You have a contact database, a signal feed, LinkedIn Sales Navigator, a notes doc, and a browser with nine tabs open on one company's newsroom. You still cannot finish the sentence "I am calling them because." That is the actual failure in account research, and it is not a shortage of data. It is a shortage of reasons.
Every category of account research tool was built to answer a different question. Databases answer who exists. Signal feeds answer who moved. Intent platforms answer who is reading something adjacent to your category. None of those is a reason to call, because a reason has to connect something that changed at the account to something you sell. Making that connection is work, and almost every tool on the market stops one step before it.
What follows is every category of account research tool, the question each one genuinely answers, and the limit each one hits. The manual baseline is in the list, because most teams are still there and deserve to know what they are comparing against.
Start with the manual baseline
Before any tool, there is a rep with a browser. The manual pass on a single account looks like this: the company's own newsroom and careers page, the last two earnings calls or funding announcements, the profiles of the two or three people who would own the problem you solve, a check for a new executive in the function you sell into, and a look at what the account already buys from you if it is a customer.
Done properly, that takes 20 to 40 minutes per account, and it works. It is also why nobody does it at list scale. Salesforce's State of Sales research, a survey of 7,775 sales professionals across 38 countries, found reps spend 28% of their time actually selling, with the remainder going to deal management, data entry and research. Thirty accounts researched by hand is most of a week, and it produces no meetings by itself.
Keep the manual pass in view as a quality bar rather than a workflow. It is the only method on this page that reliably ends with a sentence a rep can say out loud. Our account research process walks through the steps, and the pre-call research checklist is the shorter version for the ten minutes before a booked call. The rest of this page is about which tools preserve that output and which quietly replace it with a record.
The categories, and the honest limit of each
Seven categories cover almost every product sold as an account research tool. The middle column is the question the category answers well. The right column is where it stops, which is the part the comparison pages leave out.
| Category | The question it answers | Where it stops |
|---|---|---|
| Manual research | What changed here, and does it touch what I sell | Does not scale past a handful of accounts a week |
| Contact and company databases | Who exists at this account, and how do I reach them | A record is not a reason, and records go stale |
| Sales intelligence and signal feeds | What moved at this account recently | The event arrives with no link to your product |
| Intent data platforms | Which accounts are researching your category | Anonymous, probabilistic, easy to misread |
| Enrichment and workflow builders | Any question you can define, across a whole list | You have to define the question and maintain it |
| Account planning suites | What is our strategy for this one named account | Built for a handful of accounts, kept up by hand |
| Research agents that write summaries | What does this company do, in a paragraph | A summary is not a decision, and sources are optional |
Contact and company databases
This is the largest category and the one most teams buy first. It answers a real question: who works here, what is their title, what is the email pattern, how big is the company. The limit is that a record describes a state, not a change, and states go out of date on their own.
The decay is measurable. The US Bureau of Labor Statistics employee tenure release puts median tenure with a current employer at 3.9 years as of January 2024, the lowest since 2002, and at 3.5 years in the private sector. For workers aged 25 to 34, the group most of your buying committee's analysts sit in, it is 2.7 years. A list you pulled last quarter has people in it who have already moved. Refreshing that list is a maintenance job, covered in our guide to data enrichment tools, and it is not the same job as finding a reason to call.
Sales intelligence and signal feeds
Signal feeds are a real step forward on databases because they deliver change rather than state: a funding round, a new executive, a layoff, an office opening, a technology added. If you already know what you are looking for, these earn their price.
The limit is the last mile. A feed tells you a company hired a VP of Revenue Operations. It does not tell you that the reason this matters is because you sell the forecasting layer that a new RevOps leader always audits in their first quarter. The interpretation step is yours, on every alert, for every product in your portfolio. Teams that skip it end up sending "congratulations on the new role" to strangers, which is why buying signals in sales get a bad name faster than any other research input. Our comparison of B2B sales intelligence tools goes deeper on the category itself.
Intent data platforms
Intent platforms watch content consumption across publisher networks and tell you which accounts are reading about your category. At the account level, and used carefully, this genuinely narrows a list.
Forrester's own list of the ten biggest intent data mistakes names the failure modes more plainly than most vendors do. Three of them matter here: "Treating intent signals as qualifiers," "Using intent signals in a vacuum," and "Ignoring data decay." Read together, they say an intent score is an input to a reason, never the reason itself. The signal is usually anonymous at the person level, probabilistic at the account level, and stale within weeks. If you want the mechanics before you buy, start with what intent data is.
