Custom Datasets

Production-grade data, exactly as you need it

A database has the same data for every customer. Sleuth adds the data your question needs, written to your rules, then builds the dataset to your spec, checks every delivery, and sends it into your product or the systems your team runs on, every week or month.

You set the rules, and Sleuth keeps it running

You decide what the dataset has to answer. Sleuth builds it, checks it, and keeps it current, so nobody on your team builds or maintains the data.

Step 1

Tell us the rows and the rules

Which records count, which columns you need, how to score them, and how often. If you’re not sure, we’ll propose the dataset.

Step 2

Check a sample

Sleuth builds the dataset and its checks, and you review the first delivery before anything goes on a schedule.

Step 3

Get it on your schedule

Each delivery has the data, what changed since last time, what was left out and why, and the source behind every value.

Step 4

Change it when you need to

Add a column or a rule, and it’s in the next delivery, at no extra charge.

The data your question needs, written to your rules

Sleuth starts from the standard database fields, then adds the data you asked for. Every value comes with its reasoning and sources.

“We’re a contract manufacturer for antibody-drug conjugates. For every approved ADC, tell us how hard it is to make, whether the sponsor uses outside capacity, where demand is heading, and where there’s an opening for us.”

Standard database columnsColumns databases don’t usually have
ProductSponsorTargetManufacturing complexityOutsourcing signalDemand to 2028Read for a contract manufacturer
Enhertu (trastuzumab deruxtecan)Daiichi Sankyo, with AstraZenecaHER2High: about 8 payloads per antibody, and a potent payload that has to be contained.Daiichi Sankyo reports long-term CMO commitments across its ADCs, but names no Enhertu supplier.Growing: new first-line and early breast cancer uses keep adding patients.Rising demand could open overflow fill-finish or contained conjugation work if Daiichi Sankyo rebalances supply.
Adcetris (brentuximab vedotin)Pfizer (Seagen), with TakedaCD30Medium: about 4 payloads per antibody, made with standard cysteine chemistry.An outside supplier made its drug-linker under a 2010 agreement. Current suppliers aren’t public.Shrinking: US sales fell in 2025 as competition grew.Falling demand and the cost of qualifying a new site make this a backup role at best.
Zynlonta (loncastuximab tesirine)ADC TherapeuticsCD19Medium: about 2 payloads per antibody, but a very potent payload to handle.ADC Therapeutics names outside manufacturers and reports supply commitments for 2026.Stable: later-line use is steady, and earlier-line growth is uncertain.Its outsourced model makes a qualified second source the most realistic opening.

Columns databases don’t usually have

Enhertu (trastuzumab deruxtecan)

SponsorDaiichi Sankyo, with AstraZeneca

TargetHER2

Manufacturing complexityHigh: about 8 payloads per antibody, and a potent payload that has to be contained.

Outsourcing signalDaiichi Sankyo reports long-term CMO commitments across its ADCs, but names no Enhertu supplier.

Demand to 2028Growing: new first-line and early breast cancer uses keep adding patients.

Read for a contract manufacturerRising demand could open overflow fill-finish or contained conjugation work if Daiichi Sankyo rebalances supply.

Adcetris (brentuximab vedotin)

SponsorPfizer (Seagen), with Takeda

TargetCD30

Manufacturing complexityMedium: about 4 payloads per antibody, made with standard cysteine chemistry.

Outsourcing signalAn outside supplier made its drug-linker under a 2010 agreement. Current suppliers aren’t public.

Demand to 2028Shrinking: US sales fell in 2025 as competition grew.

Read for a contract manufacturerFalling demand and the cost of qualifying a new site make this a backup role at best.

Zynlonta (loncastuximab tesirine)

SponsorADC Therapeutics

TargetCD19

Manufacturing complexityMedium: about 2 payloads per antibody, but a very potent payload to handle.

Outsourcing signalADC Therapeutics names outside manufacturers and reports supply commitments for 2026.

Demand to 2028Stable: later-line use is steady, and earlier-line growth is uncertain.

Read for a contract manufacturerIts outsourced model makes a qualified second source the most realistic opening.

