Medical insights management, explained.

A question from a Midwest clinic, and the long road to someone who can answer it

One age question in a Veltrimab field note, in five steps

1. The doctor says itAn MSL hears that published results do not report response by age. The conversation is a scientific exchange.
2. The note is loggedThe MSL writes down what was said, in the doctor's words where possible, the same day.
3. Notes become a themeSomeone codes and groups notes from every channel, and sees that the question is rising.
4. A person decidesThe insights lead promotes the theme, keeps watching it, merges it, or rules it out, with a reason.
5. An owner actsA pooled analysis, an updated standard response, a slide module, a survey question. The MSL sees that it happened.

Step 3 is the subject of this guide. The people, products and figures in this example are fictional.

Unused data, counted notes, a channel leaders miss and no legal definition: four facts about field insights

MSL, scientific exchange, standard response and key insight

Fourteen terms are defined in the one-page summary.

What separates a medical insight from a field note, a theme and a key insight

In shortA note records what one person heard. An insight is what the organization can now say after analysis and validation, and it could change a plan. A key insight is one that a named person has promoted and sent to leadership. Teams that blur these words end up counting notes and calling the count understanding.

One working definition comes from a 2023 survey by Reuters Health and Within3, which called insights "relevant information which, after analysis and validation, can drive planning, actions, or decisions".1 That sentence has three tests inside it: where the information came from, what was done to it, and what it could change. In the same survey, 41% said their team was aligned on the difference between data and insights, 47% said somewhat, and 11% said there was no clear understanding.1

Four tests a real insight passes before it reaches a medical director

• It has an origin you can check. A date, a channel, a segment, and close to the doctor's own words.
• Somebody read it against other notes. Is this one doctor, or thirty in three countries? Is one colleague writing most of them?
• It could change something. Name the plan, study, document or conversation that would differ. If nothing would, keep it as a note.
• It has a home. A named role owns the next step, or a named person decided there is none.

A note that confirms what the team already believed still counts: a theme that holds steady for four quarters tells leadership something. The Medical Affairs Professional Society (MAPS) and Within3 describe the same two habits in their 2023 white paper. Good teams listen narrowly, for priorities set in advance, and broadly, by asking "Is there anything we're missing?"5 Those priorities are often written as key insight questions: the specific questions whose answers would change a plan.

MSL notes, inquiries, advisory boards, congresses, surveys: the bias each channel brings

In shortField notes are the largest channel and the most familiar, but MSLs visit doctors they already know. Medical information hears from doctors the field never meets. Advisory boards and surveys answer what you asked. Reading all five together shows more, but only if they share one coding scheme.

Teams know the channels sit apart. In a small MAPS member survey from early 2020, insights from different sources were not integrated in 47% of respondents' organizations.6 In the Reuters Health and Within3 survey, 76% said the way they store and share data has created silos, and 81% said their decisions are only sometimes or never informed by all relevant stakeholders.1 Those who said so most often wanted more from patients, specialists and payers.

The survey was run with Within3, which sells insights software, so read its framing with that in mind. Its 243 respondents were pharmaceutical, biotech and medical affairs professionals; 70% were directors or more senior.1

Scientific exchange and unapproved-use questions: the texts in force and in draft, November 2026

In shortNo regulation is written for medical insights. The work is shaped by rules on what a company's statements can prove, by FDA's draft advice on unsolicited questions, by industry codes and by privacy law. FDA's main text on unsolicited requests is still a 2011 draft, and its January 2025 guidance still carries a "not for current implementation" line.

Two terms run through this section. An off-label, or unapproved, use is one the drug's approved label does not cover, such as another condition or dose. A request is unsolicited when the doctor asks unprompted.

How a doctor's unanswered question leaves a field note

An MSL often hears a question that cannot be answered on the spot. FDA's 2011 draft recommends that questions about off-label uses go to the firm's medical or scientific representative or department, with answers limited to the question asked and documented, and with no input from sales and marketing.8

A company can add a clock of its own for getting the question to medical information. The fictional company in this guide uses one business day, and we found no public benchmark for that clock. Medical information sends the written answer, often from a standard response.

Field notes can carry other duties under a company's own written procedures. Those procedures run separately, and this guide does not cover them.

The firewall between medical and brand teams, a doctor's identity, and groups too small to show

In shortInsights flow out of medical affairs, and three limits apply. A theme about an unapproved use stays with medical. A doctor's identity stays in the source system. A count small enough to point at one person folds into "Other". Each limit needs a named approver and a record of what was removed.