Enrichment and workflow builders
Builders are the most capable category on this page. Point them at a list, define a research question per column, and they will run it across thousands of rows. Anything you can specify, you can get.
The limit is that sentence. You have to specify it, then keep specifying it as your products change, your ICP moves, and the sites you were reading change their layout. In practice a builder turns a research problem into a maintenance problem owned by one person in RevOps, and it stays healthy exactly as long as that person has time for it. That is a fair trade if you have the person. It is not a tool that answers the reason question for you.
Account planning suites
Account planning suites are the honest comparison for anyone who sells to existing customers. They give you a shared document per account: the org chart, the whitespace grid, the relationship map, the plan for the year. For a strategic account with a named team on it, that document is worth keeping.
The limit is arithmetic. These suites are designed for the accounts you can afford to plan by hand, which for most teams means the top twenty or fifty. The other several hundred names on the list get nothing. An account planning template is the right artifact for a top account and the wrong one for a list. Ask any vendor in this category how many of your accounts will actually have a current plan in six months, and the answer will be a number you could have counted on a whiteboard.
Research agents that write summaries
The newest category writes you a paragraph or a page about a company. It is fast, it reads well, and it is the easiest category to be fooled by, because fluent prose about an account feels like knowledge of the account.
Two limits. First, a summary is not a decision: knowing what a company does is not the same as knowing whether to call it this quarter, or which of your five products to raise. Second, if the output does not link to the page each fact came from, you cannot check it, which means you cannot say it on a call without risk. Treat any research output with no source attached as a draft, not a finding.
The reason test
Before you buy anything in the table above, run this test by hand on ten accounts from your own list. It takes an hour and it will tell you more than any vendor comparison. A name is worth calling when you can answer all four of these:
- What changed at the account. Not what is true about it. Something that moved.
- Where you saw it. A page you can open, not a score and not a feeling.
- When it happened. A date. A leadership change from fourteen months ago is history, not a reason.
- Which of your products it touches. The change has to point at something you sell, or it is trivia.
Examples of what fails. "They are in our ICP and they have 400 employees" fails on the first question: nothing changed. "Their intent score is 88" fails on the second: there is no page to open. "They hired a new CRO" fails on the third until you check the date. "They opened a Dublin office" fails on the fourth unless you can say why a Dublin office makes your product relevant this quarter.
What passing looks like: "They posted three requisitions for revenue operations analysts in the last six weeks, on their own careers page, and the second one names the reporting tool we replace." That is a sentence a rep can open a call with, and every part of it is checkable.
This is the test we built PitchSmart to run. It takes the list you already have, researches every account against what you sell, and returns a plan per lead: who to contact, what to pitch, why now, and what to say, with a source on every claim. The test does not change depending on who runs it. Giving the method away costs us nothing, because the method is labour and the product is the labour already done.
Why the stack grows and the research does not
The same Salesforce research found the average sales organisation running about ten tools, with 94% planning to consolidate. That number is usually read as a procurement problem. It is better read as a symptom: each purchase answered a new question, and none of them answered the reason question, so the next tool always looked necessary.
The pattern is consistent. A database gets bought because the list is incomplete. A signal feed gets bought because the list is not prioritised. An intent platform gets bought because the signal feed is noisy. A builder gets bought because the intent platform needs context added. At each step the stack grows and the rep's opening sentence stays the same generic paragraph, because nothing in the chain ever committed to a decision about a specific account.
Breaking the pattern means changing what you buy for rather than buying more. The output to hold every vendor to is a decision per account with a source attached. Everything else is an ingredient. Our walkthrough of how to research companies is the manual version of that output, and it is a fair specification to hand a vendor.
How to choose, in the order that matters
Six questions, in this order, for anything on the shortlist. The order matters because the first two disqualify faster than the rest.
- Show me the output for one of my accounts. Not the demo account. A real name off your list, chosen by you during the call.
- Can I open the source of every claim? If a fact has no link, treat it as unverified, whatever produced it.
- Does the output name which of my products it is about? A portfolio seller needs the per-product answer, not one score for the account.
- Does it ever tell me not to call? Anything that says yes to every account is a list, not research. A useful answer includes hold and skip.
- How old is each fact? Dates on individual facts, not a date on the report.
- What happens at account 400? Ask for the cost, the runtime and the failure rate at your real list size, not at ten accounts.
Most tools on the market pass two or three of these. That is fine, as long as you know which ones, and as long as you stop expecting the ones it fails to be solved by adding another tool beside it. Your list is not the problem. The missing reason is, and the only research worth paying for is research that ends in one.
