Skip to main content
All articles

Sales infrastructure: the rules layer nobody builds.

Companies without a dedicated ops person usually own the tools and improvise everything between them. The list is bought, the CRM is bought, the sending tool is bought, and the rules that connect the three live in someone's head, which is the part that breaks when that person is on holiday.

Toni MedicToni MedicSalestructSeptember 9, 20268 min readSales Operations
The short version
Illustration of an exploded axonometric assembly drawing, six rectangular component blocks separated vertically in the air with thin alignment lines between them.
Pulled apart, a sales operation is a small number of parts in a fixed order. Most arguments about it are really arguments about one layer.

"Sales infrastructure" gets used to mean whatever the person saying it sells. Tooling, if they sell tools. Process, if they sell consulting. Headcount, if they are a recruiter.

Here's the version that survives contact with a real company: it's the set of things that have to exist before a salesperson can do their job the same way twice.

The short version

  1. Five layers: where names come from, how you reach them, what the system records, the rules connecting those three, and who holds the accounts.
  2. Most teams without a dedicated ops person have bought layers one, two and three, and never built layer four.
  3. Layer four is the only one that cannot be purchased, which is exactly why it gets skipped.
  4. When a sales operation "depends on one person", that person is almost always holding layer four in their head.
  5. A tool problem produces a consistent failure. A rules problem produces an inconsistent one, which is how you tell them apart.

What is sales infrastructure?

Sales infrastructure is everything a sales team needs in place before selling becomes repeatable: the source of prospects, the channels used to reach them, the record of what happened, the rules governing all three, and the accounts everything sits in.

It is not the same as a tech stack. A tech stack is a list of purchases. Infrastructure is what those purchases do when they are connected, plus the parts that were never for sale.

The five layers

1. Sources. Where names come from. Bought lists, enrichment tools, inbound forms, referrals, events, the founder's network. Most companies have several and can't say which produces the meetings that close.

2. Outreach. How you reach those names. Email sending, LinkedIn, phone, the sequencing that decides the order and the gaps. This layer is usually the best resourced, because it has the most vendors selling into it.

3. The record. What the system knows: the CRM, the fields on it, what a stage means, what counts as a next step. This is the layer everyone points at when something breaks, and it is usually a symptom rather than a cause.

4. The rules. Who picks up a new lead and within how long. What has to be true before a deal moves. Who is allowed to change a field. What happens when someone replies while the owner is away. What the handover contains when a deal is passed on.

5. The accounts. Whose name the domains, the sending mailboxes, the ad accounts, the CRM subscription and the data are in. Boring until the day it isn't, and the day it isn't is always a day you did not schedule.

Layer four is the one that is missing

Illustration of a column of five stacked solid slabs where the second slab from the bottom is absent, and the slabs above it sag into the empty space.
Everything above the gap is still there. It just does not hold its shape any more.

Layers one, two, three and five all have vendors. You can buy your way to a list, a sending tool, a CRM and a set of accounts in an afternoon, and most teams have.

Layer four has no vendor, because it is decisions. That makes it the layer most likely to exist only as convention: things the team happens to do, learned by watching each other, never written down and never enforced by the software.

That produces a specific symptom. The operation works, and it works differently depending on who is doing it. Two reps handle the same inbound lead in two different ways and both think they are following the process, because there is no process, there is a habit with two versions.

The diagnostic that separates a tool problem from a rules problem: a tool problem fails the same way every time. A rules problem fails differently every time. If your answer to "what happens when X" is "depends who catches it", that is layer four, and no purchase fixes it.

Which layer is failing you

Read the symptom, not the complaint. The complaint is almost always "the CRM is bad".

What you observeThe layer
Plenty of activity, few qualified conversations1, sources
Good conversations, nothing lands in the calendar2, outreach
Nobody can answer a question about a live deal3, the record
Two people handle the same situation differently4, rules
It only works when one specific person is available4, rules
Nobody can say who holds the sending domains5, accounts

The two rows in the middle of that table are the ones most often misread as a CRM problem and treated with a CRM purchase, which is how a company ends up on its third system in four years with the same complaint.

Building layer four without a project

It is smaller than it sounds. Four decisions, each written in a sentence, each enforced by a setting rather than by a reminder.

Who picks up a new lead, and by when. One name or one rule, and a number of hours. Not "the team". If the answer depends on who happens to be looking at the inbox, you do not have this decision, you have an inbox.

What has to be true before a deal advances. One binary test per stage, produced by the buyer, stored in a field that can be required. This is the enforcement layer for pipeline stages and the highest-value item here.

Who may change the shape of the system. Fields, pipelines, automations. One named person, permission removed from everyone else. The detail is in who is allowed to change what.

What a handover contains. Write the list once: the next step and its date, who the buyer's decision-maker is, what they have already been sent, the constraint they named. Then require those fields. A handover that depends on a conversation fails on exactly the days that conversation cannot happen.

Nothing there needs new software. All four are settings in what you already run, which is the point: the missing layer is missing because it was never a purchase, not because it was expensive.

If you want a single question to start with, ask two people what happens to a lead that arrives on a Saturday. Different answers, and you have found layer four.

Common questions

Sales infrastructure is everything that has to be in place before selling becomes repeatable: where prospect names come from, the channels used to reach them, the record of what happened, the rules connecting those three, and whose accounts everything sits in. It is not the same as a tech stack, which is a list of purchases. Infrastructure is what those purchases do once connected, plus the parts that were never for sale.
A tech stack is the tools. Infrastructure includes the decisions that make those tools behave the same way twice: who picks up a lead and by when, what has to be true before a deal advances, who may change a field, and what a handover contains. Those decisions have no vendor, which is why a company can own a complete stack and still have no infrastructure.
A tool problem fails the same way every time. A process problem fails differently every time. If the answer to "what happens when this comes in" is "depends who catches it", the fault is in the rules rather than in the software, and buying a different CRM reproduces it exactly, which is how companies end up on a third system with the first complaint.
Because that person is holding the rules layer in their head. They know who picks up what, what makes a deal real, and what a proper handover contains, and none of it is written down or enforced by a setting. It reads as dependency on a talented individual and it is actually an unbuilt layer, which is good news, because writing four decisions down is cheaper than replacing the person.
At minimum: the next step and its date, who the decision-maker is on the buyer side, what has already been sent to them, and the constraint they named such as a deadline or a system that has to stay. Write the list once and make those fields required, because a handover that depends on a conversation fails on exactly the days that conversation cannot happen.

Sources

This post carries no external vendor claims, so there is nothing to cite here beyond the linked posts, each of which records its own primary sources. The five layers and the four decisions are a reasoned model rather than research, and are presented as such.