AfterDesk
EN
Apply

The toolkit

CRM and Pipeline Hygiene

A CRM is only worth what its worst record is worth. This course teaches you to model contacts, companies, deals and activities correctly, move a deal only on evidence, merge duplicates without losing history, and log work so the pipeline reflects reality. You finish able to build a weekly pipeline report, run a clean-up sweep on a stale database, and hold the standard that keeps it clean.

8
lessons
~58
minutes
12
exam questions

Free · No paid tier · No certificate fee

After this course

Everything, and what is in it.

What a CRM Actually Models

~7 min

Four objects, one shape

Almost every CRM, paid or free, models the same four things. A contact is one human being with one email address. A company is one organisation. A deal is one specific opportunity to sell something to that company, carrying an amount, a stage and a close date. An activity is one thing that actually happened: a call, an email, a meeting, a note. Everything else a CRM offers is decoration on those four. Learn them once and you can work in any tool a client hands you, because the buttons change and the model does not. When you open an unfamiliar CRM, your first job is to find where each of the four lives and what the client calls it. Some say opportunities instead of deals, some say accounts instead of companies.

Why the links matter

The four objects are only useful because of how they connect. A contact belongs to a company. A deal belongs to a company and involves one or more contacts. Activities attach to a contact, a deal, or both. Get the links right and one question answers itself: what is happening with this account? Get them wrong and the database lies quietly. A deal with no contacts attached is a deal nobody can follow up. An activity logged against the contact but not the deal disappears from the deal timeline, so the next person sees silence where there was a two-hour meeting. When you create anything at all, your last step is always the same: check what it is attached to.

One record per real thing

The central rule is that one thing in the world gets one record in the database. One person, one contact record, even if they wrote from two addresses. One company, one company record, even if the client's team spells it three ways. One opportunity, one deal, even if it stalls and restarts later. Break that rule and every number downstream is wrong: the forecast double-counts, the report shows more pipeline than exists, and two salespeople call the same buyer in the same week. Most of the mess you will be asked to clean up is this rule broken quietly, hundreds of times, by people in a hurry. Everything else in this course is either protecting that rule or repairing it.

Free tools that do the job

You do not need a paid subscription to work properly. The HubSpot free tier gives you real contacts, companies, deals and a pipeline with stages, and many small clients already use it. If a client has no CRM at all, a well-built Google Sheet does the same job: one tab per object, a unique ID column on each, and the company ID repeated on the contact and deal tabs so the links survive sorting. Salesforce and Pipedrive dominate larger companies and you may meet them on a client's account, but you never need to buy one to learn this. The discipline transfers exactly. A tidy spreadsheet beats an expensive CRM full of duplicates, every time.

Remember

  • A CRM models four things: contact, company, deal, activity.
  • The links carry the meaning. Always check what a new record attaches to.
  • One real thing gets exactly one record. Every number downstream depends on it.
  • HubSpot free tier or a structured Google Sheet both work. The model never changes.

Stages, and the Evidence Rule

~7 min

What a stage really is

A stage is a claim about where the buyer is, not about how hopeful the seller feels. Typical stages run something like new, qualified, meeting held, proposal sent, negotiation, closed won, closed lost. The names vary by client and the names do not matter. What matters is that each stage means one specific, checkable thing has happened in the real world. Qualified does not mean the salesperson likes the look of it. It means someone confirmed the budget, the need, and the person who signs. When a stage means a feeling, the pipeline becomes a mood chart and the forecast built on it is fiction. Your job, whether you own the data or only maintain it, is to keep each stage meaning what it says.

Exit criteria, written down

For each stage there should be a written sentence saying what must be true before a deal leaves it. Proposal sent leaves when the buyer has confirmed receipt and given a date for their answer. Meeting held leaves when notes exist and a next step is booked. If the client has no written criteria, ask for them, and if nobody has ever written them, offer to draft what you observe and have the sales lead confirm it. That single page is the most valuable document in a CRM. It turns arguments about a deal into a check against a list, and it makes your own updates defensible: you moved the deal because the criterion was met and the evidence is logged.

Move on evidence, not optimism

The discipline is simple to state and hard to hold. A deal moves forward only when something happened, and that something is logged where the next person can see it. An email confirming the meeting. A signed order form. A reply agreeing the price. No evidence, no move. This applies backwards too: a deal that has gone quiet does not stay in negotiation because moving it back would look bad. If you are updating a CRM on someone else's instruction and they tell you to advance a deal you can see no evidence for, do it and record plainly what you were told and by whom. You are not the judge of the deal. You are the record of what was decided.

