Most teams do account research. Very few of them keep it.
A rep spends forty minutes on an account before a call. They read the last two earnings summaries, find the new VP who started in March, notice the job postings that say the data team is being rebuilt, and write a good opener. The call goes well. Six weeks later a different rep picks up the same account for a different product line, opens a blank tab, and does the forty minutes again. Nothing from the first pass survived, because it lived in one person's notes and one email draft.
That is the real account research problem. Not that reps skip research. That the research has no shelf life, no shape, and no way to be compared against the account next to it.
Salesforce's 2026 State of Sales research, fielded across August and September 2025, puts the average seller at 40% of their time actually selling (Salesforce). The same body of research reports reps spending 60% of their time on non-selling tasks (Salesforce State of Sales statistics). Research is a real share of that, and unlike data entry it is work you actually want done. The waste is not the research. The waste is doing it twice.
The three ways account research decays
Before designing a process, name what breaks. In practice it is always one of three things, and they compound.
It is not reusable
The output of a research pass is usually an email. Sometimes a CRM note, sometimes a line in a spreadsheet, most often a paragraph in a sequence that gets sent and then exists nowhere. The next person to touch the account inherits nothing. Any process that ends in a message rather than a record will decay this way, no matter how good the individual passes are.
The answers have no source
A rep writes "they are expanding into Germany" in the CRM. Four weeks later a manager asks how confident we are. Nobody knows whether that came from a press release, a LinkedIn post, a job listing, or a guess. An unsourced answer cannot be checked and cannot be refreshed, so the next person has no reason to rely on it and researches it again. Sourcing is not bureaucracy. It is the thing that makes the answer reusable at all.
Every rep uses a different format
Two accounts researched by two people produce two documents that cannot be laid side by side. This is the failure that hurts managers rather than reps. You cannot prioritise a territory when every account is described in a different vocabulary, so prioritisation falls back to revenue and headcount, which is to say it falls back to whatever is already in a database field.
The five questions a research pass has to answer
A process is only repeatable if the output is fixed. Before deciding where to look, decide what you are looking for. Five questions cover the ground for most B2B teams:
- Do they have the problem we solve? Not "are they a good fit demographically". Is there observable evidence of the specific problem this product addresses.
- Is there a reason to act now? A funding round, a reorg, a new executive, a regulatory deadline, a competitor displacement, a hiring pattern. This is the buying signal, and it is what separates a good account from a good account this quarter.
- Who owns the problem? A named role, ideally a named person, with a reason to believe they own it rather than an org chart guess.
- What do they already use? Incumbents, adjacent tools, and whether the thing you displace is loved or tolerated.
- What would make this a bad use of time? The disqualifier. Teams almost never write this down and it is the highest-value field in the record, because it stops the account from being reworked every quarter by someone who does not know it was already ruled out.
Every answer gets a source link. An answer with no link is a hypothesis, and should be recorded as one.
A repeatable account research process
The process below assumes the five questions above are fixed and that the output is a record, not a message.
| Step | What you do | What it produces |
|---|---|---|
| 1. Define per product | For each product or solution you sell, write the problems it solves, what a good-fit account looks like, and the signals that indicate the problem is live. | A definition to research against, rather than a generic firmographic profile. |
| 2. Fix the source list | Decide which sources count and in what order: company site, filings, job boards, executive changes, product and pricing pages, news, review sites. | Coverage that is the same on Monday and Thursday, and across reps. |
| 3. Answer the five questions | Work the source list until each question has an answer or an explicit "no evidence found". | A comparable record per account. |
| 4. Attach a source to each answer | Link the page the answer came from. No link means the answer is a hypothesis. | Answers a manager can check without redoing the work. |
| 5. Score fit against the product definition | Compare the record to the step-one definition and score it, using the same scale for every account. | A territory you can rank rather than a pile you can read. |
| 6. Write the opener last | Draft outreach from the record, not from memory of reading the record. | Messaging traceable to evidence, and reusable when the account is picked up again. |
Notice that outreach is step six. Most teams start there and treat research as preparation for a message. Inverting that is most of the work: the record is the deliverable, and the message is a by-product of it.
