Quick answer
A pipeline that looks random, good weeks with no pattern and dead weeks with no warning, usually isn't random at all. It's two or three ICP segments running through the same funnel and getting reported as one blended number. Split the data by segment before you touch a single sequence or subject line.
I'm Hlib Storchak. I run outbound and GTM builds for B2B companies, and I've booked 2000+ meetings for clients doing it. When I audit a company's go to market for the first time, the founder almost always describes the same symptom: pipeline that swings for no obvious reason. A great week, then three quiet ones, no pattern anyone can point to. The instinct is to blame tactics: the sequence, the subject lines, the follow up cadence. Most of the time that's not where the problem lives.
Why a random pipeline is a segment problem, not a tactics problem
The mistake I see most often when I take over a company's outbound is data blended across two or three different ICPs and reported as if it were one number. A campaign hitting a well fit segment might be converting at a genuinely strong rate. A second campaign, running at the same time into a segment that's a weaker fit, might be converting at close to nothing. Averaged together, the combined number looks mediocre and unpredictable, because it's actually two different, stable numbers pretending to be one.
That's the spine of the audit I run, and it's the part most companies skip. Every other section below, ICP, personas, messaging, sales process, RevOps, exists to feed that one split: which of your segments is actually driving the result, and which one is just adding noise to the average. Fix the split and the "randomness" usually disappears on its own, because you're no longer averaging a good number with a bad one and calling the result unpredictable.
Start with revenue targets, not the audit itself
Before touching ICP or messaging, I set a 12 month revenue target and break it into quarterly goals, each one small enough to run 2 to 3 tests or iterations against. The question that matters most here isn't the number itself, it's the split behind it: how much of that revenue should come from new business versus expansion of existing accounts. A 50/50 split is common and fine. What isn't fine is not deciding, because that decision changes almost everything downstream, including which segments are worth chasing at all.
Three questions carry the rest of the audit: what's the revenue target, what has to be true for the company to hit it, and which markets and ICPs matter most to get there. Budget, headcount, and market conditions all cap how fast you can realistically grow, and skipping that reality check is how a lot of GTM plans get built around a growth rate the team can't actually execute.
Define your ICP in tiers, then track each one separately
The primary ICP, what I call V1, needs four basics nailed down: industry, company size, geography, and budget. Granularity beyond those four helps, but those four are the floor. Layer in buying signals early rather than as an afterthought: "SaaS companies under 50 people" is a workable ICP, but "B2B SaaS companies under 50 people currently hiring for 5+ SDRs" is a sharper one, and it's the difference between a segment you can message specifically and one you can only message generically.
A secondary ICP exists for a reason that's easy to miss: if your total addressable market in the primary segment is small, say a couple thousand companies, you will eventually run through it. The secondary segment is the fallback, not a second priority to chase equally hard. I've gone deeper on the mechanics of this in how to build an ICP that actually converts, which is worth reading alongside this if you haven't nailed down your primary ICP yet.
| Segment tier | What it looks like in blended pipeline data | What to actually do with it |
|---|---|---|
| Primary ICP (V1) | Reads as your "good weeks" once isolated, often 2 to 3x the blended average | Give it the majority of volume and the sharpest, most specific messaging |
| Secondary ICP | Middling numbers that drag the average down without explaining it | Keep it as a deliberate fallback, not an equal priority |
| Time waster segment | Looks like random dead weeks, high effort, near zero close rate | Demote to third priority or drop until the market changes |
Segment prioritization is the step that actually filters the noise out. The questions worth asking per segment: which ICP brings the best deals, who's easiest to close given your current sales process, and which segment wastes the most pipeline. That third question is the one companies avoid, because the honest answer is often a segment that still brings in some revenue. It's still worth deprioritizing if the fulfillment and sales cost of chasing it outweighs what it returns.
Buyer personas change with the segment, not just the industry
Once the ICP tiers are set, map the decision makers and influencers inside each one, because they aren't the same people asking for the same things across segments. Marketers and sellers inside the same account care about different KPIs. Some buyers want your solution because it saves time, others because it makes them money, and pitching the wrong motivation at the right person still loses the deal. Map who owns the budget, and keep a running list of the roughly 20 most common objections you hear, because that list is what actually gets used on calls, not the persona document itself.
