An honest inquiry-to-sale conversion is the number of sales divided by the number of leads — but only when you know exactly what enters the numerator and the denominator. Before you divide, you merge duplicates, subtract spam and sales calls, and decide whether a cancelled booking still counts as a sale. Skip those steps and the result depends on whoever happens to be counting: the same business looks like it has two different conversion rates in two consecutive months. Then comes lag — a customer asks in one month and pays in the next — so calendar comparisons mislead you before you even make an arithmetic mistake.
Below you will find the four definitions your result depends on, a table of what does and does not belong in the denominator, the cohort formula, and a worked example on made-up numbers that walks the whole path from raw contacts to a single figure. At the end there is a one-month table to fill in with your own data and three numbers worth tracking month after month.
Contact, lead and sale: the definitions your result depends on
So how do you calculate a lead-to-customer conversion rate so that it means the same thing in January and in June? You start with definitions. Without them, everyone in the company computes the ratio differently, and every argument about the result becomes an argument about words.
Four terms and one rule
Contact — any signal from the outside world: a call, a website form, a chat message, an SMS, an e-mail. You record each channel separately until you merge duplicates.
Lead — a contact that passes the bar: a new person asks about a specific service, leaves their details and can be called back. Spam and sales pitches do not pass.
Sale — the moment you define once: a paid deposit, a confirmed booking or a completed job. You pick one and keep it for the whole year, including the months when it makes the number look weak.
Cohort — all the leads from one month, tracked together until the end, no matter when the sale lands.
The basic formula and the cohort formula
conversion = sales ÷ leads — you express the result as a percentage. The calendar variant divides a month's sales by the same month's leads, and it misleads the most when customer decisions stretch over weeks.
cohort conversion = sales from a month's leads ÷ that month's leads — a sale counts whenever it actually lands: this month, next month or never.
The path of one contact:
- 01first contact
- →02duplicate merging
- →03lead qualification
- →04sale closed
- →05cancellation check
- →06month cohort
The denominator: what to subtract before you divide
A raw contact list is not yet a denominator. Before dividing, you remove four categories.
Duplicates come first: the same person calls, gets no answer, and messages you an hour later. That is one matter, not two leads. In a spreadsheet, duplicate inquiries pile up fastest because nobody merges them by phone number and e-mail address — with CRM and automations, phone, form and messengers land in one pipeline, so a duplicate shows up right away.
Spam and telemarketing: form bots, companies offering leasing, insurance or “profile upgrades”. None of them plans to buy your service.
Not your profile: a request for a service you do not provide, or a trip to a town you do not serve. A pleasant conversation does not turn into a lead.
Existing customers' matters: a reschedule, an invoice request, a follow-up stage of current work. That is service, not sales — unless you deliberately count contract extensions as new sales; then write it into the rule.
| Situation | Belongs in the denominator? |
|---|---|
| A new person asks about price and timing | Yes, that is a lead |
| The same customer calls, then writes | No — one lead |
| A telemarketer with a B2B offer | No — spam |
| A request for a service you do not offer | No — outside your profile |
| An existing customer asks to reschedule | No — that is service |
The numerator: is a cancelled appointment still a sale
The numerator has the same problem, just inverted: every unclear decision pushes the result up or down. A booking cancelled after the deposit, a refund, a no-show without cancellation — each of these either is a sale or is not, depending on the rule you adopt.
One rule, written down once
You have two honest options. First: a sale is a completed job — cancellations and refunds are subtracted from the sales count. Second: a sale is an accepted payment — cancellations stay in the numerator, and you keep a separate cancellation rate showing how much of the result is “soft”. Both are correct, provided you pick one, write it down and do not change it mid-year.
The same logic runs at the till: voids and comps read honestly only when the person at the register knows what counts as a sale and what counts as a return.
A worked example on made-up numbers: naive vs honest
A worked example on made-up numbers — plug in your own. A service business has 200 contacts in a month. After merging duplicates, 30 records disappear; after removing spam and sales calls, another 20 go — the denominator is 150 leads. That month produces 45 sales, of which 5 end in a cancellation after the deposit — the numerator holds 40.
honest conversion = 40 ÷ 150, while the naive count, with no cleaning: 45 ÷ 200. The difference in the result depends on the data mix: noise in the denominator pulls the result down, cancellations in the numerator push it up — which is why there is no universal “correction”, only clean data.
| Line | Naive counting | Honest counting |
|---|---|---|
| Contacts in the month | 200 | 200 |
| Leads after cleaning | 200 | 150 |
| Sales after cancellations | 45 | 40 |
| Conversion | 45 ÷ 200 | 40 ÷ 150 |
Lag: next-month sales break the calendar view
A continuation of the made-up example.
In February, 120 new leads arrive and 25 sales come from February's leads.
The February calendar shows 35 sales against 120 leads — because it added 10 sales from January leads to February's numerator. The February cohort currently stands at 25 sales and is still growing. The calendar inflates one month, deflates another, and no amount of careful counting fixes that — only cohorts do.