Many companies call the first limit a firewall. We found no statute or FDA rule that uses the word, and a 2022 note by the law firm Sidley says "No law or regulation expressly requires" this independence; companies adopt it to reduce risk under the Food, Drug, and Cosmetic Act and the Anti-Kickback Statute.18 So it is company policy, resting on the intended-use rule, on FDA's draft advice that sales and marketing have "no input" on responses to off-label questions, and on industry codes.9,8,14

The history is the reason it exists. In 2004 Warner-Lambert agreed to plead guilty and pay more than $430 million over the marketing of Neurontin. The Justice Department said the company used "Medical Liaisons" who represented themselves, "often falsely", as scientific experts.7 That was a promotion case with no connection to insight reporting. It shows what a blurred line looks like when a regulator reads it years later.

Your policy may differ. When a doctor asks for a written answer, the reply reaches them through the route your procedure sets, with the contact details kept in the source system.

Why a physician's name is a privacy question under GDPR, and HIPAA is usually a different one

Under GDPR, the EU's data protection law, personal data is any information about an identifiable person, and that includes a doctor named in a note. Replacing the name with a code is pseudonymization, and pseudonymized data that could be linked back still counts as personal data under recital 26.16

The European Data Protection Board adopted draft guidelines on pseudonymisation on January 16, 2025; we found no later adopted version. Its 2026 guidelines on anonymisation are a draft too.19,20 The Council's October 2026 mandate on the Digital Omnibus adds text on pseudonymized data for a party unable to identify the person. It is not law, and Parliament's position and talks are still to come.17

HIPAA, the US health privacy law, protects patients' health information held by covered entities such as providers and insurers, and a drug maker is usually not one. Its de-identification rule has two routes, a documented expert determination or the safe harbor list.21 Patient details do not make an insight better, so the practical rule is the same in both worlds.

Hiding a small group of doctors: an unstable count, an identifiable person and where to draw each line

A small count in a report table, called a small cell, has two problems. It is unstable, and it can point at someone. The worked math follows in section 07.

The insight record and coding scheme that keep this quarter comparable with the last

In shortEight fields, the same for every channel, make an insight citable. A coding scheme with a version number makes counts comparable. A percentage of agreement between coders flatters the result, so test with a chance-corrected measure, kappa, and test each code group separately.

Topic, sentiment and segment: three code groups and how a scheme decays

Most coding schemes, often called taxonomies, code a topic, a sentiment and a doctor segment. Topic is where the arguments happen. A broad code like "unmet need" absorbs anything, so every code needs a boundary line and an example of something close to it that is left out. Change the scheme and the trend breaks: recode the earlier quarter, or draw a break in the line and say why.

How a 94% agreement figure can mean strong or weak coding

Jacob Cohen objected in 1960 to plain percent agreement because it ignores the agreement two coders reach by chance.23 His kappa takes that out: kappa = (observed agreement − chance agreement) ÷ (1 − chance agreement). Work it through with two invented schemes and the same 94% observed agreement on 200 insights:

Illustrative chance figures. One review, by McHugh, says that any kappa below 0.60 "indicates inadequate agreement", that about half the data may then be incorrect, and that 0.60 to 0.79 is moderate.23

Those comments were about clinical care at one hospital, and the figure followed careful prompt tuning on a sample. They were not field medical notes. We found no published agreement figure for coding MSL notes.

The arithmetic of a rising question: windows, baselines, denominators and small counts

In shortThis section counts scientific and practice questions that doctors ask. "Doubled" is a claim with four hidden choices: the window, the baseline, the denominator and the size. State all four. Check how many colleagues and countries the rise rests on. A doubling from 48 to 99 is firmer than a doubling from 12 to 25, and both can still be nothing more than a busier field team.

Two true growth figures for the over-65 question

In the example, the over-65 theme has two growth numbers. Four weeks against the four before: (99 − 48) ÷ 48 = +106%, counted in mentions. This quarter against last: (174 − 67) ÷ 67 = +160%, counted in insights. The example's map counts mentions over four weeks, and its quarterly report counts insights. Both growth figures are true, and a report that mixes them cannot be checked.

How firm a doubling is at 99 mentions and at 25

Counts move around by chance. A rough check treats each count as a random number and asks how wide the likely range of the ratio is. It is the ratio, times or divided by e raised to 1.96 × √(1/a + 1/b), where a and b are the two counts.

