CRM best practices that survive year two

Why does the second year decide whether a CRM lasts?

Most CRM advice is written for the first month: import the data, name the stages, train the team, go live. That advice is fine, and if you are earlier than this article assumes, our guide to what CRM is covers the ground before it. It is also not where systems fail. A CRM rarely collapses in the opening weeks, when everybody is watching. It degrades later, when nobody is.

Year-two failure is quiet. The board still loads. The reports still run. What changed is underneath: open deals stopped being true months ago and nobody closed them, two reps mean different things by the same stage name, the person who built the whole thing has left, and nobody wants to say out loud that the forecast has become a guess.

Three forces do most of the damage: data going stale, stage names drifting from their meanings, and the loss of whoever owned the configuration. None is a product problem, so changing vendors does not fix them. They are maintenance problems, and maintenance responds to habits with a date attached. Each practice below names a habit, a cadence, and the failure it prevents.

Practice	What it prevents in year two	Cadence
Written stage definitions	Two people meaning different things by the same stage	Reviewed quarterly
A next step and a date on every open deal	A pipeline that is technically full and factually empty	Reviewed weekly
A field audit	Forms so long that reps fill them with anything that passes	Once a year
Automated typing	Adoption eroding because updating the record is work	Reviewed twice a year
A named owner and a named successor	The reasoning leaving with the person who built the system	Owner monthly; successor named now
The same three questions every month	Reports that cannot show a trend because they keep changing	Monthly
Re-onboarding the whole team	Workarounds spreading faster than the process	Once a year
An exit test	Learning at migration time what does not come out	Once a year

1. Write down what each stage means, then read it again every quarter

Stage names are the most durable thing in a CRM and the least stable. The names survive; the meanings drift. “Qualified” meant a budget conversation in year one, and by year two it means the buyer replied. Nothing announces the change. The pipeline report keeps producing numbers, and the numbers are now made of different things than they were.

Write one sentence per stage saying what has to be true before a deal enters it, and keep the sentence where the work happens rather than in a document nobody opens. In Senitix a stage can carry guidance text and up to five key fields, so the definition sits on the board itself. It is guidance, not enforcement: nothing stops a rep dragging a card, and nothing should. An exit criterion is an agreement between people. Software that blocks the drag only teaches the team to be inaccurate in a different field.

Once a quarter, take five open deals and read the definitions against them. If a deal does not match the sentence for the stage it is in, either the deal moves or the sentence was wrong, and both outcomes are worth the meeting. Rewrite a definition when the process genuinely changed (a new approval step, a new kind of buyer) and leave it alone otherwise. Designing the stages in the first place is a different exercise, and our guide to CRM strategy is where that one lives.

Here is what four stage sentences can look like. They are an example, not a template: what the team agrees a stage means matters more than what the stage is called.

• Qualified: someone with authority described the problem and a timeline in their own words.
• Proposal sent: a priced document reached the customer, and the team knows who received it.
• Negotiation: the question is terms, not whether to buy.
• Verbal yes: a named person committed and told the team what has to happen before signature.

2. Put a next step and a date on every open deal

The most recognizable year-two condition is a pipeline that is technically full and factually empty. Deals sit in one stage for months. Close dates get pushed to the end of whichever quarter is current. Nobody is lying; people are reluctant to write off a deal while there is still a chance, and reluctance accumulates.

The working rule is short. An open deal with no scheduled activity is not a deal, it is a hope. Every open deal carries one activity with a date on it: a call, a follow-up email, a meeting. Where the next move belongs to the customer, the step is “chase on Thursday”, not “wait”.

Two reports make this visible without a meeting: deals with no open activity, and time in current stage. Read weekly, they keep the pipeline honest for the cost of a few minutes. The Senitix AI daily digest and next-action suggestion can put the candidates in front of a rep in the morning; the rep still decides what the step actually is, and should.