Positioning, messaging, and the clarity test
Your core value proposition should fit two or three sentences, and your competitive advantage, whatever it actually is: speed, price, integrations, ease of use, needs to be named plainly rather than implied. The test I actually use is simple: could a six year old understand your offer, and can you say it in one sentence. If you can't, the offer doesn't have a clarity problem you can message your way around, it has a clarity problem, full stop.
Messaging also has to flex per segment, because different ICPs care about different parts of the same product. A single message aimed at every segment at once is usually the reason the "randomness" I mentioned earlier shows up in reply rates specifically: one segment's message lands, another segment's doesn't, and the blended reply rate looks inconsistent for a reason that has nothing to do with the copy itself.
Offer architecture, and the case study most people get backwards
Structure the core offer around roughly 80% of your customers, then build a lower friction entry offer for segments that are harder to convert cold, with the plan to upsell once you've earned the relationship. Pricing, packaging, and deliverables all need to be specified from day zero, in the contract, not worked out during fulfillment.
Tip. If you don't have a case study that's actually relevant to the segment you're pitching, wrong industry, wrong company size, wrong result, leave it out entirely. A strong offer with no social proof reads better than the same offer sitting next to proof that doesn't fit. Most teams assume any case study beats none. It doesn't.
The final test on offer architecture is the same clarity test from messaging, just under time pressure: would a cold prospect understand the offer in 20 seconds. If the answer is no, the fix isn't a better pitch deck, it's a shorter, plainer sentence.
Demand generation across outbound, inbound, and partnerships
On the outbound side, this is targeting quality, data sourcing and enrichment, deliverability and domain setup, and sequencing strategy: how many follow ups, over which channels, at what volume relative to how targeted the campaign is. I default to running email and LinkedIn together for most clients, what I'd call an all bound motion, because relying on one channel alone caps how much of a segment you can actually reach.
Inbound matters just as much: content, lead magnets, on site conversion paths, and increasingly ranking inside LLMs as well as traditional search, since a growing share of buyers are researching vendors through AI tools before they ever open Google. Partnerships and referral programs round it out, and referral pipeline in particular tends to convert faster, because the prospect arrives already warmed by someone they trust.
Sales process: where deals actually get lost
Audit the discovery call structure, the demo and pitch, how objections get handled, and how proposals get written, because a proposal that's vague on deliverables is its own source of lost deals late in the funnel. Qualification during discovery matters more than most reps treat it: a deal that shouldn't have been qualified in the first place still shows up as a "random" loss later, when it was actually predictable from the first call.
Pipeline hygiene belongs here too. A CRM full of stale, uncategorized deals makes it functionally impossible to tell whether a slow month is a real trend or five deals nobody updated in three weeks.
RevOps, CRM, and the tech stack audit
RevOps exists to make sure the systems support scale rather than quietly cap it: clean CRM structure, clear lead routing so the right person gets notified at the right moment, reporting dashboards, and defined roles so a small team can still operate with discipline. Picking a CRM by brand recognition instead of by stage is a common, avoidable miss, and it's worth reading through Attio vs HubSpot for modern sales teams if you're mid decision, since the right answer depends far more on team size and motion than on which name is more familiar.
On tooling, I'd rather name what I actually run than gesture vaguely at "a good CRM": Clay for enrichment and list building, Instantly or Smartlead for sending, and Salesforge specifically because its pricing is usage based rather than per seat, which matters once you're running several client accounts through the same stack rather than one brand. Trigify is what I use for intent and social listening signals. None of that is a universal verdict, it's what fits the way I run accounts.
The metrics that separate signal from noise
These are the benchmarks I use on video, and I'd treat them as a general range rather than a hard line: a 3 to 5% reply rate on cold outbound is a strong result, converting 40 to 50% of interested replies into a booked call puts you in roughly the top 10% of teams doing this, and closing 15 to 20% of qualified opportunities in B2B is a genuinely good outcome. Show rate on booked calls is worth keeping above 90%, since a call that no shows costs you the same effort as one that closes.
