There is no shortage of CI report templates in circulation, and most of them produce the same artefact: a long, diligent, unread document. The template is rarely the problem. The problem is that the document was organised around what was collected instead of what the reader must decide.
A CI deliverable is a decision instrument. Before structure, one question: who reads this, and what are they deciding? A brand lead deciding whether to accelerate a filing, a BD team pricing a licence, and a medical team planning congress presence need different documents built from the same record. Write the distribution list before the outline.
The five documents
- The monitoring note — weekly or biweekly, one to two pages. What changed, why it matters, what we are watching next. This is the function's heartbeat and its public reputation.
- The landscape — the maintained map of a competitive space at indication granularity. Lives as a record, not a deck; documents get generated from it when asked.
- The deep dive — a single competitor programme or event, examined because a decision is live. Commissioned, not scheduled.
- The congress readout — what was presented, how it differed from the abstract, what the corridor conversations added, delivered while it is still news. Covered properly in congress intelligence.
- The scenario brief — pre-work for a readout or launch: the plausible outcomes, leading indicators for each, and the planned response. The cheapest insurance the function sells.
Anatomy of the one that matters
The landscape-derived assessment report — the document most people mean by "CI report" — holds up when it is built in this order:
Section by section
- Assessment, first page — what changed, what it means for us, confidence level, in plain declarative sentences. If the reader stops here, the document still worked.
- Implications and recommended actions — specific to your asset and named decisions. This is the section that justifies the function's budget.
- The evidence, by competitor programme — phase, population, line, endpoints, timing, recent changes, each claim carrying its source and date. Organised by programme, never by source type.
- Event calendar — readouts, regulatory dates, congress abstract deadlines for the next two to four quarters, tied to the programmes above.
- Appendix — the completeness lives here, where it reassures without obstructing.
The so-what discipline
Every finding in the body earns its place by surviving one question: what would we do differently if this is true? A competitor's Phase II readout is a fact. "Their monotherapy response rate makes our planned comparator arm look dated, and the protocol team should see this before the amendment window closes" is intelligence. The first without the second is news coverage, and the organisation already subscribes to news.
This is also the honest filter for length. Documents balloon when analysts hedge by including everything; the hedge transfers the filtering work to the reader, who declines it by not reading. Ten pages that have been filtered beat forty that have been compiled, every time, in every organisation.
Cadence and distribution
Cadence follows decisions, not data. The monitoring note is the only deliverable that belongs on a calendar. Landscapes refresh when something material happens — a readout, a deal, an approval — because a quarterly refresh of an unchanged landscape is paid busywork, and a landscape that waits for the quarter to absorb a material event is negligent. Deep dives and scenario briefs are commissioned against live decisions.
Distribution is part of the structure. A report that goes to everyone is calibrated for no one; the practical pattern is a short assessment distributed widely, with the full document one request away. And sourcing hygiene is not optional — every claim dated and attributed, speculation labelled as speculation — because the function's long-term asset is that its documents are trusted. For what feeds the evidence sections, see tools and databases and primary vs secondary.
What to leave out
Screenshots of pipeline databases the reader could query themselves. Methodology essays — a one-line note on sources and confidence does the work. Anything already circulated in last week's note, unless its meaning changed. And the temptation to demonstrate effort: the reader is buying judgement, and padding is where judgement goes to hide.