Our arithmetic. Field notes are no designed sample, so treat this as a rough guide for reading a chart. It does not prove a change in the world.

Mentions per reporting colleague: dividing out a bigger field team

A raw count rises when the team grows, when a country joins or when someone runs a logging campaign. Suppose 20 colleagues logged 48 mentions in the first window (2.4 each) and 40 colleagues logged 99 in the second (2.5 each). The count doubled and the rate per colleague barely moved. Three denominators are worth knowing, and a chart should name the one it uses:

• Share of all insights in the window shows what the field is spending attention on.
• Mentions per reporting colleague suits a team that is changing size.
• Mentions per doctor interaction is the most faithful, if you can get the number from the CRM.

Five checks on a rise before it reaches the Monday meeting

• Spread. How many colleagues and countries? In the example, the alert rests on nine notes from three countries, and the quarter's theme on 31 MSLs in nine.
• Source. Did a survey or a board ask about the topic directly? Recording the channel shows that.
• Calendar. A window holding the main congress will look busy. A baseline should match the window's length, because a long average is smoother and flatters the short window.
• Scheme. Did the codes change between the two periods?
• The falls. Look at what dropped as well. A question that halves tells you that an answer reached people, or that nobody asks any more.

How many notes it takes to know a theme exists, and how many to understand it

Qualitative research draws a useful line. In one study of 25 in-depth interviews, the range of issues appeared by the ninth interview, called code saturation. A richly textured understanding took 16 to 24, called meaning saturation.25 Interviews are a different method from field notes, so the numbers do not transfer. The lesson does: nine notes can tell you a question exists. They rarely tell you why doctors hold the view, or what would change it.

From a rising question to an evidence gap: what data already on file can answer

In shortA theme that says "doctors want more data" is too vague to act on. A useful one names the decision the evidence would change. Then look for what the company already holds, because a pooled analysis of completed trials costs far less than a new study, and many questions need only a clearer document.

Our sample question has a real-world basis. A systematic review of 222 phase 3 trials in inflammatory bowel disease found that age was an exclusion criterion in 129 of them, or 58%. In the 32 trials that reported the figure, 763 of 14,124 patients, or 5.4%, were older than 65.26 Comorbidities were an exclusion in 76% of trials, mainly kidney, liver and heart disease.26 So a doctor asking how a drug works in her 75-year-old patient may well get no direct answer.

FDA's labeling rule shows the question is an old one. The "Geriatric use" part of a US label covers people 65 and over. When studies lacked enough older subjects to tell whether they respond differently, the rule requires a statement saying so.27 A drug can be approved for adults generally with that sentence in its label, and doctors with many older patients will still ask.

Four routes out of a theme: document, pooled analysis, real-world study, new trial

An integrated evidence plan brings together evidence work across functions and countries over a product's life. A 2026 account of these plans argues that medical affairs is well placed to lead such planning, in partnership with market access and clinical development.28 A field theme enters that plan as a stated need.

FDA's real-world evidence framework matters here: evidence from traditional clinical trials "will not be considered" real-world evidence, while routine-care data can support it, and routine care is where older and frailer patients are.29 A pooled analysis combines patients from finished trials and reads them again. It is a subgroup finding, with the limits of one: the trials were not designed to answer it, so state it as descriptive unless your statisticians say otherwise.

Many field "needs" turn out to be awareness needs: the answer exists and the doctor has not seen it. Classifying a theme before it reaches a prioritization meeting keeps that meeting short, and it protects the credibility of the gaps that do need money.

Where insight programs fail, from counting notes to reports that nobody acts on

In shortCompanies measure what is easy: engagements and notes submitted. The people doing the work want quality measured. Most of the failures that follow are failures of ownership and feedback, and they repeat: the same theme returns each quarter, and MSLs stop logging what nobody reads.

The measurement problem runs deep. In a follow-up paper by the same lead authors, 76% named difficulty defining and measuring the quality of MSL activities as a key challenge, 61% named reliance on quantitative standards that do not reflect MSLs' value, and only 40% said their organization measures MSL impact.30 That paper reports the same number of respondents, 1,023, from a convenience sample recruited through professional networks.

