For one, a likely addition to the team needs to be business or revenue operations staffed with one or two, analytically minded folks that understand both, modern marketing techniques as well as sales psychology. They can build all the various models and / or implement some of the described monitoring tools.

Second, if the needed cultural transition to transparency in reporting, commitment to analysis of facts and data no matter what the conclusions, and an emphasis on accurate sales and marketing data entry is not supported or understood by upper management, then don’t even try. Ultimately the risk is they will find it an annoying slowdown at best or a threat to their authority, at worst. The crucially important point to make here is that there is cultural acceptance of scientific forecasting – as a complement to “gut feel” management (which also has its place). And that there is an environment where sales and marketing leaders are not punished for projecting pipeline or sales shortfalls. Fact-based and data driven business analysis is a company’s equivalent of a country’s free press. If they’re not allowed to flourish, find a new home if you’re one that does trust facts and data but doesn’t get listened to, or worse, is on the outside of the political currents driving the company because of it.

Another key group of stakeholders that need to be brought along typically is the board of directors. At one company, once we had built out a robust analytics and reporting functionality that was all based on real-time cloud applications like Marketo, Salesforce, InsightSquared, Salesloft, Bizible, and the like, we gave them login credentials. They loved it because if they wanted to see where the pipeline stood for the quarter, or what was coming in for next quarter at 3 am on a Sunday morning, they could see it. Real time. We had had the conversation about not then using that data to micro-manage us (we had seen it already at 2 am that Sunday morning :), that transparency did a lot to instill credibility in our forecasts.

Disagree with our theses? Please send us your comments.

In the meantime, for some further reading, here are links to some useful articles that further elaborate what investors are looking for beyond just growth but also around predictability:

• “Venture Capital Investors Like Revenue, But They Love Predictable Revenue”
• “The One Thing VCs Love to See Even More Than Growth”

The post Scientific Forecasting – 4 of 4: Get buy-in appeared first on Premonio.

1. Different types of revenue forecasts:

In terms of forecasting techniques, we’ve often selected from three principal choices:

• Capacity vs. demand forecast
  • A common way to forecast, esp. for more than one quarter out, is to model the number of hired reps, with a hiring ramp if the model is more sophisticated, apply an average closure rate, and then use that for next year’s forecast. Technically that’s not a forecast, it’s a sales capacity model (i.e. “if we had X no. of reps and they closed $Y at closure rate Z%, then we should be making $XYZ next year”). If the pipeline velocity (how fast can they close deals?), the average selling prices (ASPs), and the average conversion and closure rates are reasonably accurate, a model like this can provide a sense of what’s possible to be sold next year.
  • A true demand forecast would be modeling the number of incoming leads, and then extrapolating from there to the eventual closed-won bookings, e.g. using a cohort model.
• On-off vs. weighted average forecasts:
  • Whether forecasting is done via spreadsheets (e.g. most sales leaders maintain a simple spreadsheet to help them track what’s out there, and also to help them sanity check what’s coming back from their business operations teams) or via applications like Salesforce or InsightSquared, an analytics overlay on top of Salesforce, there are two principal ways to calculate a forecast:
    • An “on-off method” where all possible deals are listed and their amounts, and the analyst “turns deals on or off”, i.e. totals the sum the “on” deals’ bookings. If, say, there are 90 deals in the pipeline, in this method the, say, 23 that are expected to close are added in their likely, total amounts, the resulting sum then is the forecast. As a rule of thumb, this is typically a more realistic approach for a limited number of high-priced deals.
    • Or a “weighted average method” where a likelihood of success in % is multiplied with each deal’s likely value, and the resulting multiplications are added up to a “weighted average total forecast” estimate. I.e. in this example, all 90 deals would be multiplied by their respective success probabilities and then those calculated, total expected values would all be added up together. This approach is typically better when there are many small, transactional deals.
    • This allows for some interesting comparisons: If, for example, sales management’s on-off method forecast is higher than the weighted average method calculated by a tool like, say, InsightSquared, then sales management is more optimistic about the close rate potential than the historical averages that InsightSquared uses to calculate the expected forecast. The reverse could be true, too. In any case, comparing the two methods’ forecasts can lead to interesting conversations to pressure test the assumptions behind the forecast.
  • The above two, principal techniques can be further refined by applying variations to the conversion and closure rate probabilities by sales stage, by lead source (e.g. inbound web leads or referrals typically close at higher rates than outbound, cold-call leads), by individual rep probabilities (different reps close at different rates), or by price bands (larger deals typically have lower closure rates than smaller deals).
• Gut vs. data:
  • As precise as modern forecasting methods based on tools like Salesforce, InsightSquared, or a sophisticated assembly of spreadsheets maintained by the revenue or business operations team are, they’re only as good as the data that goes into them. If, for example, the sales force does not maintain a good record of lead sources in Salesforce, then any correlations of closed-won rates with lead source types will be wrong. Or if the sales stages indicated for opportunities are overly optimistic or pessimistic, applying sales stage-based closure rate probabilities will result in biased forecasts.
  • To counteract those biases, it’s always helpful to get on the phone or send some validating emails to spot check individual opportunities and the accuracy of their data and descriptions. Prospects or reps, when spoken to in person and in private, often provide a level of insight of what’s going on in their marketplaces that the pure data can’t all convey.

