Best CRM for small business: how to choose one and use it

Ask whoever answers your phone which estimates are still open and who owes whom a call. If the answer takes ten minutes of scrolling through a handset, the system of record is somebody's memory. One owner with a good memory can run that way for years. Add a second person on the phone, or a third writing quotes, and the memory stops being reliable, because it now has to be shared and nobody wrote it down.

Buying the software is the easy half, and the published advice on it has drifted badly. HubSpot's own product page caps its free CRM at two users and 1,000 contacts. Most guides still quote the million-contact allowance, which belongs to accounts opened before September 2024. Four decisions settle whether this works: the category you buy in, the tool inside that category, the order you switch things on, and whether the outcomes ever reach your ad accounts.

Set your expectations with a current number rather than the famous one. Nucleus Research produced the "$8.71 back for every dollar spent" line that CRM vendors have quoted for a decade. In August 2023 the same firm published an update putting the figure at $3.10, a 37% decline from the $4.90 it measured ten years earlier, drawn from 63 case studies. Sixty-three self-selected cases make a thin base, so read $3.10 as a direction rather than a forecast. Any page still quoting $8.71 stopped reading its own source.

A CRM also sits inside a wider set of tools, and it is the piece that decides what the other pieces can see. We mapped the rest of it in the small business tech stack, which is worth reading first if you are still assembling the basics.

A sales CRM tracks a deal, and a deal has a contact, a dollar value, a close date and a pipeline stage. Field-service software tracks a job, and a job has an address, a two-hour arrival window, an assigned tech, a price book, an invoice and a payment. If your work involves sending a person somewhere at a time, the second record type is yours. Choosing wrong here is the most common expensive mistake in the category, and a price comparison will not reveal it.

The money follows the record type. Jobber prices the plan before the seat: Connect runs $99 a month on annual billing, with a five-user option at $149 and extra users at $29 a month each. Five seats of a $39-per-seat sales CRM cost $195 a month and buy no quoting, no dispatch and no invoicing, so the crew still books work in a spreadsheet. Most best-CRM roundups are lists of sales CRMs, which means a contractor reading one is shopping from the wrong shelf.

The same split runs through other trades. Law firms have Clio. Photographers and event planners have Dubsado and HoneyBook. Salons and studios have Mindbody and Boulevard, real estate teams have Follow Up Boss, and dental offices have Weave. When a tool exists for your trade it usually wins, because it already knows what a matter, a chart or a route is, and you will not spend a month renaming fields to pretend otherwise. The businesses this covers hardest are the ones with a truck and a schedule: plumbing, HVAC, roofing, electrical, pest control and landscaping.

Two tiers sit above Jobber. Housecall Pro covers roughly five to fifty techs and adds financing and payment depth, with published figures near $65 a month and up, which we have not confirmed at the vendor. ServiceTitan is enterprise home services sold by quote, and it lands in ten-truck shops that would run better on a tenth of the money.

No single tool wins the category. The question worth asking is which one fits the way you sell, and how many people need a login. Start with the free tiers, because three of them are real. Zoho CRM is free for three users, Bigin gives one user 500 records, and free HubSpot covers a two-person shop until the third person needs a seat.

Three traps sit inside that table. HubSpot's $7 headline is a new-customer promotion, and its pricing page carries the disclaimer that the discount is for new customers only and available for a limited time, so budget the $20 list price. Close's $9 Solo plan is a different product from the Close people recommend, because the power dialer that earns the money sits on the $99 and $139 plans. Keap's $249 covers two users and 1,500 contacts before the contact bundles escalate, which puts it eight to thirty times the monthly cost of Pipedrive or Bigin.

Annual billing is the fourth trap, and it costs the most. The discounts are real: Close runs up to 50% cheaper on annual terms, Pipedrive around 17% to 23%, Zoho and Bigin up to roughly 35%. The dominant failure here is abandonment, and prepaying twelve months makes abandonment expensive without making it any less likely. Pay monthly through the pilot, then convert once the team has gone six weeks without being reminded to use it.

Adoption dies around month three, and the cause is usually visible on day one: the pipeline stages came from a template instead of from the business. "Lead, Qualified, Proposal, Closed Won" describes a software company. A roofing company runs "Called back, Estimate scheduled, Estimate sent, Job booked, Job done, Paid." Stages written in the crew's own words get updated, and stages written by a vendor get skipped.

1. 1

   Week 1: import the active contacts, then build one pipeline

   Bring over the customers you have touched in the last two years, and archive the rest into a spreadsheet you can still search. Published estimates put contact data decay between roughly 22% and more than 30% a year, and the range is wide enough that the honest version is that a good share of your old list is wrong today. Import 8,000 stale records and the team learns on day one not to trust the system.
2. 2

   Week 2: connect the form and the phone so records create themselves

   Every web form submission and every missed call should create a record without anyone typing. Manual entry is the largest tax on adoption, and it is the first thing a busy team drops. Hidden fields can carry the lead source and the ad click ID along with the name, which you will need in the last section of this guide.
3. 3

   Week 3: answer inside five minutes, automatically

   A new inbound lead fires an automatic text and email within 60 seconds, then assigns a task to a named human. Every tool in the table above can do this in an afternoon, and the automation side of the work is where the return arrives fastest.
4. 4

   Week 4: require the lead source, then run one meeting off the pipeline

   Make source a required field on manual entry and a hidden field on every form. Then run the weekly sales or scheduling meeting from the pipeline view rather than a printout. The meeting is what makes the tool real, and a CRM nobody meets in front of turns into a filing cabinet.