Step one is also the step teams skip, and skipping it is why generic research feels useless. "Are they a good prospect" has no answer in the abstract. "Do they show evidence of the problem our compliance product solves" has one. If you sell several product lines, you need one definition per line, not one for the company. A single blended profile is how a multi-product company ends up researching every account against the average of everything it sells, which describes none of them.
Making the output comparable across a territory
Comparability is the difference between a research process and a research habit. Three rules get you most of the way.
Same fields, every account, including the empty ones. "No evidence found" is information. It tells the next person the source list was worked and came back empty, which is the difference between an unresearched account and a researched account with nothing in it. Blank fields cannot make that distinction, so they get re-researched forever.
One scale, defined in advance. If fit is scored one to five, write what a four means before anyone scores anything. Otherwise a four means "I liked this account" and the ranking is a popularity ordering.
Date every record. Research has a half-life, and different fields decay at different speeds. A funding round is durable. A named champion is not. A dated record lets you refresh the volatile fields without redoing the durable ones, which is where most of the compounding comes from.
This is also the point where intent data becomes useful rather than noisy. A third-party signal on its own tells you an account did something. A signal inside a comparable record tells you an account did something that matters given what you sell to them.
What to automate, and what to keep manual
Research effort is not fungible. Some of it is retrieval and some of it is judgement, and only one of those should be done by a person at scale.
- Automate retrieval. Finding the filings, the postings, the executive changes, the pricing page, the recent announcements. This is the bulk of the forty minutes and none of it is skilled work.
- Automate the first pass at the five questions, as long as every answer carries the source it came from. An answer you cannot trace is worse than no answer, because it looks like progress.
- Keep the disqualifier judgement with a person, at least early on. Ruling an account out is a decision with a cost, and it is the field most likely to be wrong when it is inferred rather than decided.
- Keep the relationship read manual. Whether a champion actually has budget is not on any page you can retrieve.
Data quality sets the ceiling here. Salesforce's research reports sales leaders estimating that 19% of their company's data is inaccessible, and 70% of data and analytics leaders saying the most valuable insights are trapped in unstructured data (Salesforce). Most of what a research pass needs is in that unstructured layer: prose on a careers page, a paragraph in an annual report, the way a pricing page changed. Any automation that only reads structured fields is automating the part that was never the bottleneck.
Where PitchSmart fits
PitchSmart is built around step one. You define each product or solution you sell, the problems it solves, what a good-fit account looks like, and the signals that matter. Every account on your list is then researched against that definition instead of a generic profile, and each answer links back to the source it came from. Accounts are scored for fit against the product you picked, so a territory can be ranked rather than read.
Whitespace and account-mapping tools stop at the empty cell: this account does not own that product. That is a fact, not a reason. The research pass is what turns it into one.
How to tell the process is working
Not by research volume. Four checks that mean something:
- Second-touch cost. When a second rep picks up an account for a different product, how long until they can act. If it is still forty minutes, the record is not doing its job.
- Sourced answer rate. What share of fields carry a link. This one number predicts whether anyone downstream trusts the record.
- Disqualification rate. If almost nothing is ever ruled out, the process is producing descriptions rather than decisions.
- Rank stability. Ask two managers to rank the same twenty accounts from the records. If they disagree wildly, the scale is not defined well enough yet.
None of these require new tooling to measure, and all of them are visible within a quarter.
Once the records exist and are comparable, the work downstream gets easier in ways that are hard to predict in advance. Qualification stops being a conversation about vibes. Account-based prospecting gets a list it can defend. And territory planning stops being an argument about account counts, because for the first time the accounts are described in the same language.
If you want the narrower version of this focused on a single company rather than a territory, our guide to researching companies covers the source list in more depth.
