Each grant's title and abstract are scanned for four kinds of buying signal. Points from each add up to the score shown next to a lead.
An award (usually S10) describing an intent to purchase a specific light microscopy instrument. The recipient has funds awarded right now and might be open to switching vendors. These jump straight to the top of the prospect list regardless of score.
Naming a specific microscope brand or model, such as Zeiss, Leica, Airyscan, LSM 980, and similar. The strongest signal: near absent from a grant that isn't about buying or using that equipment.
The actual microscopy method described: confocal, two-photon, super-resolution, TIRF, and similar. Real hands-on capability, not a passing mention.
The researcher runs or depends on a shared imaging facility (e.g. "microscopy core"). Counts on its own even with no specific technique named.
Just says "microscopy" or "fluorescence" with nothing specific. Worth very little, and capped so it can never carry a lead by itself.
Enter your name and work email to access the dashboard.
Your access request is under review. You'll get an email if it's approved.
Used to label the leads you triage and contact, so the rest of the team can see who's working what. First and last name, e.g. "Jared Evatt". Not a password, just a label.
How to work each of the three tools, in full. Pick a tab, everything under it is reference for that one page.
Every row is a researcher (or, for DoD, a single award) that recently received federal grant money AND shows real evidence of light microscopy work in the proposal itself. Grant money alone is not a lead, and microscopy interest without new money is not a lead either: what we look for is money plus timing.
The list covers a rolling 12 months and refreshes early every Monday morning. Each refresh re-pulls NIH, NSF, and DoD, re-scores everything, and adds the new week on top. The line under the three big numbers shows when data was last pulled and how many researchers are in it now.
It's deliberately not every federal award: awards under $100,000 are dropped unless they're instrument-purchase awards, electron and cryo-EM work is out of scope, and anything that doesn't clear a real buying-signal bar never appears. A missing researcher you know well likely just had award text with no microscopy signal.
Click anywhere on a row to expand it: the full grant title (linked to the agency's page), the funding institute or NSF directorate or DoD awarding branch, the NIH activity code, how many qualifying grants this person had, the combined total, and a line-by-line breakdown of where every point came from.
CRITICAL is not "a very high score", it's a different kind of lead and skips scoring altogether: the award itself is an instrument-purchase award that names light microscopy equipment. The money is already committed, and the buyer can choose any comparable system, not only the one written into the proposal. Four things reach this tier:
All four only qualify when the text names real light microscopy: a specific vendor or model, or a strong technique. A passing mention of "core facility" or the bare word "microscope" isn't enough, deliberately, since that used to wrongly promote electron microscopes, microCT scanners, and ultrasound into this tier.
STRONG means a score of 35+, or a named vendor or model plus a serious imaging technique: real, demonstrated capability and fresh money, without Critical's "buying right now" certainty.
MODERATE means a score of 16+, a named vendor with at least 10 points, one strong technique, or a named imaging facility. Worth a look, but thinner evidence and more judgment calls.
The grant's title and abstract are scanned for four kinds of language. Every match adds points, and the total is the score. Expand any row for the exact arithmetic, term by term.
Four rules fix real false positives:
The two New PI badges are the ones to watch: fresh money, no incumbent vendor, and time-sensitive.
Leads are grouped into Monday-start weeks by when this tool found them, not by when the grant was awarded. Each lead still shows its true award date in its own column.
The two differ because a lead can reach us long after the award. Most NIH records are published within about a week, but a minority arrive months late, and a lead that showed up in this Monday's pull but was funded in May used to get filed into an old, collapsed week nobody reopens. Now everything a run turns up lands together in that run's week, and a week stops changing once it has closed.
Browsing everything, the newest week is open and older weeks sit collapsed beneath it. Filter to a rep, a territory or a search term and every group with a match opens, so a filtered view never looks empty just because your leads are in older groups. Each header shows the lead count, split by tier, and share already reviewed. Initial backlog at the bottom holds everything found before weekly tracking began; it is one historical batch, not a week, and it is where most of any one territory currently sits. Expand all weeks and Collapse all switch between browsing everything and resetting to the current week.
Every row asks one question first: Good fit or Bad fit. Good fit unlocks Contacted underneath it, marking that you reached out. Clicking Contacted again clears the row back to untouched, with a confirmation first since it erases who marked it and when.