The five-minute rule in week three has the strongest evidence in this guide, and the study behind it is older than most people assume. Professor James Oldroyd's lead response research, published in 2007 with InsideSales.com, read three years of data across six companies, more than fifteen thousand leads and over one hundred thousand call attempts. Calling a web lead within 5 minutes instead of 30 changed the odds of reaching that person by 100 times, and the odds of qualifying them by 21 times. The sponsor sold dialer software, so weigh it accordingly. Nothing since has pointed the other way.

The gap between knowing that and doing it is a workflow problem. Moniveo moved lead response from 48 hours to 5 minutes by putting the routing in software instead of a person's inbox, which we wrote up in this case study. We build that wiring as part of our automation work, and you can also build it yourself in an afternoon with the native automations in any tool above, which is the honest recommendation for a shop of three.

The scare number is sourced badly. Dozens of pages report that 70% of CRM projects fail and credit Gartner, and nobody can produce the report. Rates from surveys that do exist run from about 18% at AMR Research in 2005 to roughly 70% at Butler Group in 2002, because each firm counted a different thing as failure. Gartner's own analyst Ed Thompson has pushed back on the failure narrative in print.

The mechanism is easier to pin down than the rate. People stop entering data, and a record nobody updates is worse than a spreadsheet, because the spreadsheet never claimed to be the source of truth. That happens when the stages do not match how the work moves, when the fields exist to feed a management report rather than to save the user a step, and when the person who bought the tool never opens it.

Headcount is a poor threshold for the buying decision. The trigger is usually an event you can name. A customer called three weeks ago and never heard back. A second estimator gets hired, and the founder's phone stops working as the shared record. A key employee leaves and takes the relationships along, which is the sharpest trigger there is. The trades should build this in the slow quarter, when the phone is quiet, rather than in the week the air conditioners start failing.

Everything above improves the inside of the business. This last play changes what your advertising costs. Left alone, Google optimizes toward whatever you counted as a conversion, and for most small advertisers that is a form fill. It then buys more of whatever fills forms: tire-kickers, wrong-service requests, job applicants and spam. The account looks busy while the calendar stays empty.

Offline conversion import closes the loop. When a lead becomes a qualified lead or a won job, you send that outcome and its dollar value back to the ad platform, and Smart Bidding starts buying customers rather than clicks. Your ads pass a click identifier called a GCLID into the form, the CRM stores it on the record, and a stage change triggers the upload. Google's own documentation reports a median 10% increase in conversions for advertisers who paired first-party data, meaning email addresses and phone numbers, with GCLIDs, compared with a standard offline import.

A separate Google figure, a median 5% conversion lift on Search and 17% on YouTube, describes enhanced conversions for web, which is a different setup. Read it as a signal about first-party data rather than a forecast for this build, and do not stack the two numbers.

Two constraints decide whether the loop works at all. Uploads carry a deadline measured from the ad click: 90 days for a standard offline conversion import, 63 days for enhanced conversions for leads. A business with a four-month sales cycle that dutifully uploads closed-won gets nothing imported and concludes the method is broken. Pick a milestone that lands inside the window instead, such as qualified lead, estimate sent or appointment held, and give it an estimated value you can defend.

That June date is worth acting on early, because a homegrown script or an older Zapier recipe needs rebuilding before then. Imported conversions also take about three hours to show up, which is worth knowing before you assume an upload failed. The tracking layer underneath all of this is covered step by step in our Google Ads conversion tracking guide.

Meta runs the same loop through its Conversions API for CRM, where you send lead-stage updates back so the system optimizes for lead-to-sale instead of lead volume. Meta's guidance is to pick a milestone whose conversion rate sits between roughly 1% and 40%, which rules out both a rare high-ticket closed sale and any stage nearly every lead reaches. None of it works without the prerequisite from week four: a lead source and a click ID on every record. If your CRM cannot say which channel produced the last 50 jobs, the budget behind local paid ads is a guess.

The decision compresses into three checks. If your work happens at an address on a schedule, start with field-service software and stop reading comparison tables built for sales CRMs. If you sell a deal over calls and quotes, pick the cheapest tool that models your pipeline as it stands, because you can migrate later and you probably will. Then count the logins you will need in twelve months, since that number, not the feature list, sets the bill.

Six months from now, ask the question you started with. Which estimates are open, and who owes whom a call. If somebody answers from a screen in under a minute, the tool took, and the offline conversion loop is your next build. If the answer still comes out of a memory, the stages are wrong, and no other vendor will fix that for you.