Eight common errors in medical insights work, and why each one happens

• A target on notes submitted. Four short notes then score higher than one good one. Rural territories lose to teaching hospital clusters. The trend line starts to show how the field responds to being measured.
• Twelve ways of coding one idea. Each MSL's habits, and each country's vocabulary, split a single question into several small ones. MAPS' webinar names inconsistent capture first among its pain points.4
• The vivid quote as the finding. One memorable remark from a famous name outweighs thirty quiet ones. A theme should carry the quote as one citation among many.
• Reporting by volume. The largest theme leads every quarter, so the report describes the same stable subject for years, while a smaller question rises unseen.
• A statement nobody can trace. Someone summarized the notes by hand, and the link from a sentence back to the exchange was lost. A medical director who cannot check a claim discounts it.
• Upward only. Insights go up to leadership and never sideways or back down. The roundtables behind the MAPS white paper describe siloed systems that need multi-year IT projects to untangle.5
• No ending. The same theme appears in four reports, each time as "emerging". Four endings cover nearly every case: promote, keep watching on a stated condition, merge, or rule it out.
• No feedback. MAPS' webinar speakers list, as a pain point, MSLs who feel their insights are "disappearing into a black hole".4 Other roundtable participants report that MSLs see insights as a distraction from their real job.5 When nothing visibly happens, the notes get shorter, then stop.

What weak insights cost medical affairs, and how far insights software has spread

In shortSurveys point to launches at risk, wasted budget and data nobody uses. No one publishes a dollar cost of a late insight, or a reliable market size for this software, and we say so. Tools are common and rarely automated: 61% use technology to collect data, and 8% call it highly automated.

Those are opinions in a survey run with a software vendor, and they measure fear better than loss. No one publishes these figures: insights per MSL, the cost of an insights program, how late a typical theme reaches leadership, how many MSLs work worldwide, and the size of the insights software market. Section 18 shows how to measure your own.

Report rhythm is better known, though from an old and very small sample. In the second of two MAPS member surveys from 2020 (40 responses), an insight review process appeared established in 74% of organizations, usually involving medical directors. Reports went out monthly in 44% and quarterly in 38%. We read a summary of this MAPS page, not its full text.32

Some companies are redrawing field medical roles, as in a 2025 account of a "strategic scientific advisor" at one company, built around early engagement and evidence generation.33

What AI changes in reading field notes, and what regulators say about tools like this

In shortLanguage models can code free text at a useful level once tuned, and can group inquiries by meaning. The evidence is thin and little of it comes from this field. No agreement score tells you whether a theme matters, so that call stays with a person. Most AI rules for drug development do not reach this work directly, and your own policy matters more.

Policies lag behind those expectations. In that survey only about a third of organizations had policies on AI use by MSLs.34 In a small 2026 survey by a software company, 49% of 79 respondents said they review every AI output by hand; the survey says it was not designed to be statistically representative.35

What the studies of AI coding show, and where they run out

• Tuned coding can reach useful agreement. On 778 clinician comments from one hospital, a general model reached kappa of 0.76 and 0.78 after careful prompting and calibration on a 20% sample.24
• Theme finding is uneven. On 40 health text messages, two models matched 5 of 7 themes that human analysts found on their own, and 6 or 7 of 12 when given a framework to code against. On the themes both found, coding agreement with the human coders ran from 36% to 47%. The sample was small and far from field medicine.36
• Clustering inquiries has been shown. A University of Toronto tool grouped medical information inquiries by meaning, a demonstration of the method that reports no accuracy figure we could quote.37
• No test on this work. We found no peer-reviewed test of AI theme detection on MSL notes against an expert reference.

Models also drift in ways a test should look for: broad codes collect anything, rare codes get few examples, sentiment is harder across languages, and a model that agrees with itself on two runs has shown only consistency. A confident wrong coder would score well on that.

Which AI rules reach an insights tool, as of November 2026

The practical stance follows: treat each output as a claim that a person checks, keep every sentence linked to its source, record who decided what, and write down what the tool may and may not do. A 2026 review of digital medical affairs puts cross-functional collaboration and governance for responsible AI among the conditions for success.40

Who turns a field note into a decision today, and where an insights lead's quarter goes

In shortSix roles touch an insight. One person usually does most of the reading. A MAPS webinar lists five parts of a working program: collect, analyze, communicate, act, and give feedback at every step. Reading and coding take the largest share of a by-hand quarter on our illustration, and the report arrives weeks after the quarter closes.

From MSL to medical director: the six people who carry a theme

Where twenty-five working days go in a by-hand insights quarter

These days are spread over the weeks after a quarter closes, in between other work, so the elapsed time runs longer. In the fictional example the team signed its last three reports 41, 38 and 35 days after quarter end.