Bad fit is for a lead you've rejected: already a customer, the wrong kind of lab, the money clearly spoken for. The row fades into the background rather than disappearing, and clicking Bad fit again brings it back. Nothing is ever deleted.
Use both buttons generously, Bad fit especially: a rejected lead marked stops a teammate reaching the same conclusion twenty minutes later, and it's the only way this tool learns what a bad lead looks like. It counts as work in the weekly reviewed figure exactly like Good fit does.
The first time you mark anything you'll be asked for your first and last name, remembered in this browser as a label, not a login, so the row can say "Contacted by Dana Whitfield." Status is shared and live for the whole team, one status per lead.
Once a lead is marked Contacted, one more question appears: did they write back? Answer with thumbs up or down, a fact rather than a judgment, and the only way anyone can tell whether these leads are any good.
A thumbs up prompts was it a real prospect?, shown only on leads that replied. Tapping a lit thumb again clears it, so a wrong tap costs nothing.
Leads contacted more than two weeks ago with nothing recorded get an Awaiting result flag, collected by that filter and counted in the Monday email. Two weeks is deliberate, so a slow but genuine reply doesn't get recorded as a dead end.
These answers will eventually show whether Critical leads outperform Moderate ones, and whether the scoring needs changing, which takes a few months of real answers.
Draft Customized Email researches the person and writes a first draft: it searches PubMed for recent papers, finds their lab or faculty page, and writes a short note grounded in that work. Each step ticks past, usually 20 to 45 seconds. Available on NIH and NSF leads.
First it asks how technical the email should read. Straightforward is plain language, safe with no preparation. Informed is the default: one specific paper and one real imaging question. Technical names the imaging method, so only pick it if you can back that up in a qualification or discovery meeting. Your last choice is remembered.
What comes back has three parts: a For your reference box explaining the framing (never part of the email itself), the draft in an editable box, and Sources used listing what it actually read.
The email opens a scientific conversation, not a pitch: a real question about their work, never a product mention, closing with a request for a short meeting.
Nothing is ever sent for you. Use Copy, paste into your mail client, and edit it to sound like you, checking every claim first. The draft is saved, so the button becomes View Draft and reopens it instantly, and Regenerate writes a fresh one at the same depth.
There's one saved draft per lead, shared by the team; the stamp above shows which depth it was written at and who ran it. Rewrite at writes a new one at a different level, replacing the saved copy for everyone.
Dashboard access signs you up automatically for a short digest after each weekly refresh, no button or form needed. It names every new Critical lead, the total new leads, and how many from the past 30 days are unreviewed. Every lead is announced exactly once, so a late-arriving one still gets its turn.
My Territory, in the header of the grant leads and faculty pages, scopes your Monday email to specific reps: your own name, or several if you manage a region. Leave nothing picked and you get every lead, the default.
Every digest has an unsubscribe link, effective immediately. If you change your mind, sign up again here; dashboard access is unaffected either way.
The Feedback button, in the header of the grant leads and faculty pages, opens a board the whole team can see. Post a bug or idea, and upvote anything already there instead of re-posting it. The button's number is how many items are on the board.
Access is by work email; a company address is recognized straight away, anything else creates a request reviewed by hand, with an email if approved. Sessions last 30 days per browser, and signing in on any of the three pages carries over to the other two.
Scores and tiers come from the grant text as published, so an odd-looking lead usually traces back to an odd abstract. Expand the row first, the evidence breakdown almost always explains it. If it still looks wrong, post it on the Feedback board with the name and institution.
Every row is a researcher who started a faculty post in the last 12 months, the early half of the same pipeline as the Grant Leads tab: a new hire normally gets a startup package with equipment money a year or two before their first funded grant shows up over there. Working a name here means reaching them while the money is unspent and no vendor relationship exists yet.
It's deliberately not every new faculty appointment in the country. A row only appears when the researcher published the move themselves, on ORCID or in their own post, so the list is small, recent and verifiable rather than large and inferred.
Click anywhere on a row to expand it: full research focus, the ORCID record, department and title, every source link, and why it cleared the bench-research check.
Which one you'd rather have is a judgment call about your territory, so hire type is shown next to the name, not scored or ranked, with no equivalent of the grant page's tier here.
The ? Hire types button in the filter bar opens a shorter, color-matched version for a quick check while you work.
