We Analyzed 25,537 Sales Calls. Here’s What We Learned.

Chris Orlob by

“We need to figure out the ROI justification and which teams we’re going to roll this out to,” my prospect said with confidence. “Once we do that, I can’t think of anything stopping us from moving forward.”

Despite that seemingly strong buying signal, my heart sank.

I had just started my new position at Gong and I was asked to meet on-site with this potential customer to move them closer to signing a deal.

So why did my heart fall into my stomach instead of race with excitement upon hearing that?

Any other sales professional would have been excited to hear that phrase uttered from their prospect’s lips. It seemed like such a sure thing, on the surface.

But I had a piece of knowledge most sales leaders and professionals have never had…until now.

Two Counterintuitive Sales Forecasting Insights Revealed by Artificial Intelligence

By the end of this short post you’ll understand why the above story makes sense, despite its counterintuitive first impression.

You’ll learn why that above so-called “buying signal” actually has a negative correlation with win-rates and forecast accuracy.

You’ll also discover the other side of the same coin: how a single cautious, indecisive word uttered by your prospect can increase (yes, I said increase) forecast accuracy by 73%.

Let me explain.

Bringing Science to the B2B Sales Conversation

For the sake of context, we at Gong have built an automatic sales conference call recording platform with conversation analytics running on the back end using transcription, AI, and machine learning technologies.

We used our own platform to analyze 25,537 anonymized B2B sales calls from 17 customer organizations in search of data-driven sales conversation insights.

For the curious, here’s a quick summary of how we gathered this data:

  • As mentioned, we analyzed 25,537 B2B sales calls from 17 customers. These customers were mainly mid-market, high growth SaaS companies. All of these calls were account executive conference calls conducted on platforms like GoToMeeting, Zoom, Webex, etc., rather than SDR or prospecting calls
  • Each call recording was mapped to its corresponding CRM record. This allowed us to analyze calls against outcomes like win-rates, forecast accuracy, sales cycle length, and revenue produced
  • As the calls were recorded, they were also speaker-separated, cleaned, and transcribed from speech-to-text
  • Finally, we used Gong’s conversation analytics capabilities to analyze the calls and transcripts, auto-categorizing events within each call such as key moments, topics discussed, and seller/buyer behaviors

Among a handful of other insights, here are the two counterintuitive trends we discovered in terms of forecast accuracy signals.

Beware the Phrase “We need to figure out __________________”

Think back to the story I told at the beginning.

Remember when my “heart sank” despite what seemed like a strong buying signal?

Here’s why…

Every B2B sales professional worth his or her salt will eventually ask the “timing question.”

“When do you estimate moving forward with this project?”

“What does your timeline look like for purchase?”

“How soon do you foresee getting this agreement executed?”

It turns out when your prospect responds to your timeline question with some variation of “We need to figure out ________________,” there is an unmistakable negative correlation of getting that deal closed within your estimated forecast.

Your odds of closing that deal on time drop significantly when your prospect utters those words (or some variation of them).

Counterintuitive, but true.

Now you understand why my heart sank in the story at the beginning of this post

“Probably” Justifies “Happy Ears”

On the other side of the same coin, we also discovered a response to the timeline question that has a strong positive forecasting correlation.

The word “probably.”

In other words, when you as a sales professional ask for timeline and the prospect cautiously responds with the word “probably __________________________,” you have a much higher likelihood of closing the deal within the estimated forecast.

Again, counterintuitive but true.

After all, the word “probably” isn’t exactly strong, decisive, confident language.

Only in hindsight does it make sense: the prospect is likely responding cautiously because of how seriously they are considering the purchase.

They don’t want to get backed into a corner by a pushy sales rep with happy ears, so they verbally distance themselves.

A Sales Conversation Insight Generation Machine

We’ve only revealed the tip of the iceberg.

25,537 analyzed sales calls reveals many more insights than just the two we covered here, as I’m sure you can imagine.

There are many more AI-driven sales conversation revelations we discovered such as:

  • “Talk-to-listen” ratio trends against win-rates
  • How often pricing is discussed in winning sales conversations
  • The exact time window top performing AEs discuss pricing
  • A specific type of language that increases sales win-rates 32% on average (hint: it’s not “assumptive language”

Go here to get the report covering the full range of insights we discovered in this analysis.