Close date and amount are data

Stage is the headline, but close date and amount decide the forecast. A close date is the date the buyer is expected to sign, not the end of a quarter someone hopes for, and not a date nobody has touched since March. When a close date has passed and the deal is still open, that is not a small formatting problem: it is the most common single reason a forecast is wrong. Flag every one you find. Amounts follow the same rule. Use a number the buyer has seen, or the client's stated standard price for that product, never a figure you invented to fill the cell. If the real amount is unknown, the next lesson explains what to do.

Remember

  • A stage describes the buyer's position, not the seller's confidence.
  • Every stage needs written exit criteria. Draft them if nobody has.
  • No evidence, no stage move. Deals are allowed to move backwards.
  • A close date in the past is the loudest error in any pipeline.

Duplicates and How to Merge

~8 min

Where duplicates come from

Duplicates are not a sign of careless people. They are the normal result of many hands and several doors. A form creates a contact with a personal address, then the same person emails from work. An import runs twice because the first one seemed to fail. Two salespeople create the same company because one typed Acme Ltd and the other Acme Limited. A trade show list arrives as a CSV and lands on top of everything. Understanding the source matters, because the same source will keep producing duplicates after you have finished cleaning. When you deliver a deduplication job, saying where the duplicates came from is often worth more to the client than the merge itself, and it is the part most people leave out.

Find them without guessing

Work from exact keys first, then loosen. Export the contacts, and in Google Sheets build a helper column that lowercases and trims the email, then use COUNTIF on that column to count how many times each value appears. Anything above one is an exact duplicate and needs no judgment. Next, match companies on the website domain rather than the name, because domains are unique and names are not. Then, and only then, look at softer signals: same last name plus same company, same phone number, company names that match once you strip Ltd, Inc, GmbH and punctuation. Each pass leaves a shorter list needing more thought. Never start with the fuzzy pass, or you will spend your night on judgment calls the exact pass would have settled.

Deciding what is really the same

Two records that look alike are not always the same thing. Two people can share a name inside one company. A parent company and its subsidiary can share a website and still be separate accounts with separate budgets. Shared inboxes belong to no single human and must never be merged into a person. Ask one question: if I merge these, does one real thing end up with one record, or do two real things get flattened into one? When you are not sure, do not merge. Put the pair on a review list with both record links and the reason you hesitated, and let the client decide. An unmerged duplicate is untidy. A wrong merge destroys information nobody can rebuild.

Merging without losing history

A merge is not a delete. Done properly, one record survives and everything attached to the other moves onto it: activities, deals, notes, files, form submissions. Before you merge anything in bulk, export the full list of both sides so the original state can be reconstructed, because most merges cannot be undone. Choose the survivor deliberately rather than by default. The record with the richer history is usually the better survivor, even when the other has newer field values, because activities are hard to recreate and a phone number is easy to copy over. Then check what the merge kept: CRMs typically keep the survivor's field values and discard the other's, so any value you wanted from the losing record must be copied first.

When not to merge

Some pairs should stay apart, and knowing them saves you from a bad night. Do not merge a person into a shared inbox. Do not merge two companies because the names match if the domains differ. Do not merge deals: two open deals on one account are often two genuine opportunities, and if they truly are duplicates the right move is usually to close one as a duplicate so the history stays visible, following whatever the client's process says. And never merge in bulk on your first pass through an unfamiliar database. Do a sample of twenty by hand, show them, get confirmation, then scale. The sample costs an hour. Reversing a bad bulk merge can cost a client their year.

Remember

  • Duplicates come from repeat imports, forms and inconsistent naming. Name the source, not just the count.
  • Match on exact keys first: email, then domain. Save fuzzy matching for last.
  • Ask whether merging leaves one real thing with one record. If unsure, do not merge.
  • Export both sides before merging. Most merges cannot be undone.
  • Merge twenty by hand and get them confirmed before touching hundreds.

Fields, and Why Blank Wins

~7 min

Which fields actually matter

Every CRM has a handful of fields the business genuinely runs on. Usually it is deal owner, amount, close date, stage and lead source, plus the domain and country on the company. These feed the reports, so an error in them travels. Before you start any CRM work, ask which fields the client's reports actually use, and treat those as sacred. If nobody can tell you, look at what the last report grouped by. Knowing which five fields matter changes how you work: you spend your attention where a mistake would show up in a board meeting, instead of spreading it evenly across forty fields nobody has read since the system was set up.