2. Lengths of forecasting periods:

We talked about investors’ interest in predictable cashflows, and thus predictable revenue, and ideally, they’d like to have a rolling six quarter forecast that’s pretty accurate, and that also gets updated early enough if / when circumstances change. For example, a large deal gets pushed a year, macroeconomic conditions change, a major competitor is launching a predatory campaign against your company, or a new startup threatens your installed base. Forecasts typically get calculated for three time horizons, utilizing different approaches for each time horizon:

• In quarter forecasts:
  • Tools like Salesforce or InsightSquared (or others) handle this very well; and you can select from on-off or weighted-average options.
  • Or the oft mentioned, tried-and-true spreadsheets’ primary focus typically is on the ongoing quarter, as well.
• Two quarters out:
  • InsightSquared has several multi-quarter pipeline forecasting options that we have found to be quite accurate (assuming the data entered into the CRM system is accurate, of course).
  • If you do have a cohort model (i.e. can track how different lead source cohorts propagate through the pipeline over time), you can essentially reverse its calculational direction and estimate how much closed-won business you are likely to generate by extrapolating from incoming pipeline information in combination with known sales velocities and closure rates.
• Three to six quarters out:
  • For longer term forecasting, top down, market size models are an option. Top down refers to the process of starting with an overall market size (usually a “TAM” or total available market), and then breaking that down into smaller and smaller chunks that describe the “SAM” (i.e. the served available market). Breakdowns might be by major geography, by target customer size, by price point (i.e. enterprise deals vs. small, medium size business or SMB deals), or by vertical (say, finance and IT).
  • A second, longer term method is sales capacity modeling, hopefully layering in a hiring ramp, that estimates how much revenue could be booked by a sales force of a certain size by when. Marketing pipeline generating capacity typically gets estimated similarly but by calculating how much marketing spend would translate into how much $ pipeline.
  • A simple example illustrates this: If “marketing yield” (i.e. amount of pipeline generated per $1 of program spend) is $10, then a $1M marketing budget should generate $10M in pipeline. If in turn, sales reps, say, generate 33% of their bookings themselves and close 25% of all $ opportunities, then that $10M marketing pipeline should result in $2.5M in closed won from marketing sources, and another $1.25M from sales “self gen” (i.e. bookings they generated). The overall sales and marketing spend efficiency could then also be calculated once personnel costs are known.

Don’t turn that dial, we’ll be back with more …

The post Scientific Forecasting – 3 of 4: More how to’s appeared first on Premonio.

1. Overcoming common cultural challenges to scientific forecasting:

As more and more leads get generated from online behavior, which now can be tracked down to individuals’ browsing patterns, the amount of data that needs to be crunched to correctly attribute the origins of a pipeline is ever growing. And with it comes both, a greater burden on the analyst to be familiar with at least basic data collection and reduction techniques (e.g. multi-variate regression analysis, cohort analysis, basic probability theory, etc.; all available for study online), as well as a commitment to adapt and revise demand generation strategy based on what the data says.

