Sales forecasting is a pain for most B2B sales managers. They are expected to produce accurate forecasts, but often lack the large sales volumes required for statistical forecasting. As a result, most rely on subjective appraisals of their sales pipelines: What do they contain today, and how does this compare to past experience?
While such comparisons are more art than science, they do leverage a solid insight: Designing, monitoring, and analyzing the sales process are key to improving B2B sales forecasts. Formalizing the link between the sales pipeline and its outcome is thus an enticing challenge for B2B companies.
Mathematically, we are looking for function f in the following equation:
sales forecast = f (sales pipeline)
Breaking up the sales pipeline
"Sales pipeline" is a weak input for a potential equation—we must break it into more specific parts. It is good methodology, and will also increase the data available for analysis. For example, SalesClic research shows that looking at pipeline transitions (opportunities moving from one pipeline stage to another) instead of just pipeline outcome (closed opportunities) increases data volume by a factor of three.
Three dimensions of the sales pipeline are relevant.
1. Attributes of the opportunities. The sales pipeline is (hopefully) full of opportunities, all of which have attributes that affect the outcome of the sales process. Product line, client type, and deal size are obvious candidates, but the list is potentially infinite.
2. Attributes of the sales process. The structure of the sales process influences forecast accuracy: The number of pipeline stages, their duration, the corresponding tasks, and participants are decisive inputs in our equation. A transactional process in one or two stages, concluded in a few weeks and involving one decision maker, will be easier to model than a more complex process.
3. Attributes of the salesperson. Finally, the salesperson actually running the process will influence its outcome. We are looking here for indicators of competence and effort. They are notoriously difficult to isolate, and most companies rely on proxies, such as activity volume (e.g., number of calls) or quota attainment.
So function f would ideally include:
- A minimum of three to four opportunity attributes
- Three to four factors summarizing sales process structure
- Three to four proxies for salesperson attributes
We would then regress function f on past sales process data, identify the statistically significant factors and their coefficients, and use the calibrated function for forecasting purposes.
That is serious data crunching, with equally serious limitations:
- It may be possible for companies with vast consulting and computing resources, but for most sales managers, it is a pipe(line) dream.
- There is a trade-off between sophistication and robustness. The more sophisticated the model, the more sensitive it is to changes in the company's environment. Function f as imagined above would require frequent recalibration. We can even doubt it would work above salesperson level in a given company.
- Producing accurate forecasts is not enough. We also want salespeople to understand the corresponding calculations. "Black box" forecasts would be perceived as just another enforcement tool.
So let us resist the golden algorithm fallacy, and look for practical (and rational) ways to improve sales forecast accuracy.
My company suggests five simple tips to improve B2B sales forecasting. They can be implemented easily in the context of traditional CRM software.
1. Refine your sales process. The more realistic the sales process, the easier to determine its outcome. Realistic in this context means: "reflecting the decision process of the buyer." Measuring the dynamics (e.g., conversion rates, time to wins, durations) of a well-designed pipeline will considerably improve your understanding of its true potential.
2. Explain your assumptions. It is inevitable and advisable for B2B sales forecasts to include a judgmental element: Your salespeople may be biased, but they also possess incomparable market knowledge. Decision theory shows that writing a one-sentence explanation for such judgments increases their reliability significantly.
3. Build detailed sales simulations. Especially in the presence of "elephants" in your pipeline (opportunities so large they could make or break your forecast), going beyond the traditional best-case and worst-case scenarios and building detailed, opportunity-by-opportunity simulations is a great help.
4. Use multiple forecasting methodologies. One of the most consistent results of forecasting research over the past 25 years is that averaging three to four simple forecasts is often better than relying on a single, more sophisticated methodology. We would advise a combination of "enlightened judgments" (see Explain your assumptions, above) and nimble calculations based on the historical data in your CRM database.
5. Monitor forecast errors. Incredible but true: Most B2B companies strive to produce more reliable sales forecasts but do not monitor forecast errors rigorously. Since closing dates are so important in a B2B context, we would recommend measuring forecast errors opportunity by opportunity, for the opportunity amount and closing date. That should give you useful indications on how to improve forecast accuracy.
Thomas Oriol is a founder and director at Nimble Apps Limited, the Dublin-based publisher of SalesClic, a sales pipeline visualization, analysis, and forecasting solution for B2B companies. Prior to founding Nimble Apps, he spent 10 years advising European technology companies on mergers and acquisitions.