Blank is honest, guessed is poison

A blank field says we do not know. A guessed field says we know, and it is wrong. The first can be filled by anyone in five minutes. The second can sit undetected for years and be counted in a forecast. So when a value is not in front of you, leave it empty. Do not put a placeholder date. Do not set the amount to the average deal size to make the pipeline look complete. Do not pick the first option in a list just to get past a required field. If a system forces a value before it will save, that is worth flagging rather than defeating, and if you must save anyway, record exactly which records were forced and why. Completeness is not accuracy.

Picklists beat free text

Any field you will ever count should be a picklist, not a text box. Lead source typed by hand becomes webinar, Webinar, web-inar and web in a single quarter, and the report shows four sources where there was one. Industry, country, deal type and stage all belong in fixed lists. When you are asked to clean a free-text field, the deliverable is usually two things: the cleaned column, and a short list of the values you mapped onto each other so the client can see your decisions and turn the field into a picklist afterwards. If you are building a spreadsheet CRM, use data validation from a named list on a hidden tab. Two minutes of work prevents the mess permanently.

Decide the format once

Half of CRM cleaning is not judgment, it is choosing a format and holding it everywhere. Phone numbers in international format with the country code. Dates in one unambiguous format, remembering that 03/04 means March to one reader and April to another. Company names without the legal suffix unless the client wants it. Websites as bare domains, lowercase, no www and no trailing slash. Names capitalised properly rather than shouted in capitals because that is how the import arrived. Write your chosen formats at the top of your delivery note, or on a short tab called standards. The value of a standard is not that it is the best possible choice. It is that everyone after you can follow it.

Remember

  • Find the five fields the client's reports use, and guard those first.
  • Blank means unknown and is fixable. A guessed value is a lie that spreads.
  • Any field you will count should be a picklist, never free text.
  • Write your formatting standards down so the next person can follow them.

Logging Activity So It Is True

~7 min

What counts as an activity

An activity records that something happened between the seller and the buyer: a call, an email, a meeting, a demo, a quote sent. It has a date, a type, the people involved, and what came of it. Internal chatter is not an activity. Neither is a task someone plans to do. The line matters because activity counts get reported and compared, so if planned calls are logged as completed ones, the report says the team spoke to sixty prospects when it spoke to twenty. Attach every activity to the deal as well as the contact whenever a deal exists. An activity attached only to the person vanishes from the deal history, which is exactly where the next person will look.

Log it the same day

Memory decays faster than anyone believes. A call logged three days later loses the buyer's exact objection, the name they mentioned, the date they said they would decide. Those are the details that make the record worth keeping. If you are logging on behalf of a salesperson from their notes or recordings, do it within the working day the material reaches you, and log what the source says rather than what you infer from it. When something in the source is unclear, mark it unclear instead of smoothing it over. A record saying a decision date was mentioned but not captured is more useful than a confident date that turns out to be invented.

Notes someone else can use

Write the note for a colleague who knows nothing about the deal and picks it up in six months. Three things carry almost all the value: what the buyer actually said, what was agreed, and what happens next with a date. Keep quotes as quotes when the exact words matter, especially about price, timing and competitors. Cut the adjectives. Great call, very positive tells the next reader nothing, while they asked for pricing on forty seats and said budget clears in the new fiscal year tells them everything. Length is not quality. Five accurate lines beat a page of impressions, and five lines actually get read, which the page does not.

The next step is the point

If a CRM has one field that predicts whether a deal closes, it is the next step with a date. A deal with no scheduled next step is drifting, whatever stage it sits in, and a good sales lead scans that column before anything else. So every activity you log should end by setting or updating the next step. If the source material does not say what the next step is, leave it empty and flag it rather than inventing something plausible. Empty next-step fields are the fastest way to find the deals in trouble, which is precisely why they must be honestly empty rather than politely filled.

What never goes in a note

A CRM note is a business record. It can be read by the whole company, exported in a legal dispute, or shown to the buyer by accident. So it holds facts and agreements, not opinions about people, not guesses about their finances or their personal life, and nothing you would not repeat with them in the room. Never paste passwords or card details into a note; if a buyer sends them, raise it through the proper channel and keep them out of the record. On AfterDesk, your channel for anything sensitive is the delivery note to the operator, never a message to the client, because you have no contact with them at all.