Who hasn’t seen the classic battles raging over whether sales or marketing “sourced” a lead? As if that matters nowadays. It’s like trying to decide which shot won the basketball game. Did the prospect turn into a win because s/he read a white paper, came to the website, talked to an SDR, come to the booth, or followed the referral provided by a colleague? The answer is yes to all the above. The trick is to know which of these are the most efficient and success-prone channels, not who “sourced” them, so you know where to invest more and where less. To mitigate this issue, we now ask who “captured the lead”?

Which means there need to be forecasts across the entire pipeline – from customer success to marketing to sales. However, while marketing forecasts are in terms of leads, sales and customer success forecasts are in terms of $ pipeline or numbers of expected deals. For there to be a reliable model, there needs to be a good understanding of the analytics underlying a company’s lead lifecycle and specific pipeline conversion metrics.

Another, closely related issue is the need for management to support the conclusions of the data analysis, even if they head in a direction not aligned with prior thinking. A recent CEO we worked with was so convinced of the rectitude of his target market size instincts that any evidence to the contrary was being questioned to death, and the business analysts “punished” with oppressive data requests to justify their alternative conclusions. At some point his team just reported to him what he wanted to hear, and net new business pipeline generation continued to produce at 50% of the needed goal.

A variation on that theme can originate from the head of sales, where the implicit, old school message to the reps issuing their forecasts is that they better “make my number”, i.e. personally commit to a forecast. Vs. a more modern, analytics-driven approach to forecasting where they’re just reporting the state of their pipeline, say on a weekly basis, without pressure to generate a guaranteed outcome. Once SDRs and sales reps have begun to trust their management, they will report what’s in their pipelines so it can be made visible; and if need be, more pipeline generation initiatives can be launched. Vs. the old school approach where reps tend to wait with reporting what’s in their pipelines until a week or two before quarter end to not get shot over forecasted deals that didn’t pan out (see sidebar story below).

1. 1. :
   2. In-quarter corrective actions:

Guess with which of the above approaches is there more predictability and more of an opportunity to counter-steer early in the quarter if pipeline generation is below goal? The modern, analytics-enabled approach also allows the forecaster to normalize between different sales reps’ innate forecasting tendencies; some are optimists prone to over-forecasting while others are more cautious and prefer to “sandbag” their numbers. In the end it all settles out into whatever the closed-won number is, but during the quarter as one tries to predict the outcome, compensating for reps’ own forecasting tendencies increases the overall forecasting accuracy.

Knowing how to compensate for forecasting biases and have a sense, say 4 or 5 weeks into the quarter how it will turn out, allows sales and marketing management to counter-steer (see sidebar story). Here are some practical, rapid response options for in-quarter corrections:

1. 1. • Supporting sales pipeline building:
        • “Air strikes” – Reps can call in “air strikes”, i.e. help from upper management to help close iffy deals; e.g. the head of products or the CEO personally call a prospect to address their concerns. This can improve the “likelihood to close” of unsure deals.
        • “Local pipeline generation” – Marketing can utilize their demand generation infrastructure to drive leads to a local sales rep whose pipeline may be below goal. A call to action might be to attend a local event, and this allows driving in-territory brand awareness and introduce the local sales rep to prospects in his / her area.
      • Supporting marketing pipe generation:
        • Typically, this exercise for marketing revolves around which leads can be spooled up quickly? One example is “local pipeline generation” mentioned above. A challenge of in-quarter corrective action can be in terms of time: Some demand generation campaigns can take a few weeks to set up and be productive – so, unless there is visibility in the quarter, the in-quarter impact can be limited.
        • Another approach involves digging into the various pipelines and see which ones have rapid pipeline velocity (i.e. move to the next higher stage quicker than others). We’ve found that SEM campaigns can spool up quickly, as can additional nurturing emails around an interesting piece of content or a local event that prospects may be interested in. Also, improving the conversion and click-through rates of high-trafficked website pages is a relatively quick way of generating leads.