What to do with a lead that goes quiet between one month and the next is covered in a separate piece on follow-up automation.
Channels: one customer, one source
One customer fills in a form, calls and messages you within the same week. In your statistics that is still one person and one matter — one lead. The source is the channel of first contact: if the form came first, the sale from that lead belongs to the form, even if the final confirmation arrived by phone.
The first-source rule answers “where do customers come from”, not “which channel they fancy most along the way”. Without it, you count the same success twice and inflate the channel that is merely the second touch.
RODO: one customer card instead of three copies
Merging duplicates is not just statistical hygiene. A UODO guide quotes art. 5 of RODO (Poland's GDPR): personal data must be adequate, relevant and limited to what is necessary; correct and updated where necessary; and data that is inaccurate in light of the processing purposes must be erased or rectified without delay. The guide was written for schools, but the principle comes straight from the regulation and applies to every data controller.
For conversion measurement the practical conclusion is this: one card per customer, merged by phone number and e-mail address, instead of three copies in three spreadsheets, each remembering something different. Whether to merge entries or delete the surplus in your specific case is a call for your DPO or a lawyer, not for an internet template.
To make channels land in one place by themselves, connections between phone, form and messenger into one database help — as integrations do in a CRM system.
Where contacts disappear before anyone counts them
Before you even get to dividing, some contacts do not exist in any database. Calls answered on staff members' private phones, after-hours conversations, messenger threads never passed on — all of these are leads that statistics see as air.
What to do depends on the hole: how much missed calls cost in a company gives a formula for pricing the loss on your own numbers, one queue instead of five inboxes gathers channels in one place, and the analysis of why customers don't leave inquiries explains where contact never arrives at all. A form that does not lose messages is the subject of Lead forms.
Context: who actually measures conversion in a system
Where there is no CRM, the same numbers live in a spreadsheet or in someone's head — and that is where duplicates multiply while cohorts do not exist.
Do it yourself: a one-month table and three numbers
Take one closed month and walk the path in a spreadsheet:
- Collect all the month's contacts in one place — phone, form, messengers.
- Merge duplicates: one person is one lead, regardless of channel count.
- Subtract spam, sales calls and existing customers' matters.
- Count sales under your written rule and subtract cancellations.
- Divide sales by leads and record the honest conversion.
- Come back to this cohort next month and add the sales that landed later.
Three numbers you read off the table
30 ÷ 200 = 15% in the example above — shows how strongly channels overlap.Once set up, these three numbers can be collected without retyping: we write about reporting automation separately, and ready-made views come from Analytics and BI and Dashboards.
What counting looks like when a system does it
At Aura, this path looks like this:
Channels in one pipeline → one customer record → case status → a monthly report with numbers you can put next to revenue.
A human sets the rules — what counts as a sale, what as a duplicate, what closes a lead — and the system makes sure the rule is the same in January and in December. The risk moves elsewhere: a badly written rule spoils the result faster than manual counting, which is why it is worth walking the previous section's path by hand once, even in a spreadsheet.
See how it works in practice: CRM and automations brings contacts from phone, form and messengers into one pipeline, AI Reports describe leads and sales in plain language every week, and Lead forms take the customer through a few short steps instead of one long list of fields. None of them promises miracles — it promises that you compare like with like.
Frequently asked questions
Should a cancelled appointment count as a sale?
It depends on the rule you adopt once for the whole year. If a sale is a completed job, a cancellation is subtracted from the numerator. If a sale is an accepted payment, the cancellation stays and you track a cancellation rate alongside. The worst option is the third one: switching definitions from month to month.
Why does conversion jump from month to month?
Mostly because of lag and mixed cohorts: sales from last month's leads land in the current calendar figure, while some of this month's leads have not closed yet. On top of that, the share of spam and duplicates in raw contacts varies. Cohort counting removes both causes.
How do you count a customer who wrote and called the same day?
As one lead. You record the first channel as the source and treat the rest as duplicates merged by phone number or e-mail address. Two records for the same person inflate the denominator and depress the result.
How do you know if you have too many duplicates?
Measure the duplicate share in a month's contacts and compare it with your own previous months, not with someone else's numbers. A rising share means customers write across many channels while the database fails to connect them. A stable share is simply the price of multichannel contact.
Should conversion be broken down by channel?
Yes, but on cleaned leads and with the first-source rule. Then you see which channel brings people who buy and which only generates contact. Without duplicate merging, a channel comparison turns into a comparison of which channel tends to be second.
Is merging customer cards compliant with RODO?
Art. 5 of RODO requires data to be adequate, relevant, limited to what is necessary, and correct and updated where necessary — which points to one accurate card rather than several diverging copies. Whether to merge, rectify or delete an entry in your case is a call for your DPO or a lawyer; an internet template will not replace an assessment of your actual situation.