“We distribute cell and gene therapies to treatment centers. For every approved gene therapy, tell us how hard it is to handle, what a site needs to be ready, where demand is heading, and what that means for us.”

Standard database columnsColumns databases don’t usually have
ProductSponsorGene or targetHandling burdenWhat a site needsDemand to 2028Read for a distributor
Hemgenix (etranacogene dezaparvovec)CSL BehringFactor IX geneMedium: refrigerated rather than frozen, but each kit is made for one patient and used within 24 hours of dilution.Refrigerated storage, aseptic preparation, and a hematology team at a CSL-trained center.Growing: more US centers are trained and more patients are treated each year.Deliver refrigerated and timed to the infusion appointment, with no stock held at the site.
Luxturna (voretigene neparvovec)Spark Therapeutics (Roche)RPE65High: stored at −65°C or colder, with 4 hours from dilution to use.A designated eye gene therapy center with an operating room and a retinal surgeon.Shrinking: US treatment volume is likely to taper by 2028.Frozen delivery to a handful of designated eye centers, timed to the day of surgery.
Casgevy (exagamglogene autotemcel)Vertex, with CRISPR TherapeuticsBCL11A enhancerHigh: liquid-nitrogen vapor at −135°C, thawed one vial at a time and infused within 20 minutes.An authorized transplant center with liquid-nitrogen storage and a cell therapy team.Growing: first cell collections are rising ahead of infusions.A cryogenic shipment made for one patient, sent to an authorized center and timed to the infusion day.

Columns databases don’t usually have

Hemgenix (etranacogene dezaparvovec)

SponsorCSL Behring

Gene or targetFactor IX gene

Handling burdenMedium: refrigerated rather than frozen, but each kit is made for one patient and used within 24 hours of dilution.

What a site needsRefrigerated storage, aseptic preparation, and a hematology team at a CSL-trained center.

Demand to 2028Growing: more US centers are trained and more patients are treated each year.

Read for a distributorDeliver refrigerated and timed to the infusion appointment, with no stock held at the site.

Luxturna (voretigene neparvovec)

SponsorSpark Therapeutics (Roche)

Gene or targetRPE65

Handling burdenHigh: stored at −65°C or colder, with 4 hours from dilution to use.

What a site needsA designated eye gene therapy center with an operating room and a retinal surgeon.

Demand to 2028Shrinking: US treatment volume is likely to taper by 2028.

Read for a distributorFrozen delivery to a handful of designated eye centers, timed to the day of surgery.

Casgevy (exagamglogene autotemcel)

SponsorVertex, with CRISPR Therapeutics

Gene or targetBCL11A enhancer

Handling burdenHigh: liquid-nitrogen vapor at −135°C, thawed one vial at a time and infused within 20 minutes.

What a site needsAn authorized transplant center with liquid-nitrogen storage and a cell therapy team.

Demand to 2028Growing: first cell collections are rising ahead of infusions.

Read for a distributorA cryogenic shipment made for one patient, sent to an authorized center and timed to the infusion day.

“We forecast approval timing for ADC programs. For every approved ADC, tell us what sped it up or slowed it, when approval first looked likely, which earlier ADC it most resembles, and what that teaches us.”

Standard database columnsColumns databases don’t usually have
ProductTargetYears to approvalWhat sped it up or slowed itEarliest signalClosest comparableRead for a forecast
Enhertu (trastuzumab deruxtecan)HER24.3Sped up: no dose-limiting toxicity in the first trial, then one single-arm Phase 2 in heavily pretreated HER2-positive breast cancer.Oct 2016: 5 of 12 patients already treated with Kadcyla responded in the first human trial.Kadcyla, the earlier HER2 ADC in the same breast cancer setting.A strong early signal in patients who failed the last drug in the class can justify a fast single-arm path.
Kadcyla (trastuzumab emtansine)HER26.8Slowed: FDA refused to file the first application in 2010, so approval waited for the randomized EMILIA trial.Dec 2006: one of the first seven patients in the first human trial had a partial response.Adcetris, which reached approval on single-arm response data.When regulators want randomized evidence, add the length of a Phase 3 to the forecast.
Besponsa (inotuzumab ozogamicin)CD2214.0Slowed: a Phase 3 in aggressive lymphoma stopped for futility in 2013 before the program turned to leukemia.May 2011: 20 of 36 patients with relapsed ALL responded in a Phase 2 study.Mylotarg, which uses the same calicheamicin payload in a blood cancer.A failed first indication can add years, so date the forecast from the indication that worked.