Don’t turn that dial, we’ll be back with more …

The post Scientific Forecasting – 2 of 4: How to’s appeared first on Premonio.

Well, here’s a common reality: While often significant amounts of time, energy and precision go into the accuracy of the expense forecast (aka “the budget”), on the revenue side the norm with many startups more often than not is to simply extrapolate growth with usually just a basic spreadsheet or “back of the envelope” calculations. Based on those “estimates”, executive teams often then commit growth plans to their boards (and with them EBIDTA and cashflow projections, profitability metrics, etc.) that ultimately not only drive department budgets, hiring plans, etc., but also their own compensation. All based on “rough estimates”. What could possibly go wrong, you say, right? Read the box below for one story of what did go wrong.

But before diving deeper into what’s at stake with good or bad forecasting, first a bit of theory: When you calculate a number (let’s call it P for Profit) by subtracting a highly precise number with a low uncertainty band (call it E for Expenses; say +/- 1%) from a highly imprecise number with a high uncertainty band (call it R for Revenue; say +/- 10%), the uncertainty band of the calculated number P is greater than the worst of the two uncertainties, i.e. the +/- 10%.

Here is an illustrative example: Say a start-up is expected to lose $2M a year on expected revenues of $25M, or -8%. Now, if the revenue uncertainty is +/-10%, then the year could end at $22.5M or at $27.5M. I.e. the expected profits might range from a loss of $4M to breaking even, depending on the how expenses scale with revenues. So, in this example, 10% revenue uncertainty could translate into a profitability range of -18% to breaking even. Now, how many investors would be willing to pay a premium for the resulting profit or cashflow prediction that has an almost +/- 100% uncertainty around it?

But instead of improving the precision of revenue forecasts, it’s too common to instead rely on adjusting expense budgets, typically quarterly and sometimes more often, to counteract the fluctuations in the actual revenues compared to their original, imprecise forecasts. The resulting rituals of quarterly expense adjustments, re-budgeting, vendor trimming, or even layoffs, etc. aim to ensure the expense lines track the undulating non-attainment of the revenue forecasts. The trouble is that while revenue can almost immediately go down, this common approach to managing the bottom line can cause problems because many expenses cannot be ramped down that fast – even if they would need to be adjusted after a revenue decline, it may not be possible due to existing expense commitments.

In companies not knowing how to precisely manage their forecasts, we’ve seen these expense budget re-adjustments literally consume 10% to 20% of affected departments’ time (often it’s the marketing department as they tend to have the highest non-personnel budgets). That’s wasted time with zero productivity gain (e.g. reduced pipeline from marketing), not to mention the resulting frustrations and morale hits. And to add insult to injury, all too often the hapless “expense victims” get blamed for having “runaway spending”, when the real culprits are companies’ inability to precisely forecast revenue.

Alright, I get why forecasting precisely is important, so what do I do about it?

Many organizations think they know how to forecast growth, but at the end of the day, their forecasts often really neither aren’t all that precise nor accurate. For one, many folks we talk to confuse “sales forecasts” with “forecasting sales”.

The former is what the head of sales puts together to manage, typically this quarter’s sales expectations. Which is different from the latter, which ideally is a rolling 2 to 6 quarter revenue and bookings forecast for the company that’s the basis of its financial projections and budget allocations, especially for marketing departments who have to have a longer planning horizon as some expense commitments get made a year out or more. As we’ve discussed, the latter, especially when done precisely, has a lot of upsides (investors pay a premium, more stable budgets) and avoids downsides (accurate and meaningful analytics, more stable organization, higher morale).

From years in the trenches and having made a living by pegging my own compensation to the accuracy of my forecasts, in the next three blogs we’ll publish tips and how-to’s on how to improve forecasting precision (and thus accuracy) by applying some basic common sense and scientific principles to forecasting.

Don’t turn that dial, we’ll be back with more …

The post Scientific Forecasting – 1 of 4: Why is it important? appeared first on Premonio.