The other half of the practice is closing things. Give the team a short list of lost reasons and use it: price, timing, no decision, competitor, no reply. A deal closed with a reason teaches you something about where the process leaks. A deal that quietly ages into nothing teaches nobody anything, and it inflates every forecast it appears in right up until it disappears.

3. Audit the fields once a year and retire the ones nobody fills

Fields accumulate. Every one of them was added in a reasonable meeting by someone with a reasonable question. The cost arrives later and is paid by other people: a form long enough that a rep on a busy Friday fills it with whatever passes validation. Bad data rarely comes from carelessness; it comes from forms that ask more than the moment can support.

Once a year, put every field on one page and ask three questions of each. Who reads it? In which report does it appear? What decision changes when the value changes? A field that fails all three is not neutral. It is a tax collected on every record anyone creates, forever.

One more question is worth asking on its own: is anything on the list a separate thing wearing a field’s clothing? A dealer, a vehicle, a site, a shipment. When the answer is yes, the value does not belong in a text box on an account. It belongs in a record with its own fields and its own history, which is what custom objects are for in Senitix. Where records live and how they relate is its own subject; our guide to the CRM database takes it up.

• Keep: a report or a routine decision depends on it, and someone can name which.
• Rename: people fill it inconsistently because the label asks an ambiguous question. Fix the label before you blame the team.
• Retire: stop requiring it, hide it from the form, export the values, then remove it. Retiring in that order means nothing is lost if you were wrong.

4. Let the system do the typing, not the thinking

Adoption does not collapse; it erodes, and it erodes where the CRM asks a person to retype something that already exists somewhere else. Every message copied by hand into a note is a small tax, and by year two people stop paying it, first on small deals, then on the ones that matter.

So remove typing wherever the record can fill itself. Connect the mailboxes and calendars (Gmail, Google Calendar, Microsoft Outlook, Microsoft Calendar or any IMAP/SMTP account) so correspondence and meetings land on the record without anyone forwarding anything. Turn the two or three reminders everyone forgets into automations. Let Senitix AI draft the reply and suggest the next action, with nothing sent until the rep confirms; most of the delay in a follow-up lives in the blank page, not in the decision. Plan details for all three are on plans and pricing; check them before you build the week around it.

Then draw the line at judgment. An automation may create a task, notify an owner, or stamp a date. It should not move a deal to the next stage, because a stage change is a claim about reality and a person has to own it. Systems that advance deals on their own produce beautiful pipelines and useless forecasts.

Some of the typing will disappear later rather than now. Saved email templates, macros and assignment rules are coming soon, and messaging channels (WhatsApp, Instagram, Messenger and Telegram) will land conversations on the record once they are released. Build the practice around what exists today and let the rest improve it. What AI decides and what it must not is set out in our guide to what AI in a CRM does today.

5. Name an owner, and name the successor now

A CRM that decays usually had one person who really understood it. They chose the stages, wrote the automations, and knew which report was the real one. Then they changed jobs. What leaves with them is not access (access transfers in an afternoon); it is the reasoning. Nobody remaining knows why a field exists, so nobody dares remove it, and the system slowly becomes a thing the team works around rather than in.

The owner is a role, not a job title: a few hours a month for someone close to the selling, with the authority to say no to a new field. Their work is the yearly audit, the quarterly stage reading, the monthly reports, and the decision log.

The log is the part that outlives them. Keep it inside the CRM rather than on a laptop, and write down four things: why each custom field exists and who asked for it, which automations run and what they touch, who holds which permissions, and, where approvals are in use, what the approval chain is and who agreed to it. In Senitix CRM a record keeps its own change history, showing who changed which field and when, and a setup audit trail logs changes to the configuration itself. The decision log records why, which is the half no system writes for you.