The reason these numbers matter more than they look is that they're the tool that actually tests the segment blending theory from the top of this piece. If your blended reply rate sits at 2%, under benchmark, split it by segment before assuming your copy is weak. It's common to find one segment already sitting at 6%, comfortably above benchmark, dragged down by a second segment converting at near zero. That's not a copywriting problem, and no amount of subject line testing fixes it. I've written separately about how a company wide activity number can hide exactly this kind of problem in why SDR activity metrics are lying to you, if you want the fuller argument for why blended numbers mislead before you ever get to segments.
Execution gaps: the six places a company gets stuck
When the numbers still don't add up after the sections above, the gap is usually one of six things: a skill gap, where the team doesn't know how to build the automation or process a goal requires; a resource gap, missing budget, headcount, or tooling; a strategy gap, targeting the wrong market entirely; a process gap; a sales and marketing alignment gap, one of the most common I run into; or a leadership gap, missing a specific kind of operator, often technical, that the rest of the team can't substitute for. Naming which one you actually have matters more than the audit itself, because the fix for each is completely different.
The 90 day roadmap that follows the audit
Once the audit surfaces the real problem, I run it as a 90 day build, not a single fix. The first 30 days are foundation: fixing infrastructure, clarifying ICP and segments, tightening messaging so it's segment specific rather than generic, and cleaning the CRM so routing actually works. The next 30 are acceleration: rebuilding the outbound engine across the channels that fit, launching the offer, improving the demo and follow up frameworks based on what the first 30 days revealed, and standing up weekly reporting per rep so bottlenecks surface fast. The final 30 are scaling: adding channels, ramping outbound volume deliberately rather than all at once, building RevOps automation, and using the deals won and lost so far to tighten the pitch further.
| Phase | Focus | What "done" looks like |
|---|---|---|
| Days 1-30: foundation | Fix infrastructure, clarify ICP and segments, tighten messaging, clean the CRM | Every lead is categorized and routed to the right person |
| Days 31-60: acceleration | Rebuild outbound, launch the offer, refine demo and follow up, weekly reporting | Bottlenecks are visible per rep, not hidden in a blended average |
| Days 61-90: scaling | Add channels, ramp volume, build RevOps automation, tighten the pitch | Pipeline growth is deliberate, not a lucky week |
Key takeaways
- A pipeline that feels random is usually two or three ICP segments blended into one reported number, not a tactics failure.
- Split reply rate, call conversion, and close rate by segment before rewriting a single sequence.
- A secondary ICP is a fallback for when you exhaust a small primary market, not an equal priority to chase alongside it.
- If a case study doesn't genuinely fit the segment you're pitching, cut it. No proof beats the wrong proof.
- 3 to 5% reply rate, 40 to 50% interested to call conversion, and 15 to 20% close rate are the ranges worth benchmarking against.
- The audit isn't the fix. It's the input to a 90 day build: foundation, acceleration, then scaling.
FAQ
Why does my pipeline feel random even though I'm following a consistent process?
Because "consistent process" and "consistent segment" aren't the same thing. If the same sequence and cadence run across two ICPs with very different fit, the result will look inconsistent even though the process didn't change, because you're really watching two different outcomes averaged into one number.
How do I know if a segment is worth keeping as secondary rather than dropping?
Look at the cost of chasing it against what it actually returns: how many clients you can bring on, at what velocity, and how much pipeline and sales time it consumes relative to your primary ICP. A segment that closes occasionally but eats a disproportionate share of sales time is a candidate to demote, not necessarily to drop entirely.
Should I really cut a case study instead of using any social proof I have?
Yes, if it doesn't fit the segment you're pitching. A case study from the wrong industry or the wrong company size invites the prospect to find reasons your result won't apply to them. An offer that stands on its own, without proof that raises more doubt than it resolves, usually converts better.
What's the first metric I should check if I suspect segment blending?
Reply rate, split by segment rather than reported as one company wide number. It's the fastest one to isolate and the one most likely to reveal a strong segment being dragged down by a weak one.
Do I need all of this audit before I can fix anything, or can I start with one section?
You can start with ICP and segment tiers alone and get real signal fast, since it's the section most of the "randomness" traces back to. The rest of the audit, messaging, sales process, RevOps, matters, but ICP and segmentation is the one worth doing first.