Columns databases don’t usually have

Enhertu (trastuzumab deruxtecan)

TargetHER2

Years to approval4.3

What sped it up or slowed itSped up: no dose-limiting toxicity in the first trial, then one single-arm Phase 2 in heavily pretreated HER2-positive breast cancer.

Earliest signalOct 2016: 5 of 12 patients already treated with Kadcyla responded in the first human trial.

Closest comparableKadcyla, the earlier HER2 ADC in the same breast cancer setting.

Read for a forecastA strong early signal in patients who failed the last drug in the class can justify a fast single-arm path.

Kadcyla (trastuzumab emtansine)

TargetHER2

Years to approval6.8

What sped it up or slowed itSlowed: FDA refused to file the first application in 2010, so approval waited for the randomized EMILIA trial.

Earliest signalDec 2006: one of the first seven patients in the first human trial had a partial response.

Closest comparableAdcetris, which reached approval on single-arm response data.

Read for a forecastWhen regulators want randomized evidence, add the length of a Phase 3 to the forecast.

Besponsa (inotuzumab ozogamicin)

TargetCD22

Years to approval14.0

What sped it up or slowed itSlowed: a Phase 3 in aggressive lymphoma stopped for futility in 2013 before the program turned to leukemia.

Earliest signalMay 2011: 20 of 36 patients with relapsed ALL responded in a Phase 2 study.

Closest comparableMylotarg, which uses the same calicheamicin payload in a blood cancer.

Read for a forecastA failed first indication can add years, so date the forecast from the indication that worked.

Rather build datasets yourself? Studio builds them for any questionOr build datasets yourself in Studio

Data your product or operations depend on

Sleuth sends each delivery, weekly, monthly, or quarterly, to the tools your team already works in, with alerts in Slack or Teams when something changes.

In your product

For when your customers need biopharma data directly in your product. Sleuth sends a weekly feed with a source behind every value.

In your operations

The list a planning, site, or account team works from each month, in the tools it already uses, next to what you already know.

In your models and agents

Structured predictions your own tools can use, such as when a trial will start or how likely approval is, each with its reasoning.

Weekly

A feed straight into your product, with what changed since last week.

Monthly

A planning list in the tools your team already uses.

Quarterly

A scorecard in Excel or CSV, ready for a quarterly review.

Talk to us about your systems

Checks written for your data, run on every delivery

Nothing reaches your product or your team unchecked. Sleuth checks every value when it first takes it in, then checks your dataset against your spec before each delivery.

Checks written for your spec

Your rules for which rows count, how each column is filled, and what has to tie out across the data. Checked for every value prior to delivery.

Every value labeled

Each value is marked sourced, inferred, or unknown, with its sources and a confidence score, so you know which ones to lean on.

A change log with every delivery

Each delivery is checked against the last one, and the change log shows what changed, why, and where it came from.

Case Study

A weekly dataset inside a customer’s product

A health-tech company helps hospitals decide which advanced therapies to offer and when to start getting ready. Sleuth built the dataset its platform shows them. Every weekly delivery goes through checks written from its spec.

Their column

What it answers

Hospital preparation lead time

How long a hospital needs to get ready before it can give the therapy

Infrastructure tier

What it takes on site, from an infusion chair to a cell-processing lab

Eligible patients per academic center

How many patients a large center would treat each year

Expected price tier

Where the price is likely to land, from modality, indication, and precedent

Approval confidence, 0 to 100

How likely approval is, scored from five factors

Read the full story

Tell us the dataset you need

Set up in Claude, ChatGPT, or Copilot