Then name the successor before you need one and have them do a month of the work alongside the owner. Handover as a scheduled event costs an afternoon. Handover as an emergency costs far more than that, and you find out the price at the worst possible moment. If a partner ran the original setup, the rule is the same: the reasoning has to end up in your building. Our guide to CRM implementation covers what to capture while a rollout is still fresh.

6. Ask the same three questions every month

Reporting decays in the opposite way to data: not through neglect but through enthusiasm. Someone builds a better version of the pipeline report. Someone else filters it differently for a board deck. By year two the same question has several answers, some of them disagree, and the meeting opens by deciding which number to believe.

Pick a small standing set and then leave it alone. Three questions are usually enough: what closed and against what expectation, what is at risk, and where deals stop. Same definition, same filter, same day of the month, read by the same people.

Three rules keep it stable. Save a report rather than rebuilding it each time. Add a new one only after the same question has been asked twice. And when a definition has to change, write down the date it changed and keep the old version beside it for a quarter, so the step in the trend line comes with an explanation.

Be strict about it, because a trend needs continuity more than precision: a slightly crude report read the same way all year tells you more than a perfect one rebuilt every month. Reporting is also the honest test of the field audit: if no report reads a field, the audit already has its answer. If you need candidates for the standing set, our guide to CRM benefits and how to measure them writes each measure as a fraction you can read the same way every month.

7. Re-onboard the whole team, not just the new hires

Training happens once, in the week of go-live, to people who were mostly going to be fine anyway. Everyone who joins afterward learns from the person sitting next to them, which means they inherit that person’s workarounds along with the process. Two years in, the way the team uses the CRM and the way it was designed to be used are two different systems sharing one login screen.

Book an hour a year and run it as a working session rather than a demo. Cover the stage sentences, which fields are required and why, what the lost reasons are for, and the paths people improvise around most: raising a quote and revising it (in Senitix, quotes carry numbered versions and produce a PDF), attaching a document, handing an account to a colleague.

The most useful part costs nothing. Ask each person to open their own worst record and talk through it. What they explain is the distance between the design and the practice, and it makes a better training list than anything you would have written in advance. Record the session and keep it where a new hire can find it in week one.

And if four people separately invented the same workaround, the fault is in the design rather than the team: fix it in the field audit, not in the training.

8. Test the exit once a year

The annual exit test takes an afternoon and answers three questions at once: can you get your data out, does anyone still have access who should not, and is anything being kept that should already have been deleted.

Export everything and open the files. Not a report: the records. Contacts, accounts, deals, activities, notes, attachments. Check the parts that exports usually lose: history, file attachments, and the links between records that make a contact belong to an account. Every Senitix plan, Free included, exports records as CSV or Excel files, and a full data export can be scheduled. There is no one-click connector between Senitix and another CRM in either direction, so a file export is the way out and a file import the way in, which is exactly why the files deserve a yearly look before a migration depends on them.

Then read the user list against the payroll. Live logins belonging to people who have left are the easiest finding to act on and the least excusable to leave. Review which records each role can reach (on paid Senitix plans, the access explorer shows why a given person can see a given record) and write down what you decided and why. Rule-based sharing is coming soon, so for now that review is a manual one. So is the duplicate check: Senitix can merge the pairs you find, while configurable matching rules that would catch them automatically are coming soon.

Finally, retention: what does your own policy say you delete, and has any of it actually been deleted? The point is not compliance theater. A system you can leave is a system you can trust, and a quiet afternoon is a better time to find out which one you have than the week you have to migrate.

Where should you start if you are already in year two?

Do not attempt all eight. Take the two that describe the failure you can already see (usually the stale pipeline and the missing owner) and put a name and a date against each. A habit without a date is an intention.

Then schedule the two annual ones now, on whatever date you are reading this: the field audit and the exit test. They cost an afternoon each and they catch the problems that show no symptoms until they are expensive. If you are still choosing the system these practices will run on, the list of CRM features worth checking is the earlier read; if you want to see how the parts sit on one record, that is what Senitix CRM is.