Remember

  • An activity is something that happened, never something planned.
  • Attach activities to the deal too, or the deal history looks empty.
  • Log the same day, and record what the source says rather than what you infer.
  • Every deal needs a next step with a date, or an honestly empty one.
  • Notes are business records. Facts and agreements only.

The Weekly Pipeline Report

~8 min

What a sales lead is asking

A sales lead opens your report with four questions in mind. Will we hit the number this quarter? Which deals are actually moving? Where is it stuck? And can I trust these figures? Everything you include should answer one of the four. That is why a report listing every open deal is useless: it answers none of them and hands the reader the work you were paid to do. Write for a person with six minutes who will act on two lines of it. The best pipeline reports are short, the same shape every week, and boring in a specific way: the reader knows exactly where to look, so the changes jump out at them.

The six numbers to lead with

Start with total open pipeline value against target, because coverage is the first question. Then the count of open deals and the average size, which explains whether the total comes from many small deals or one large one. Then what closed won and closed lost since the last report, with the loss reasons if that field exists. Then what entered as new. Then the deals whose close date has slipped into the past. Present them as a small table with this week, last week, and the difference. The difference column is where a busy reader looks first, and it is the column most people forget to include.

Movement matters more than totals

A total tells you where you are. Movement tells you where you are going. So the heart of the report is what changed: which deals advanced a stage, which moved backwards, and which have not moved at all since the last report. That last group is the one nobody wants to write down and the one the sales lead needs most. List the deals with no activity in the last two weeks, with the owner and the days since the last touch, and keep it to the largest ten so it stays readable. You are not accusing anyone. You are showing where attention will pay, which is the entire job.

Build it from an export

You rarely need more than a CSV export and Google Sheets. Export the open deals with these columns: deal name, company, owner, amount, stage, close date, created date, last activity date. Paste it into a raw tab and never edit that tab, so your work can always be rebuilt from the source. Do the arithmetic on a second tab using SUMIF by stage and by owner, COUNTIF for the deal counts, and TODAY minus last activity date for staleness. Keep the report itself on a third tab, formatted for reading, with the export date written at the top. Same columns, same order, every week. That consistency is what turns a spreadsheet into a report people rely on.

Say what the data cannot say

Every report should end with a short data quality note, because the numbers mean little without one. Say how many deals have no close date, how many have a close date in the past, how many have no amount, and how many have no owner. Give the counts plainly and do not soften them. If the export was taken mid-import, or a stage was renamed this week, say that too. This section protects the reader from acting on a number that is thinner than it looks, and it protects you, because a figure you flagged is never a figure you hid. On AfterDesk the same honesty goes into your delivery note.

Remember

  • The lead wants four answers: coverage, movement, blockages, and whether the figures are trustworthy.
  • Lead with six numbers and a difference column against last week.
  • Stalled deals are the most valuable list in the whole report.
  • Export to a raw tab, calculate on a second, present on a third.
  • Close every report with a data quality note. Counts, not softening.

Stale Pipeline and the Sweep

~8 min

How a pipeline goes stale

A stale pipeline has a look you learn to recognise within a minute. Close dates clustered on the last day of quarters that have already passed. Deals sitting in one stage for months with no activity. A stage everyone quietly stopped using. Owners who left the company still holding thirty deals. Amounts that are suspiciously round, all ending in triple zeros. Contacts with no company attached. The same buyer appearing four times. None of these are opinions: each is a filter you can run and a count you can report. Start any clean-up by producing those counts, because they tell the client how big the job is and give you a before figure to show against your after figure.

Before you touch anything

Take a full export first, every object, and keep it untouched until the work is approved. That export is your undo button, and in most CRMs it is the only one you get. Then agree the rules in writing before you apply them: what counts as stale, what happens to deals owned by someone who left, whether closing a dead deal needs a reason code. Ambiguity here is not a detail, because a sweep multiplies it by a thousand records. On AfterDesk that agreement comes from the brief and from the operator, so if the brief does not settle a rule, ask before you sweep rather than after. One question costs minutes. One wrong rule costs the whole job.

Sort into four buckets

Work each stale deal into one of four buckets and nothing else. Revive: there is recent evidence of life, so it stays open and gets a real close date and a next step. Reschedule: it is genuinely alive but the date was fantasy, so the date moves with a note explaining why. Close lost: nothing has happened for months and nobody can point to a reason it is still alive, so it closes with a reason recorded. Escalate: you cannot tell, so it goes on a list for a human who knows the account. Four buckets handle almost any sweep, and the fourth is what makes the other three safe to apply.