Focus tags come from what each person actually publishes on, never their department name: a psychology department can house an imaging neuroscientist, so department alone never decides who's in or out.
The two source dates matter for timing. An ORCID start date is usually reported late, so the job began before we saw it, sometimes by months. A post can run the other way: someone announcing in August that they join next January is dated August, and their startup money doesn't exist yet. Open the post before writing to anyone you treat as urgent.
The dark bars group rows by month, using that signal date, so the page reads like an ongoing hire history. The catch: a late-reported ORCID start date can land a brand-new lead in a collapsed bar from months back. A row with only a year on record gets its own "Sometime in [year]" bucket rather than a guessed month.
That's what the NEW tag is for: it marks anything found in roughly the last 30 days regardless of which month bar it's under, and a collapsed bar with a fresh row shows an "N new" count on its header. Inside a bar, the freshest signal date sorts to the top.
Good fit and Bad fit are the first call on every row. Contacted unlocks once you choose Good fit. A hire marked Bad fit fades into the background rather than disappearing, and one more click brings it back. Your name goes on the row so nobody works the same hire twice; clicking Contacted again clears it, with a confirmation first.
The three figures at the top mirror the Grant Leads tab, read against this dataset: untouched first-faculty-job hires, hires in the last 30 days, and how much of the newest batch the team has worked.
Door-to-Door is a lookup, not a list. Type a PI's name, an institution, or both, and it pulls that person's active NIH-funded projects straight from RePORTER, whether or not they show up on the other two tabs. Use it before a cold visit, when a name comes up in conversation, or for anyone the automated lists haven't flagged, because their award isn't recent enough or the abstract alone didn't clear the scoring bar.
There's no tier, score, or team queue here, and nothing you look up is tracked as a lead the way the other tabs track Good fit, Bad fit and Contacted. It answers one question on demand: what does this person's public work say about their imaging needs.
Enter a PI name, an institution, or both, either alone is valid. Multiple matches produce a pick list with each candidate's institution and current project title. A search with no results at a named institution offers to retry on the institution alone, in case the name was slightly off.
Picking a match loads a quick reference card: the grant's own abstract and up to two recent publication abstracts, pulled straight from RePORTER and PubMed with nothing added. Meant to feel instant, and costs nothing to generate.
Generate Prospecting Profile is the deeper option: it reads the same material and has Claude write an assessment of what the science is, how central imaging is to it, and what it implies about instrumentation needs. Steps tick past as it runs, unless someone already generated this exact PI's profile in the last six months, in which case it comes back instantly. That's a quiet cache, not a shared list: it saves repeat NIH/PubMed/Claude calls when two reps look up the same person, but nobody can see a profile already exists until they search for that PI themselves.
Every claim must trace back to a specific grant abstract or paper Claude actually read; anything it can't support that way is left out. Sources used are listed at the bottom of the card.
Every profile is rated high, medium, or low, but the card only speaks up when it's low: a banner says plainly the evidence was thin, rather than padding out a guess. No banner means medium or high. Low is common for a brand-new PI with little published; treat it as a starting point, not a verdict.
Generate Outreach Email uses the same drafting engine as the Grant Leads tab. Pick a depth first: Straightforward is plain language safe with no preparation, Informed is the default and engages with one specific paper, and Technical names the imaging method, so only choose it if you can back it up yourself. See the Grant Leads tab above for the full explanation, it's the identical choice and email, just reachable here for a person with no active lead driving them onto that list.
The list under the search box is personal: the last ten people you've looked up, newest first, so you can reopen a card without searching again. Nobody else can see it, and this is separate from the six-month cache above: a profile can come back instantly because someone else generated it first, without either search showing up in the other's list.
Same sign-in as the rest of the site, see the Grant Leads tab above. Signing in on any of the three pages carries over to the other two.
Built by Jared Evatt, for the whole team.
A few recent updates to the site:
Bugs, feature ideas, anything that'd make this more useful, visible to the whole team. Upvote what you agree with instead of asking Jared in person.
A short Monday email listing the leads that have landed since the last one, with a link straight back here. One email a week, nothing else. Every email has an unsubscribe link at the bottom.
Optional: pick one or more reps and the Monday email only shows leads in those territories -- a manager can pick their whole region. Leave it as "No territory" to keep getting every lead, same as today.
Looking up what this researcher has published and where they work, then writing a short, personal note. This usually takes 20 to 45 seconds.