Keep in mind: there are many more to come over the next months and quarters. Be sure to keep tabs on us by visiting Gong.io. And if you’re a B2B sales leader with an account executive team of at least 10 and you’re interested in seeing what Gong is all about, go here to request a Gong demo.

Senior Director of Product Marketing

  • Hi Chris,

    when one holds a senior position in product marketing, these types of ‘timing questions’ are sure to come from her lead prospect. I always too feel the heat just like you. Your article just spoke my heart out.

    Nice one!

    I really enjoyed reading it.

  • How about some old fashioned human intelligence?
    If you are well into the ‘sale’ and you are not dealing with Power, someone who can write a cheque, you’ll be wasting a lot of time on people who can’t say yes. Analyze audio bitstreams all you want, the basics are the same.

    No decision makers in sight, you’ll be stalled 85x out of a 100.

    • Hey Marty – we certainly advocate using human intelligence to sell! That’s actually the point of this research – to guide human intelligence rather than replace it. Sales will never be reduced completely to a science with this technology, just as marketing analytics didn’t completely reduce marketing to a science.

      However, technology, tools, and data like this are like an “ironman suit” for sales professionals. Neglecting to use them will leave you at a disadvantage against your better-equipped competitors.

  • My point is that the research as described didn’t reveal anything. The responses referenced are entirely consistent with selling to people who can’t say yes (have no authority), are budget (compliance) focused, and do not ‘buy value’ (things that improve corporate performance).

    Maybe I missed it, or its hidden in the data not yet provided.

    Fundamental qualification techniques would have predicted what this machine intelligence was supposedly discovering. What it was discovering is that these opportunities were not qualified for closing, that the KDMs we’re not present, and that what the KDMs might need in order to prioritize the buy above other options, had not been determined.

    Solid human intelligence would have prevented many of the conversations used to discover that a lot of time was being wasted.

    Don’t get me wrong. I love great software that can provide deep insight.
    Or as I’ve commented on before, when the AI Sales Bot doesn’t make plan, who gets fired – the sales manager bot, or the AI Programmer? 🙂

  • >> if you’re a B2B sales leader with an account executive team of at least 10 and you’re interested …

    No because of that comment no longer interested – you just disqualified yourself – what if say I’m doing 100k MRR with 2 sales execs and growth of north of 45%?

    Like a lawyer a sales person asks question a) to discover, b) to exclude

    but never asks a closed question when exploring and collating evidence

    Rookie error !

    • Thanks for expressing your opinion so forcefully James.

      The goal certainly isn’t to exclude small (yet growing) sales teams.

      The machine learning is dependent on gathering a certain amount of data to work with. Sales teams of less than 8-10 or so simply don’t generate enough call volume to discover insights tailored to them.

      • Chris – then tailor wording to goal – we are a pattern recognition startup and understand need for data volume – but reject disqualification on basis of stasis rather than change – especially given things in startup land change fast.
        I hope your response clarifies what you actually meant for your other readers.

  • When closing deals, it is so important to maintain a positive mindset. Saying that, “We need to figure out” really does indicate that there is some kind of negative obstacle in the way. The goal of any sales rep is to remain positive so that these obstacles don’t become insurmountable barriers.

  • Great insight shared Chris, We at callhippo follow the same strategy and therefore we have came up with a solution i.e callhippo.com a virtual phone system which allows small business and enterprises to get local numbers for more than 40+ countries and setup their support center in less than 3 minutes. The main features of callhippo is that it allows managers and directors to analyze and monitor the calling activity on routine basis which can directly affect in data driven strategy making and redcution of costs. It also contains variety of features like: Call Recording, Call Forwarding, IVR, Custom music, Set working hours, Make and receive calls from browser, Forward calls to your smartphones. So with callhippo you will never miss a customer call again!

  • Marah Sayaman

    This is brilliant, Chris! I’ve picked up a lot of Dos and Don’ts from and I’m excited to put what I’ve learned from this in practice soon. Again, if ever you or your readers are looking for something to revolutionize your calls, give Tenfold a try. Tenfold is a phone intelligence platform that integrates CRMs and phone systems quickly, enabling companies to enhance prospect and customer interactions. Hope you check us out! https://www.tenfold.com/tenfold-explainervideo