Propose, do not delete

Your default output from a sweep is a proposal, not a completed action. Build a sheet with one row per record showing the current values, the proposed values, and the reason, then let the owner approve in batches. Yes, it is slower. It is also the difference between a contractor a client trusts with their database and one they never let near it again. Deletion in particular is almost never yours to perform. Closing a deal as lost keeps the history; deleting it destroys reporting for every past period that deal appeared in. If someone asks you to delete records, ask for that instruction in writing and ask what they want preserved. Archive over delete, every time.

Delivering a sweep

Deliver three things. First, the before and after counts on the same measures you started with, so the improvement is a number rather than a claim. Second, the exceptions list: what you could not decide and why, with enough detail that someone else can finish it. Third, the rules you applied, written plainly, because those rules are what stops the mess returning and they usually become the client's new standard. Keep your working files inside the task. On AfterDesk that means no copies on your own drive after approval, no uploads to outside tools, and nothing from that database used as an example anywhere. Their pipeline leaves your machine when the task closes.

Remember

  • Staleness is measurable: past close dates, no activity, departed owners, suspiciously round amounts.
  • Take a full export before any bulk change. It is your only undo button.
  • Sort every stale deal into revive, reschedule, close lost, or escalate.
  • Propose changes for approval. Archive over delete, always.
  • Deliver before and after counts, an exceptions list, and the rules you applied.

Keeping It Clean

~6 min

A habit, not a project

Every database you are asked to rescue was cleaned before, often more than once. The reason it is dirty again is that the clean-up was a project and hygiene is a habit. The rules that hold are the ones attached to a moment: nothing saved without an owner, no deal advanced without evidence logged, no import run without a duplicate check, every call logged the same day. If you work on a client's CRM regularly, propose one small standing routine rather than another large clean-up, and show the counts each time so the value stays visible. Clients pay for the big clean-up. What they actually need, and quietly value more, is that it never becomes necessary again.

The data dictionary

A data dictionary is one page saying, for each field that matters, what it means, who fills it, when, and what the allowed values are. Lead source means how this contact first reached us, chosen from this list of eight, set once and never changed afterwards. It is boring to write and it settles nearly every argument a team can have about their data. If a client does not have one, offering to build it from what already exists is among the most valuable things you can hand over, and it costs you a couple of hours. Keep it where the CRM is, not in somebody's inbox, and put the date and the author at the top.

Guard the doors

Data gets in through a few doors, and hygiene is mostly guarding them. Every import should be checked before it runs: column headers mapped deliberately, a duplicate check on email and domain, a test with twenty rows, and a note of the file name and date so the batch can be identified later. Web forms should write into picklists rather than free text wherever they can. Integrations that create records automatically deserve a look every month, because a quiet one can generate thousands of records nobody asked for. When you inherit a mess, the door is usually still open. Finding it and saying so is worth more than the cleaning.

A weekly and monthly rhythm

Weekly, take fifteen minutes: deals with close dates in the past, deals with no next step, new records missing an owner, and anything created by an import since the last check. Monthly, take an hour: run the duplicate passes again, review the picklists for new junk values, check for owners who have left, and compare the completeness of the reporting fields against last month. Write the same short summary each time, with counts. The rhythm matters more than the length. Fifteen honest minutes a week keeps a database in a state that would otherwise take two full days a year and a great deal of arguing.

Confidentiality while you work

A CRM is the most sensitive file most companies own: their customers, their prices, their losses. Treat exports accordingly. Do not upload a CRM export to a spreadsheet cleaner, a deduplication website, or an AI tool to speed the job up, unless the client or the brief explicitly allows it, and on AfterDesk that means the brief and nothing else. Work in the client's own environment, or in files that stay inside the task. Delete every local copy once the work is approved. Never use a real pipeline as a portfolio sample, even with names removed, because deal sizes and industries identify companies far more easily than most people expect.

Remember

  • Clean-ups fail because hygiene is a habit attached to moments, not a one-off project.
  • A one-page data dictionary settles most arguments a team has about its data.
  • Check every import: mapped headers, duplicate check, twenty-row test, dated file name.
  • Fifteen honest minutes each week beats two full days a year.
  • CRM exports are highly sensitive. They never leave the task and never become samples.

Sit the exam

The courses are free. The work is real.

Twelve scenario questions. Pass at ten. Three attempts a day. The bar is the point.

No paid tier. No certificate fee. No upsell. Not now, not later.

Create a free account

You can start the first course tonight.

The curriculum

Everything, and what is in it.