Spreadsheets, a CRM module, an outside agency or an insights platform

The over-65 question in Botminds, from Monday's theme map to a named owner

In shortMonday's scan shows 48 mentions becoming 99 in four weeks, sourced to nine notes. Priya Ranvelle, the insights lead, reads them, opens the theme, promotes it to a key insight, and four actions go to their owners. The badge on the map now names who owns the next step.

1

Twelve themes on one map, sized by mentions and colored by change

Priya's home is a map. Each bubble is a theme, sized by mentions from September 7 to October 4 and colored by the change on the four weeks before. A dashed outline marks a theme that stays medical-only. A small badge on each bubble gives the initials of the person who owns the next step, or "!" when it waits for her. The theme "Patients over 65: dosing and response" has gone from 48 to 99 mentions, +106%.

2

Replaying Monday's scan, last month against this month

Priya can replay the scan. The bubbles start at last month's size and grow to today's, so she sees which theme moved and which only looks large. Sequencing is still the biggest theme at 153 mentions, and it grew 2%. A report written by volume would lead with sequencing again.

3

The nine MSL notes behind the alert, each with its place and date

The first question in the room is usually "says who?". The panel answers it with nine notes, each with its segment, place, colleague and date. A doctor in Illinois asks for remission rates in over-65s on steroids. A pharmacist in Ohio asks whether the maintenance interval needs adjusting with reduced kidney function. Priya can see at a glance that several colleagues in three countries logged them. Each note opens its source record, and the doctor stays a segment.

4

A theme statement with every sentence cited to its notes

She opens the theme. Four agents built it in 9.6 seconds: 174 insights from five channels, de-identified; coded, with two follow-up requests flagged; grouped into four sub-themes, +160% on Q2; and matched to evidence gap EG-14 and medical plan objective MPO-3. Each numbered citation in the statement opens the insight behind it, with the passage highlighted on the right. A heat table shows where it is heard, month by month.

5

Her decision: key insight, keep watching, merge, or not a theme

An agent, the evidence and plan linker, proposes four actions, each with an owner and date: a pooled response analysis of patients 65 and over from VLT-302, VLT-305 and the extension (Omar Haddad, evidence generation lead, January 30), an MSL slide module (Lena Ortiz, field medical director, November 15), an update to standard response SRL-0412 (Sam Patel, medical information, October 31), and two questions for the Q4 survey (Priya, November 3). The agents suggest a key insight at 91% confidence. The choice is hers, and she can change any owner.

6

Four actions sent, and a note for the leadership team

She promotes it. A dialog lists the four actions and offers an optional note: "Raised in Q2; doubled this quarter. The pooled analysis can answer it from data we already hold." The company holds 212 patients aged 65 and over across two trials and a long-term extension, 9% of 2,356, so this question can be answered without a new study. The analysis is descriptive, and the statisticians will say how far it goes.

7

The badge on the map changes from "!" to SP

Back on the map, the over-65 bubble no longer waits for her. Its badge reads SP: Sam Patel owns the next step, updating SRL-0412 on use in older patients, due October 31. The panel keeps what was already done: EG-14 logged in Q2 with its priority raised to high, and a Hamburg physician's question sent to medical information as request MIR-2026-08812.

A physician's written question, a theme nobody owns, and the map cut by channel

In shortThree screens around the main story. A doctor's question the MSL could not answer becomes a drafted request that waits for a person. A theme with no next step shows a dashed badge. A filter resizes every bubble, so the insights lead can test whether a rise lives in one channel.

The Wisconsin question that waits for a person to confirm it

One insight in the theme, INS-30871, records a question the MSL Emily Castellan could not answer in the meeting: how long to wait before judging response in patients over 75. The physician asked for a written answer. The insight coder flagged it and drafted the request with the passage highlighted. The physician's contact stays in the CRM. Priya confirms before 5:00 pm, and medical information sends the answer. The company's rule is one business day.

Access delays: done, with a dashed badge and nobody on the next step

Not every bubble ends in an action from medical. The access delays theme carries a dashed badge: its appeal summary is done, and coverage questions sit with market access. The panel says so, and offers an Assign button in case that is wrong. A gap like this is what a quarterly read-out tends to hide.

MSL notes alone: how the bubbles resize when one channel is filtered

Filter to MSL notes and the over-65 theme shows 58 mentions against 28, +107%. The rise holds with medical information and the other channels taken out, which is what a careful reader wants to see. Filters by region work the same way.

