What a Pipeline Database Can Miss in Your Competitive Landscape, and What It Costs to Get It Right

A Ferma comparison found 86 IBD programs absent from one traditional pipeline database's results, including three in Phase 3. We examine the gaps and estimate the analyst time and labor cost of building a reviewed landscape.

If you build competitive landscapes from a pipeline database, the workflow is familiar. Select the disease, filter by modality and development stage, export the programs, then build the phase grids, company profiles, and executive brief. That starting list shapes which competitors make it into the analysis.

Two requirements matter at the outset: relevant programs need to appear in the results, and their development status needs to be current. A stopped program listed as active can make the field look more crowded. An active program missing from the export can stay outside the team's assessment unless someone finds it through another source.

We examined that starting point in inflammatory bowel disease (IBD). Ferma compared its program record with the results of a traditional commercial pipeline database for a defined landscape question. The database returned 194 programs; Ferma returned 206. A difference of just 12 in the totals concealed a much larger difference in which programs appeared and whether they were still active.

For a CI team evaluating AI for pharma competitive intelligence, the useful comparison covers the whole job: the quality of the program set, the analysis built from it, and the effort required to produce and maintain a reviewed brief. The IBD findings below address the starting data. Ferma's workflow and the accompanying time and cost estimates then follow the work through to the deliverable.

About the evidence: Ferma conducted and scored the IBD comparison without independent adjudication. Eight entries returned only by the database remain unresolved. The workflow time estimates were reviewed by industry experts Ferma works with who regularly build competitive landscapes. They include human review and describe expected effort, not timings recorded in an IBD deployment. Cost calculations use an assumed hourly rate and workload.

What did the IBD database comparison reveal?

The question was specific: oral small-molecule and peptide programs in inflammatory bowel disease, from preclinical through Phase 3.

Ferma ran the comparison on 14 August 2026 against an established commercial pipeline database accessed through a live subscription. Both sides were assessed in the same week against the same publicly available evidence. All 194 programs returned by the commercial screen carried an active development status.

Ferma found 86 programs absent from the database results, including three in Phase 3. It also identified 63 stopped, suspended, or terminated programs that the database results listed as active. The table shows how the two returned sets overlap and differ.

The 63 programs classified as stopped, suspended, or terminated represented approximately 32.5% of the commercial screen's returned set. Ferma excluded them from its active landscape. Across the shared and database-only sets, four database citations could not be retrieved.

Eight entries that the database listed as active remain unresolved. Ferma's screening rules exclude entries that are not distinct assets, such as target classes and platforms, as well as development stages carried over from another indication. The comparison did not determine whether each of these eight entries is a valid exclusion or a Ferma miss.

Program status was assessed against the most recent retrievable primary evidence available at the snapshot. A program was counted as absent when it did not appear in the comparator's returned set for this query.

Ferma built its own program set and scored both sides without independent adjudication. The table above reconciles the totals for this defined query on 14 August 2026. These findings do not establish completeness across the entire IBD pipeline or describe the commercial database as a whole.

For the team building the landscape, the Phase 3 omissions warrant a closer look at which competitors reach the brief. The inactive entries warrant checking how crowded the active pipeline appears. A small difference in total program count can therefore conceal changes that matter to a phase grid, competitor prioritization, or the shortlist for further diligence.

How does Ferma turn a pipeline question into a competitive brief?

Once the program set has been checked, the team still needs to compare competitors, identify the findings that matter, and prepare the brief.

Ferma's pharma competitive intelligence platform connects its primary-source evidence base to agents that prepare the analysis and deliverable. The team defines the objective, selects the evidence fields and analytical views it needs, and reviews the result.

For the scope tested above, an IBD workflow would look like this:

1. Define the question and boundaries. For example: “Map oral small-molecule and peptide programs in IBD from preclinical through Phase 3, and identify competitors that warrant further diligence.” Specify the populations, inclusion rules, and intended audience.
2. Choose the evidence needed for the answer. Configure fields such as sponsor, mechanism, stage, efficacy, and safety, as relevant to the question and available evidence. Define what the team needs to compare before the analysis runs.
3. Build and enrich the program record. Bring together primary-source evidence, reconcile different names for the same asset, and check status and development stage within the relevant indication and population. Keep source links available for inspection.
4. Apply the selected CI methods. Generate phase grids, company strategy assessments, and analyses of potential gaps in the competitive landscape. Methods developed with CI practitioners organize the evidence around the question; the team assesses what the patterns mean.
5. Prepare and review the brief. Generate an executive summary from the evidence and analysis, inspect the findings, and revise the agent plan or report as needed. Download a PowerPoint presentation for further editing and presentation preparation.

This is the proposed application of Ferma's workflow to the IBD question. The historical IBD benchmark assesses program records; it does not establish the quality or turnaround of this finished brief.

The practical connection is that the same evidence base supports the landscape, analytical views, and executive summary. A team can examine the basis for a conclusion, pursue a follow-up question, and assess whether the answer supports the decision at hand.

