can take that. I did. So did everybody else who builds in this space.

Then it says the sentence that reframes the whole asset. Neither one provides a complete national local office roster.

Go look at the state table on that page. Federal and state numbers in all fifty states. The county column reads zero in every state except Texas. School board coverage is Texas only.

So the part of this database that came free is the part everybody already has. The part nobody has is the county commissioner, the city council seat, the school board trustee in a town of two thousand people. There is no national file for that. It does not exist. Somebody has to go get it one roster at a time.

That local layer is the only part of this I actually own.

One district, and what it costs

The first refresh listed on that page is Harleton ISD in Harrison County. Seven trustees. Checked September 8, 2026, against the district own roster page.

The entry does not just list them. It records that six are listed as elected and one as appointed in July 2026 with no term expiration listed. It notes the superintendent is an administrative role and is not one of the seven. It notes one earlier name in the file is retained with a departure date that is unconfirmed. It notes an older spelling of one trustee name was matched to the existing record instead of counted as another seat.

One school district. Seven seats. Four separate caveats.

There are 1,194 Texas school districts with records in the system and 61 districts flagged as research priorities.

Now multiply that out. That is your answer for why no free national local file exists.

I am not naming those trustees here. They sit on a public roster and their records are public, but this piece is about how the data gets built, not about any one of them, and not one of them did anything.

One more boundary worth stealing

The site also runs a free claim path for candidates and officials, open to every party including nonpartisan and school board offices, with no payment required to request one.

It states that a claim is not an endorsement, a ballot certification, or verification of every statement. And it states that payment cannot buy a better grade, remove sourced criticism or approve a claim.

Write that rule down before somebody offers you money to move a number. It is a whole lot easier to hold a line you already published than one you are deciding in the moment.

The counts that look sloppy until you read them

Three different totals show up across this site, and they mean three different things.

The home page counter reads 16,835 public profiles. The coverage page reads 17,041 searchable records. The data page reads 8,914 official records.

That looks careless until you read the coverage page headline, which gets there first. A profile, a school district, and a verified officeholder are different counts.

I ran the math myself. 8,914 official profile records plus 8,135 named school research records is 17,049. The page publishes 17,041 searchable records and notes that 8 reviewed identity links were joined. 17,049 minus 8 is 17,041.

It reconciles. That subtraction is mine, not theirs, but it holds.

What the page will not give you is the total number of unique people, or how many are current elected officeholders. It says so directly. Not yet established.

That is the correct answer. I would rather publish not yet established than publish a number I have to walk back later.

What this means for your business

Take the free thing. Everybody does, and there is nothing clever about refusing to.

But be honest with yourself about what it is. A free national file. A platform follower count. A marketplace buyer list. A directory anybody can copy in an afternoon. None of that is yours. You are renting it, and so is the guy across town, at the same price.

The part you own is the part you had to go get. The customer list you built with your own hands. The local knowledge nobody else bothered to write down. The records you checked yourself and dated.

And the test for whether you really own it is simple.

Can you say what is missing from it?

If you cannot answer that, you do not have an asset. You have a pile.

Publish the gaps. It is the only way anybody can tell the difference between a dataset and a dump.

What I have to fix

I am going to put the fix on the record with a date on it, so it is not just a nice paragraph.

8,615 past due checks is too many. The field exists. The queue is visible. Now it has to get worked, county by county, the same way Harleton got worked.

And a smaller one. A reader who lands on the home page sees 16,835. A reader who lands on coverage sees 17,041. Neither page tells them why those are different numbers. The coverage page explains the concept once you get there. The home page does not send you. That is one sentence and one link, and I will take the note.

If you want to help, the source submission page does not require an account.

Do not threaten anybody. Do not harass anybody. Do not contact anybody in my name.

Send the roster. Send the agenda. Send the filing with the date on it.

Put eyes on the record.