A financial-services client came to me with a familiar mess. Their inside-sales floor had a hundred agents working personal-loan upsell, and every agent opened the morning to a personal list view of five hundred leads. They picked whatever caught their eye, dialed, and logged a task by hand if they remembered. Six-month-old leads sat untouched at the top of the list while the genuinely hot inbound ones aged out unworked. Conversion sat at 1.2 percent, and nobody could say which leads were being neglected or why.
The instinct in that situation is to buy a faster dialer or write a stricter SLA. Neither fixes the real problem, which is that the reps have no system telling them what to do next. They are making the prioritization decision a hundred times a day, badly, instead of doing the work. That is exactly the gap Sales Cadences and High Velocity Sales are built to close, and it is why I almost always frame the project to the client as "we are going to take the next-action decision away from your reps" rather than "we are buying an engagement tool."
That framing matters because it sets expectations correctly. A Sales Cadence is not a marketing drip. It does not send anything on its own. It is a sequenced playbook that drops human tasks into each rep's queue, branches on what actually happened, and logs every touch as activity. Get that distinction wrong in the discovery conversation and you will spend the rollout fighting a client who expected automation to do the selling.
A cadence is a rep-action playbook, not a drip
The mental model I give every client is this: a cadence routes work, it does not perform it. Each step is something a person does, a call, a personalized email, an SMS, a LinkedIn touch, with a wait between steps and a branch on the outcome. Salesforce surfaces those steps every morning in a Work Queue, so the rep opens the app and sees "these are today's calls, these are today's emails, these are overdue," sorted by priority instead of by accident.
For the loan client I designed a fourteen-day outbound cadence that looked roughly like this: a call at 10am on day one, the intro email that same afternoon, a follow-up call on day three if the first did not connect, an SMS on day five through their existing provider integration, a last-attempt call on day seven, and an automatic move to Disqualified on day fourteen if nothing landed. Ten steps over two weeks, which sits right in the range I trust for outbound. I have learned the hard way that a thirty-step cadence just trains reps to skip steps, while a three-step cadence never builds enough contact to convert anyone. Seven to twelve steps across one to two weeks is the band I defend in design reviews.
The single most underrated step type is Wait. It is what keeps a cadence from reading as spam. I put a minimum of a day between an email and a call on the same theme, I run the waits on a Business Hours calendar so nothing fires over the weekend, and on a recent engagement I had to manually load the local public holidays into Business Hours because Salesforce ships no calendar for that region. That is a small thing that nobody catches until the cadence sends a "checking in" call on a national holiday.
Branching and Listeners are where designs go wrong
Branching is what makes a cadence smart, and it is also where I see the most self-inflicted damage. Cadences branch three ways: on call outcome, on email engagement through a Listener step, and on field criteria. On the loan cadence, the day-one call branched on outcome the obvious way, "connected and interested" exits the cadence and creates an opportunity, "voicemail" diverts to an email, "no answer" waits for the day-three call, and "bad number" exits straight to a data-quality status. That is clean because it is one level deep.
The trap is depth. My first draft of an earlier cadence had four nested layers, call outcome into email-open into field value into a custom condition, and it was effectively impossible to reason about or hand off to the client's admin. I now hold a hard line at two branch levels. Anything more complex gets pushed out to a Flow that sits beside the cadence, where the logic is visible and testable, rather than buried in a canvas of nested branches.
Listener steps deserve their own warning. A Listener waits for an email to be opened, clicked, or replied to before advancing, and on paper it is elegant. In practice, open tracking relies on a pixel that Apple Mail Privacy and Outlook secure mode quietly strip, which produces two failures at once. You get false positives, where a mail client preloads the image and the cadence thinks the prospect engaged and skips the call, and you get false negatives, where a real open never registers and the lead sits stuck because the Listener has no exit. I now configure every Listener as "wait for the event OR a maximum of N days" so there is always a fall-through, and for consumer segments I skip Listeners entirely and branch on call outcome instead, where the signal is real. I reserve open-and-click branching for B2B audiences where engagement is more reliable and the privacy blocking is less universal.