How much analyst time could an AI-assisted IBD landscape require?

The workflow estimates allocate 40 human hours for a conventional IBD landscape and four with Ferma, including human review. Actual effort varies with the scope, data volume, and review requirements.

Both workflows are assigned the same finish line: a reconciled landscape with supported statuses and stages, source links, unresolved cases identified, and a reviewed executive brief. The conventional workflow uses the subscription and the team's normal research tools. The Ferma workflow uses the agents plus human setup, verification, interpretation, and final review.

The estimated human effort by stage is:

Using those estimates, the difference is 36 hours of staff capacity per initial landscape, a 90% estimated reduction in human effort.

The largest estimated reductions sit in collecting, reconciling, and verifying evidence. The estimates also allow for reviewing the analysis, correcting the output, and preparing it for use. Actual work must continue if material errors remain after the budgeted review.

Agent runtime is separate from human effort. With an assumed 35-minute execution budget, a schedule in which machine execution and all four hours of human work happen sequentially would take 4 hours 35 minutes, excluding queues, handoffs, and other delays. This is a modeled schedule, not a measured turnaround.

How do you calculate the cost of AI competitive intelligence?

Calculate the human hours required for an accepted deliverable, multiply them by the relevant loaded labor rates, and add attributable software, subscription, and run costs. Count review and corrections in both workflows.

Assume a blended, fully loaded analyst and reviewer cost of US$175 per hour. This is a scenario input, applied equally to both workflows.

The difference is $6,300 in estimated staff capacity value per initial landscape. That capacity could be used to investigate the newly surfaced competitors, test assumptions with primary research, or answer the next strategic question.

The broader commercial question is the combined cost of data access and synthesis. Include software, subscriptions, direct run costs, and any purchased services not already counted as labor. Allocate those costs to the same workload and period for each workflow, without double-counting services included in a subscription. Those prices are not supplied here.

Estimated full cost difference = $6,300 + conventional non-labor costs − Ferma non-labor costs.

Released staff capacity and reduced spending are different outcomes. A cash-saving claim requires evidence of spending actually avoided or removed. The calculation values available time and makes the remaining cost inputs explicit.

The estimate changes if the Ferma workflow needs more human review and correction. Holding the conventional budget at 40 hours gives the following scenarios:

These are three scenarios, not measured ranges. They show how the economics change with a different review and correction burden.

What does it cost to keep an IBD pipeline landscape current?

Once a landscape exists, the next question is often what has changed. A new readout, development decision, or competitor milestone can alter the assessment. Ferma's platform includes monitoring and alert options for tracked landscapes, using the same evidence base to support follow-up analysis.

An IBD refresh still requires reviewing new evidence, updating affected programs, assessing the implications, and revising the brief. Its cost depends on the volume and significance of the changes and the human review required. In the scenario below, both workflows can reuse their prior landscape.

For a refresh within the same scope, the estimated human effort is eight hours conventionally and one hour with Ferma, including review. That gives an estimated difference of seven hours, an 87.5% effort reduction, and $1,225 of staff capacity value at the same assumed hourly rate. A major scope change or extensive new evidence may require a larger budget.

An annual scenario shows how initial builds and refreshes contribute differently. Assume a team completes 24 landscapes of comparable workload and 96 limited refreshes over a year.

This is an annual capacity scenario before non-labor costs. The workload and effort assumptions describe a hypothetical team; actual volume, case complexity, adoption, and review requirements determine what a deployed team can realize.

What should CI teams take away?

Three findings carry the weight of this comparison.

The starting record decides more than the totals suggest. A difference of 12 between 194 and 206 returned programs concealed 86 programs found only by Ferma, three of them in Phase 3, and 63 returned entries whose latest retrievable evidence showed development had stopped. A landscape inherits those differences before any analysis begins, and the totals alone would never surface them.

Data quality and synthesis are one workflow, not two purchases. The evidence base that produced the program set is the same one that supports reconciliation, the analytical views, and the reviewed brief, with source links available for inspection at each step. The links make verification possible; they do not establish factual accuracy on their own. Correcting the starting record and producing the deliverable stop being separate jobs.

The economics follow from the workflow. The expert-reviewed estimates put human effort at 40 hours conventionally and four with Ferma per initial landscape, worth an estimated $6,300 in staff capacity at the assumed rate. Those are estimates with stated assumptions, not measured deployment results. A team can replace them with observations from its own scope by recording the corrections and effort its own review requires.

None of this requires taking a vendor's word for it. The comparison method here works on any landscape question your team already owns: run both workflows, examine the evidence behind the differences, and count the effort honestly on both sides.

Run this comparison yourself. Everything in this analysis can be repeated on a landscape question your team already owns. Choose one, compare what your current database returns with what Ferma returns, and judge the differences against the primary evidence. Then count the effort it takes each workflow to reach a brief your team can use. When you are ready to run it, reach out to us to get access to Ferma for your team.

Frequently Asked Questions