All twelve themes in a table, by status, region and evidence gap

The Themes screen lists every theme with its count, its 13-week shape, its change on Q2, its top region, its evidence gap, its progress and its status. It can filter to those needing a decision, and it searches the quotes.

The Q3 insight report, every sentence cited, and a separate brand version with each removal listed

In shortThe report writer works from Priya's decisions, and undecided themes appear as "for discussion". The firewall guard builds a brand version and lists four removals. Medical governance approves that version, and the medical director signs with the meaning of her signature.

Priya presses "Draft the report". The report writer takes about a minute and produces the medical version: 2,946 insights, 13% more than Q2, grouped into 12 themes, with the over-65 theme first and a recommendation that the leadership team endorse the pooled response analysis. The draft carries 25 citations. Dr. Maya Chen, the medical director, can click any of them to see the note behind a sentence.

What the brand team may see: four removals with a reason for each

The firewall guard, an agent, builds the brand version from the sharing rules. It removes theme T-11, an unapproved use that stays medical-only. It drops city and practice from quotes. It folds two small region cells into "Other". It leaves out medical information request detail. An item it cannot classify is removed. Nothing reaches the brand team before Dana Okafor, in medical governance, approves it.

Dr. Chen's electronic signature and the meaning she signs

Priya sends the report for sign-off. Dana approves the brand version. Dr. Chen signs with a password and a stated meaning: "I approve this report for the medical leadership team." In the sample data this report closed 7 days after quarter end, against 41, 38 and 35 for the three before. None of that is a customer's result.

The action board, the theme momentum dashboard and the medical rules you set yourself

In shortEvery key insight ends on a board with an owner, through to impact measured. The dashboard ranks themes by change, shows the road from theme to finished action, and tracks time from quarter end to signature. The rules are settings you change, and each change is versioned and recorded.

Six agents read and draft while the insights lead, governance and the medical director decide

In shortEach agent has one job and a stated limit. People decide themes, confirm requests, approve the brand version and sign. Insights come in from five read-only sources, and cited themes, owned actions and a signed report come out. Your CRM and medical information system stay the systems of record.

Who signs what: the insights lead, the owners, governance and the medical director

Controls that support your own compliance and audit work come built in: separate roles for preparing, approving and signing, an audit trail of every agent step and decision, versioned rules, and export of the record. The company behind it is SOC 2 Type II audited and manages information security to ISO 27001. Whether any record falls under Part 11 or similar rules is your quality team's call.

Five insight channels and your evidence register in; cited themes and a signed report out

The CRM, medical information system and registers you keep

Design targets, to be measured in your own run: 2.1 days median from a captured insight to its theme, 97% of insights coded by agents, 94% agreement with the insights lead on a 200-insight sample. These are the solution's goals, and none is a customer result.

What late themes and unread field notes cost you, and how to find your number in a week

In shortNobody publishes the dollar cost of a late insight. The costs you can count are reading hours, evidence work done twice and standard responses updated late. The bigger ones sit outside a spreadsheet: a question left open for a quarter, and MSLs who stop writing. Measure your own in a week, starting with last quarter.

Seven costs of a late medical insight you can size from last quarter's records

MSLs who stop writing, doctors who stop asking: losses outside any spreadsheet

What automating insight work pays back: reading days, MSL admin hours and earlier decisions

In shortCount the days your team spends reading, coding and writing, and the time MSLs spend on insight admin. Then count only the hours that will go to other work. Time to a decision matters more, and it is hard to price, so report it next to the money.

Working the sum for your therapy areas and field team

Score your insights program for automation: definitions, channels, deciders and the firewall

In shortEight questions, each scored 0 to 2, plus one for reflection. Scores of 0 to 6 say explore first, 7 to 11 say run a pilot, and 12 to 16 say you are ready to implement.

Eleven tests for an MSL insights platform, from clicking a citation to seeing the brand view

In shortGood vendors, us included, answer with a screen and a number on your data: a sentence that opens its note, kappa by code group against your own coders, and the view a brand user gets. Warning signs sit beside each answer.

Replaying a quarter you already know: a ten-week insights pilot with a decision at the end

In shortTake one therapy area and one past quarter. Hand-code a sample first. Run the software on the quarter and compare it with your codes, your report and your memory. Then run live for one month. Decide on numbers you set before you began.

Coding agreement, citation checks, planted firewall items: eight pilot measures

Fourteen answers on MSL insights, emerging themes and who may see them

In shortFourteen short answers. Each stands on its own.