A few other production failures recur often enough that I now design against them by default. When a lead's owner changes, the cadence steps stay assigned to the old rep, so if that rep has left, the calls route to an inactive user and dangle forever; I build a Flow on owner change to reassign the steps. Opt-out has to live at the record level, on a single Do Not Contact field that every cadence checks as an exit criterion, otherwise a prospect who opts out of cadence A keeps getting hit by cadence B. And bulk-adding leads needs a guardrail, because a manager who drops two hundred leads onto one rep in a minute can generate six hundred tasks for that rep's morning and burn them out on day one.
High Velocity Sales is the engine, but only for the right reps
Sales Cadences are one piece of High Velocity Sales, now branded Sales Engagement. The other pieces are the Work Queue, the integrated engagement tools like the Lightning Dialer and email tracking, and the optional Einstein Lead Scoring add-on. The Work Queue is the part reps actually live in: overdue first, then highest-scored leads, then new ones, with the dialer auto-dialing from the record and the call outcome advancing the cadence automatically.
The license is not cheap, around 75 dollars per user per month at list, and the most important advice I give buyers is to scope who gets it. High Velocity Sales earns that price for genuine inside-sales reps doing thirty-plus calls a day. For field reps making five calls a day it is wasted spend, and those people are better served by the Inbox and Einstein Activity Capture features they already have. I have watched clients buy floor-wide licenses out of fairness and then struggle to justify the renewal because half the seats barely touch the Work Queue.
Einstein Lead Scoring carries a similar warning. It scores each new lead on conversion probability, and on the loan account the segmentation was genuinely useful once it was trained, leads scoring above seventy converted at eight percent while leads under thirty converted at three-tenths of a percent. But the model needs roughly a thousand historical closed-won leads to mean anything. A client six months into their CRM does not have that, and turning Einstein on too early produces near-random scores that destroy rep trust in the whole system. I tell clients plainly to wait six to twelve months after the HVS go-live before enabling it.
Prove the license pays, and measure honestly
The conversation that decides whether this program survives is the one with the CFO, so build the ROI case before go-live, not after. That means capturing a clean baseline in the three months before rollout, calls per day per rep, emails per day, conversion rate, pipeline generated per rep. On the loan account the post-rollout numbers were dramatic, calls per rep per day went from 65 to 95 simply because reps stopped spending time picking leads, conversion moved from 1.2 to 3.5 percent, and lead aging past seven days collapsed from 35 percent to 5 percent. Against a license cost in the low tens of thousands, the incremental pipeline made the renewal an easy signature.
The discipline that keeps those numbers honest is measuring effectiveness alongside volume. If you track only calls per day, reps learn to dial fast and connect with no one, so every volume metric on the dashboard needs its effectiveness partner: calls paired with connect rate, emails paired with reply rate. I deliberately deprioritize Email Open Rate now, because Apple Mail Privacy inflates it across every iOS recipient whether they opened the mail or not; reply rate and meetings booked are the metrics I trust. Two more traps catch teams repeatedly. Pipeline gets double-counted when an SDR creates an opportunity that an AE later owns and both claim credit, which is a policy decision the client has to make, not a report you can quietly fix. And Time to First Touch reads as garbage when "Lead Created" is really the timestamp of a 2am bulk import, so I add an eligibility timestamp field that clamps the clock to business hours and measure against that.
The principle underneath all of it is the same one I open every cadence project with. The tool's job is to remove decisions from the rep's day so they spend their time selling instead of triaging, and your job as the consultant is to make sure the structure you build is shallow enough to maintain, honest enough to measure, and scoped to the people who will actually use it.
Implementing or optimizing Sales Cloud? Our Salesforce team runs discovery, designs the sales process, and configures Sales Cloud on production engagements. Get in touch ->
See our full platform services for the stack we cover.
